Construction method of training data set of electric power meteorological large model and related device

Abstract

The invention belongs to the technical field of electric power meteorology forecast, and discloses a construction method of an electric power meteorology large model training data set and a related device. The method comprises the following steps: acquiring data for generating a large power meteorological model training data set; the data comprises global meteorological original reanalysis data, static data and multi-source heterogeneous electric power meteorological historical multi-year observation data; constructing an analysis model for generating a training data set; inputting the data of the generated electric power meteorological large model training data set into an analysis model for analysis and simulation to obtain electric power meteorological reanalysis data; and forming an electric power meteorology large model training data set based on the electric power meteorology reanalysis data. According to the invention, the technical problems of insufficient industry specific requirements and insufficient coverage depth of the training data set of the existing electric power meteorological large model are solved, and the practicability of the electric power meteorological large model is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power meteorological forecasting, and in particular relates to a method for constructing an electric power meteorological large model training data set and related devices. Background Art

[0002] In recent years, breakthroughs in artificial intelligence (AI) technology have driven the rapid development of large-scale meteorological models. Using deep learning techniques, these models can extract complex atmospheric motion patterns from massive amounts of meteorological data, significantly improving the accuracy and efficiency of weather forecasts. They have become a key technology supporting renewable energy power forecasting, disaster warning, and resource scheduling. However, the performance of these models is highly dependent on the quality, coverage, and applicability of the training data. Currently, mainstream large-scale meteorological models (such as Pangu-Weather and GraphCast) use ERA5 (ECMWF ReAnalysis version 5) reanalysis data as their training datasets.

[0003] The ERA5 reanalysis dataset, developed by the European Center for Medium-Range Weather Forecasts (ECMWF), is currently the most widely used meteorological reanalysis dataset worldwide. Using Four Dimensional Variational (4D-Var) technology, ERA5 integrates data from multiple sources, including satellite, sounding, and ground-based observations, to produce a high-resolution (horizontal resolution of 0.25°×0.25°, 137 vertical layers) reanalysis product covering global atmospheric, oceanic, and surface variables from 1979 to the present. ERA5 provides over 60 meteorological variables, including temperature, humidity, wind speed, and radiation. Its high physical consistency and strong spatiotemporal continuity make it a benchmark dataset for training large-scale meteorological models.

[0004] However, as a general-purpose reanalysis data, ERA5 focuses on global climate research and numerical model verification, and is not optimized for the refined needs of specific industries such as electricity. Directly using ERA5 data as a training set for large-scale power meteorological models has the following shortcomings: First, it lacks industry-specific needs. As a global data set, ERA5 cannot meet the power industry's needs for high-precision, specific regions, or specific meteorological variables. The resolution of ERA5 data makes it difficult to depict the micro-scale meteorological processes that are of greatest concern to new energy sites and power equipment, such as the wake effect of wind farms or extreme winds on power lines. Second, the data coverage depth is insufficient. ERA5 cannot accurately cover the key variables of power applications, and the assimilation process does not include field observation data for power meteorology, such as wind and light measurement data from new energy sites and online micro-meteorological monitoring data from transmission lines. This seriously restricts the application level of large-scale models in the power industry. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and related devices for constructing a large-scale power meteorological model training data set to solve the technical problems of insufficient industry-specific requirements and insufficient coverage depth of existing large-scale power meteorological model training data sets, thereby improving the practicality of the large-scale power meteorological model.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for constructing a large-scale electric power meteorological model training data set, comprising: Acquire data for generating a large-scale power meteorological model training data set; the data includes global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data for many years; Constructing an analysis model for generating a training data set; inputting the data for generating the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data; A large power meteorological model training dataset is formed based on the power meteorological reanalysis data.

[0007] A further improvement of the present invention is that the step of obtaining data for generating a large power meteorological model training data set specifically includes: Obtain global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data for many years, and preprocess them to obtain data for generating a large power meteorological model training data set.

[0008] A further improvement of the present invention is that the step of performing pretreatment specifically includes: Preprocess the global meteorological raw reanalysis data and convert the format into the universal binary format GRIB; Preprocess the static data to make the spatial resolution and projection coordinates of the static data consistent with the analysis model; The multi-source heterogeneous historical observation data of electric power meteorology are preprocessed and converted into the format required for assimilation of the reanalysis model.

[0009] A further improvement of the present invention is that: in the step of obtaining global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data, the European center's ERA5 (ECMWF ReAnalysis version 5) data, the United States' FNL (Final Operational Global Analysis) data, and China's CMA-RA (China Meterorlogical Administration ReAnalysis) data are collected as global meteorological original reanalysis data; Obtain one of SRTM (Shuttle Radar Topography Mission), GTOPO (Global Digital Elevation Model), GlobCover (Global Land Cover Map), and MODIS (Moderate-resolution Imaging Spectroradiometer) as static data; Multi-source heterogeneous power meteorological historical observation data include: data from ground meteorological stations, sounding stations, weather radars, meteorological satellite data, wind and light measurement data from new energy sites, online micrometeorological monitoring data from transmission lines, and wind measurement radars and wind profiler data for power grid facilities and equipment.

[0010] A further improvement of the present invention is that the steps of constructing an analysis model for generating a training data set and inputting the data for generating the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data specifically include: Obtain the configuration of the boundary layer scheme, cloud microphysics scheme, land surface process scheme, and simulation integration time step of the analytical model; Construct a power meteorological gain matrix of spatiotemporal four-dimensional approximation to obtain continuous analysis windows and cycle strategies; Based on the acquired analysis model's boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and circulation strategy, as well as the constructed power meteorological gain matrix of empty four-dimensional approximation, analysis and simulation are carried out to obtain power meteorological reanalysis data.

[0011] A further improvement of the present invention is that: in the steps of obtaining the boundary layer scheme, cloud microphysics scheme, land surface process scheme and configuration of the simulation integration time step of the analysis model, the boundary layer scheme adopts the MYNN (Myer-Nakanishiand Niino) scheme; The cloud microphysics solution uses WDM6 (WRF Double-Moment 6-class) or Morrison 2-Moment solution; The land surface process scheme adopts Noah LSM (Noah Land Surface Model) or CLM4 (CommunityLand Model version 4); The simulation integration time step was 18–30 s.

[0012] A further improvement of the present invention is that the steps of constructing a spatiotemporal four-dimensional approximation power meteorological gain matrix and obtaining a continuous analysis window and a cycle strategy specifically include: The constructed power meteorological gain matrix is: (1) in, Represents the model state, are the physical and dynamic terms of the original analysis model, is the pre-processed power meteorological observation value, is the power meteorological gain matrix; The duration of the continuous analysis window is 3 hours; the cyclic strategy adopts a sliding window method, and the simulation state is saved at the end of each window as the initial value of the next round of numerical integration.

[0013] A further improvement of the present invention is that: the boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and cycle strategy based on the acquired analysis model, and the constructed empty four-dimensional approximation power meteorological gain matrix are used to carry out analysis and simulation, and in the step of obtaining power meteorological reanalysis data, the analysis and simulation task is divided into several cycles according to the cold and hot start of the analysis model, each cycle is determined to be one day, the analysis model is cold started, and the sliding window within each cycle is set to the hot start of the analysis model, and each cycle is regarded as a complete simulation analysis; each cycle is simulated separately, or multiple cycles are simulated in parallel.

[0014] A further improvement of the present invention is that the step of forming a large power meteorological model training data set based on the power meteorological reanalysis data specifically includes: Conduct simulation effect verification and deviation correction on power meteorological reanalysis data to obtain corrected power meteorological large model training data; The corrected large-scale electric power meteorological model training data is segmented and stored according to the preset data storage format, analysis period naming and data description to obtain the large-scale electric power meteorological model training data set.

[0015] In a second aspect, the present invention provides a device for constructing a large-scale electric power meteorological model training data set, comprising: An acquisition module is used to acquire data for generating a large-scale power meteorological model training data set; the data includes global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data for many years; An analysis module is used to construct an analysis model for generating a training data set; input the data for generating the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data; The generation module is used to form a large power meteorological model training data set based on the power meteorological reanalysis data.

[0016] In a third aspect, the present invention provides an electronic device comprising a processor and a memory, wherein the processor is configured to execute a computer program stored in the memory to implement the method for constructing a large-scale electric power meteorological model training data set.

[0017] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the method for constructing a large-scale power meteorological model training data set is implemented.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for constructing a training dataset for a large-scale power meteorology model, comprising: obtaining data for generating the training dataset; constructing an analysis model for generating the training dataset; inputting the data for generating the training dataset into the analysis model for analysis and simulation to obtain power meteorology reanalysis data; and forming the training dataset for the large-scale power meteorology model based on the reanalysis data. The present invention utilizes a multi-source fusion analysis method based on globally released original meteorological reanalysis data and combines it with real-time monitoring data to construct a high-resolution (up to 3 km) analysis field that balances global consistency with regional detail. This method accurately reproduces meteorological variables of key interest in power applications, such as regional localized wind fields, extreme gales, and wake effects. This method addresses the technical issues of existing training datasets for large-scale power meteorology models, which lack industry-specific requirements and insufficient coverage depth, thereby improving the practicality of large-scale power meteorology models. The present invention deepens the data coverage of key variables by directly integrating real-time monitoring data of new energy sites (such as wind speed from wind turbine SCADA) and online micrometeorological data of transmission lines (such as wind speed, temperature, and humidity) into the numerical model analysis process, thereby strengthening the observation coverage of key variables. This ensures that the resulting analysis field not only has temporal and spatial consistency, but also truly reflects the microscale meteorological characteristics around power facilities, making up for the shortcomings of traditional global meteorological original reanalysis data (ERA5 data) that are difficult to meet local needs due to the macroscopic sampling targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings: Figure 1 A flowchart of a method for constructing a large-scale power meteorological model training data set is provided for an embodiment of the present invention; Figure 2 A detailed flowchart of a method for constructing a large-scale power meteorological model training data set provided by an embodiment of the present invention; Figure 3 A flowchart of a method for constructing a large-scale power meteorological model training data set provided by another embodiment of the present invention; Figure 4 A schematic structural diagram of a device for constructing a large-scale electric power meteorological model training data set is provided in accordance with an embodiment of the present invention; Figure 5 The figure is a schematic structural diagram of an electronic device of the present invention. DETAILED DESCRIPTION

[0020] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other.

[0021] The following detailed description is an exemplary description and is intended to provide further detailed description of the present invention. Unless otherwise indicated, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention.

[0022] The present invention proposes a method for constructing a large-scale power meteorological model training data set, which can be divided into three major steps: First, basic input data is collected and preprocessed for the generation of the training dataset, the output of which will serve as the input for the analytical model built in the second step; Secondly, customized configuration and optimization for power meteorological needs are carried out from different perspectives, and an analytical model is constructed to generate a training data set for the large power meteorological model. This analytical model uses the data generated in the first stage as input. After analysis and simulation, the generated simulated data will serve as the input for the third stage. Finally, after testing the effectiveness and correcting the deviations of the simulation data generated in the second link, a long-term, multi-year historical power meteorological model training data set will be constructed.

[0023] The specific steps are as follows: Step S1 is the first link, which completes the collection and preprocessing of input data; Steps S2 to S4 belong to the second link, which respectively constructs an analysis model for generating a training data set from different levels, including parameter configuration of the analysis model and physical scheme optimization, key parameters and sliding window configuration for continuous absorption of observation data, and segmented simulation and automated configuration for long-term historical data analysis; Steps S5 and S6 are the third link, in which Step S5 evaluates the effect and corrects the deviation of the simulation data generated in the second link, and based on this, Step S6 will eventually generate a large-scale power meteorological model training data set with many years of history.

[0024] See also Figure 1 As shown, an embodiment of the present invention provides a method for constructing a large-scale electric power meteorological model training data set, comprising the following steps: S1: Acquire and preprocess the various data required to generate a large model training dataset. This data includes global meteorological raw reanalysis data, high-resolution simulated static data, and multi-source power meteorological observation data. Preprocessing includes data formatting, projection, and quality control.

[0025] S2: Obtain parameter configurations and dynamic and physical process schemes for analytical models tailored to power meteorology needs. Power meteorology requires high accuracy in atmospheric boundary layer characteristics, turbulence, severe convection, and precipitation. We propose selecting the optimal numerical model configuration and physical scheme combination based on spatial resolution and regional climate characteristics.

[0026] S3: Acquire key parameters and sliding window configurations for the analysis model to continuously assimilate multi-source power and meteorological observation data. This paper uses a spatiotemporal four-dimensional approximation method to assimilate power and meteorological observation data into a numerical model, and proposes assimilation parameters, a continuous assimilation window, and a loop strategy to ensure the quality of power and meteorological reanalysis data.

[0027] S4: Based on the simulation scheme of the analysis model set in steps S2 and S3, the analysis and simulation of long-term historical data are carried out to obtain power meteorological reanalysis data. A strategy based on cold and hot start long-term historical simulation tasks is proposed, which includes segmented simulation, automated configuration and operation scripts, and multi-task parallel submission of simulation tasks, effectively shortening the duration of long-term historical simulation tasks.

[0028] S5: Perform result verification and bias correction on the power meteorological reanalysis data obtained in step S4 to obtain corrected power meteorological large model training data. Simulation effect verification and bias correction are performed on the power meteorological reanalysis data produced in the previous steps to further improve the accuracy and practicality of the power meteorological reanalysis data.

[0029] S6: Based on the corrected power meteorology model training data obtained in step S5, a training dataset for the power meteorology model is generated. The storage format for long-term historical simulation data, the naming of reanalysis periods, output elements, and data descriptions are determined. This generates a reanalysis dataset with a time resolution of one hour and a cumulative duration of nearly 40 years to support the training of the power meteorology model.

[0030] In one embodiment, see Figure 2 As shown, step S1 specifically includes: S11: Obtain global meteorological original reanalysis data released globally. Currently, the mainstream global meteorological original reanalysis data in the world mainly include the European Center's ERA5 data, the United States' FNL data, and China's CMA-RA data; In a specific implementation method, these three types of global meteorological original reanalysis data are collected separately, spatially covering the northern hemisphere of my country and temporally accumulating historical data of the past 30 years.

[0031] S12: Obtain high-resolution static data (topography and other static data) required for high-resolution simulation. High spatial resolution is a primary characteristic of specialized datasets used to train large-scale power and meteorological forecasting models. To more accurately construct numerical simulation systems for training datasets, high-resolution static data such as topography and geomorphology, such as those from SRTM, GTOPO, GlobCover, and MODIS, are required.

[0032] In a specific implementation, the embodiment of the present invention obtains one of SRTM, GTOPO, GlobCover, and MODIS.

[0033] S13: Acquire multi-source, heterogeneous historical power meteorological observation data over many years. Power meteorological observation data is one of the key input sources for constructing the training dataset for large-scale power meteorological forecasting models, ensuring sufficient accuracy and spatiotemporal continuity of observation fields during the reanalysis process. Multi-source, heterogeneous power meteorological observation data includes: conventional ground-based meteorological station and sounding station data; unconventional weather radar and meteorological satellite data; wind and light measurement data from renewable energy stations; online micrometeorological monitoring data from transmission lines; and various types of wind radar and wind profiler data for power grid facilities and equipment.

[0034] S14: Preprocessing of various collected input data. First, preprocessing of global meteorological raw reanalysis data, converting its format to the General Regularly-distributed Information in Binary (GRIB) format, which is readable by the analysis model. Second, preprocessing of static data, ensuring that its spatial resolution and projection coordinates are consistent with those of the analysis model. Third, preprocessing of multi-source power meteorological observation data, including quality control: removing missing, duplicate, or unreasonable observation data, and converting multi-source power meteorological observation data to the format required for assimilation in the reanalysis model.

[0035] In a specific embodiment, the multi-source electric power meteorological observation data is converted into the format required for assimilation of the reanalysis model, which is Little_R or OBSGPS.

[0036] In one embodiment, see Figure 2 As shown, step S2 specifically includes: S21: Optimizing the configuration of the boundary layer scheme for the analytical model. Common boundary layer parameterization schemes include over ten, including YSU, MYJ, QNSE, the Mellor-Yamada-Nakanishi-Niino scheme (MYNN), ACM2, and MRF. High resolution is particularly important for power meteorological forecasting, so this paper selects the MYNN scheme, which provides turbulent kinetic energy (TKE) forecasting.

[0037] The MYNN scheme, or Mellor-Yamada-Nakanishi-Niino scheme, is a widely used boundary layer parameterization scheme in atmospheric numerical models. Originally developed for the Mellor-Yamada-Nakanishi-Niino-Eddy Diffusivity-Mass Flux (MYNN-EDMF) scheme, it has been used in the U.S. National Oceanic and Atmospheric Administration (NOAA) Operational Rapid Update (RAP) and High Resolution Rapid Update (HRRR) forecast systems since 2014.

[0038] S22: Optimizing the configuration of the cloud microphysics scheme for the analytical model. Common cloud microphysics parameterization schemes include nearly 40, including WSM3, WSM5, Eta, Thompson, Milbrandt 2-mom, WDM5, WDM6, Morrison 2-mom, P3, and Jensen ISHMAEL. Power meteorological forecasting is of great interest to microscale processes such as precipitation and strong winds. Therefore, this paper selects WDM6 or Morrison 2-Moment schemes, which provide multivariable cloud water and rainwater forecasting and are more suitable for fine-grid simulations.

[0039] S23: Optimize the configuration of the model's land surface process scheme. Common land surface process parameterization schemes include approximately 10, including NoahLSM, RUC LSM, Noah-MP, CLM4, and Pleim-Xiu. This paper selects Noah LSM or CLM4, which help more accurately simulate soil thermal and moisture changes, improving the simulation of ground temperature and humidity.

[0040] S24: Optimize the time step for the simulation integration of the analytical model. Due to the constraints of numerical simulation stability, the time step for the simulation integration is generally 6 to 10 times the spatial grid size. The present invention intends to construct a training dataset with a spatial resolution of 3 km × 3 km, so the integration time step is preferably within the range of 18-30 seconds.

[0041] S25: Iterative testing of power meteorological reanalysis simulation experiments. Based on the preliminary configuration of the boundary layer scheme, cloud microphysics scheme, land surface process scheme, and simulation integration time step in the previous steps, small-scale simulation experiments are carried out and iterative testing is carried out to ensure the stability of the analysis model and the reliability of the simulation results.

[0042] In one embodiment, see Figure 2 As shown, step S3 specifically includes: S31: Constructing a power meteorological gain matrix for spatiotemporal four-dimensional approximation. The present invention uses a spatiotemporal four-dimensional approximation method to gradually approximate the state of the analysis model to the observed value using an additional term in each integration time step of the analysis model, thereby ensuring that the analysis model state is corrected according to the latest observation in each integration time step. The form is as follows: (1) in, Represents the model state, namely the meteorological elements of wind speed, temperature, and humidity, are the physical and dynamic terms of the original analysis model, is the pre-processed power meteorological observation value, is the power meteorological gain matrix, whose size is the approximation time of the observation element in the analysis model The inverse of , which mainly controls the correction strength of the analysis model.

[0043] The present invention sets different relaxation times based on sensitive tests of power meteorological observation factors, specifically: temperature, humidity and other continuous factors, hours; intermittent factors such as wind speed and direction, hours; and precipitation event factors, Hour.

[0044] In addition, since the power meteorological monitoring data is mainly concentrated in the atmospheric boundary layer, a shorter approach time is used in the lower layer of the model (1-2 km near the ground) , while in the upper troposphere, the approach time is appropriately extended .

[0045] S32: Set the continuous analysis window and loop strategy. To construct a multi-year historical training data set for the large power meteorological model, it is necessary to ensure that the latest observations are introduced in each integration time step of the analysis window, and to ensure that the analysis model operates stably over a long period of time. That is, the problem of pattern drift accumulation and the connection between the cyclic fields need to be solved during continuous analysis. The present invention determines the analysis window as a cycle of every 3 hours based on the time resolution of the large model training data set. On this basis, the sliding window method is adopted, and the simulation state is saved at the end of each window as the initial value of the next round of numerical integration, thereby realizing simulation analysis for many consecutive years.

[0046] In one embodiment, see Figure 2 As shown, step S4 specifically includes: S41: Segmentation of long-term historical simulation tasks. Multi-year historical simulation tasks are divided into several cycles based on the cold and hot start of the analysis model. Each cycle is determined to be one day. The analysis model is cold started, and the sliding window within each cycle is set as the analysis model hot start. Each cycle can be considered a complete simulation analysis.

[0047] S42: Automated control script for long-term historical simulation. Following the flow of all the aforementioned steps, starting with "data processing - model optimization - analysis settings - long-term simulation," write a Perl or Python automated control script to form a job script for the long-term historical simulation task.

[0048] S43: Submit a long-term historical simulation analysis task. Submit the job script for the long-term historical simulation task on a high-performance cluster (HPC). If hardware conditions permit, multiple job scripts can be submitted in parallel, with each job processing multiple time periods to improve the efficiency of the long-term historical simulation.

[0049] In one embodiment, see Figure 2 As shown, step S5 specifically includes: S51: Select observational and simulation analysis data for comparison. Observational data includes long-term, stable, and reliable data from conventional meteorological stations, wind and light measurement data from new energy stations, and online micrometeorological monitoring data from transmission lines. Simulation analysis data extracts the results generated in step S4 that match the observed temporal and spatial data.

[0050] S52: Statistical simulation analysis of data error indicators. Deviation indicators such as RMSE, Bias, MAE, correlation coefficient, and TS (Threat Score) are calculated for the entire period and certain key periods (e.g., summer and winter). This examines the error distribution and simulation accuracy of different factors over time and in different regions.

[0051] S53: Correct the deviation of the simulation analysis data. If linear deviation is present, simple linear regression is used to correct the power meteorological reanalysis data; corrected training data for the power meteorological large model is obtained. If the deviation is complex in time or space, random forest or neural network methods are used to perform multivariate correction of the power meteorological reanalysis data; corrected training data for the power meteorological large model is obtained.

[0052] In one embodiment, see Figure 2 As shown, step S6 specifically includes: S61: Determine the storage format for the large model training data set. Because the training data set for the large power meteorological forecast model is large and dispersed, and requires multiple subsequent accesses and secondary development, the data storage format is determined to be NetCDF4 to improve the efficiency of model training and business system use, meeting meteorological simulation and forecasting standards.

[0053] S62: Naming the analysis period for the large model training set data. The time resolution of the large power meteorology model training set data is one hour. That is, new observation data enters the model every hour, and a higher-precision data set is generated through analysis. Therefore, each analysis period is named with a "YYYYMMDDHH" time granularity accurate to the hour.

[0054] S63: Data Description for the Large Model Training Set. The data for the large power meteorology model training set is stored hourly. This means that the minimum unit for model training and secondary development is a compressed NetCDF4 file of an hour. To ensure reliable data reading, data description is required. Add grid coordinate information to the data file, including the latitude, longitude, altitude, and pressure layer of each grid point, as well as data variable descriptions, including variable name, physical meaning, and unit.

[0055] This invention not only effectively enhances the meteorological services provided by the power industry, but also incorporates real-time monitoring data from renewable energy stations and transmission lines. Simulation results show that at a 3km resolution, the root mean square error (RMSE) of wind speed forecasts for wind farms can be reduced to over 1.2m / s, while the accuracy of photovoltaic power station irradiance forecasts is improved by approximately 18%. Furthermore, this method achieves breakthroughs in high-resolution field analysis capabilities. By employing regional nested grids and adaptive terrain filtering, the spatial resolution of the analysis field is increased from approximately 0.25° (approximately 25km) in traditional global models to 3km. This method is capable of analyzing wind speed gradients within a 50m height range in complex terrain, with prediction errors kept below 5%.

[0056] See also Figure 3 As shown, an embodiment of the present invention provides a method for constructing a large-scale electric power meteorological model training data set, comprising: S100, obtaining data for generating a large-scale power meteorological model training data set; the data includes global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data for many years; S200, constructing an analysis model for generating a training data set; inputting the data for generating the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data; S300: forming a large power meteorological model training data set based on the power meteorological reanalysis data.

[0057] In a specific embodiment, the step of obtaining data to generate a large power meteorological model training data set specifically includes: Obtain global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data for many years, and preprocess them to obtain data for generating a large power meteorological model training data set.

[0058] In a specific embodiment, the step of performing pretreatment specifically includes: Preprocess the global meteorological raw reanalysis data and convert the format into GRIB format; Preprocess the static data to make the spatial resolution and projection coordinates of the static data consistent with the analysis model; The multi-source heterogeneous historical observation data of electric power meteorology are preprocessed and converted into the format required for assimilation of the reanalysis model.

[0059] In a specific embodiment, in the step of obtaining global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data, ERA5 data from the European Center, FNL data from the United States, and CMA-RA data from China are collected as global meteorological original reanalysis data; Obtain one of SRTM, GTOPO, GlobCover, and MODIS as static data; Multi-source heterogeneous power meteorological historical observation data include: data from ground meteorological stations, sounding stations, weather radars, meteorological satellite data, wind and light measurement data from new energy sites, online micrometeorological monitoring data from transmission lines, and wind measurement radars and wind profiler data for power grid facilities and equipment.

[0060] In a specific embodiment, the steps of constructing an analysis model for generating a training data set; inputting the data for generating the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data specifically include: Obtain the configuration of the boundary layer scheme, cloud microphysics scheme, land surface process scheme, and simulation integration time step of the analytical model; Construct a power meteorological gain matrix of spatiotemporal four-dimensional approximation to obtain continuous analysis windows and cycle strategies; Based on the acquired analysis model's boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and circulation strategy, as well as the constructed power meteorological gain matrix of empty four-dimensional approximation, analysis and simulation are carried out to obtain power meteorological reanalysis data.

[0061] In a specific embodiment, in the step of obtaining the boundary layer scheme, cloud microphysics scheme, land surface process scheme and configuration of the simulation integration time step of the analysis model, the boundary layer scheme adopts the MYNN scheme; The cloud microphysics solution uses the WDM6 or Morrison 2-Moment solution; The land surface process scheme adopts Noah LSM or CLM4; The simulation integration time step was 18–30 s.

[0062] In a specific embodiment, the steps of constructing a spatiotemporal four-dimensional approximation power meteorological gain matrix and obtaining a continuous analysis window and a cycle strategy specifically include: The constructed power meteorological gain matrix is: dX / dt=F(X)+G(X_obs-X)(1) Where X represents the model state, F(X) is the physical and dynamic terms of the original analysis model, X_obs is the preprocessed power meteorological observation value, and G is the power meteorological gain matrix; The duration of the continuous analysis window is 3 hours; the cyclic strategy adopts a sliding window method, and the simulation state is saved at the end of each window as the initial value of the next round of numerical integration.

[0063] In a specific embodiment, the boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and cycle strategy of the obtained analysis model, and the constructed empty four-dimensional approximation power meteorological gain matrix are used to carry out analysis and simulation, and in the step of obtaining power meteorological reanalysis data, the analysis and simulation task is divided into several cycles according to the cold and hot starts of the analysis model. Each cycle is determined to be one day, the analysis model is cold started, and the sliding window within each cycle is set to the hot start of the analysis model. Each cycle is regarded as a complete simulation analysis; each cycle is simulated separately, or multiple cycles are simulated in parallel.

[0064] In a specific embodiment, the step of forming a large power meteorological model training data set based on the power meteorological reanalysis data specifically includes: Conduct simulation effect verification and deviation correction on power meteorological reanalysis data to obtain corrected power meteorological large model training data; The corrected large-scale electric power meteorological model training data is segmented and stored according to the preset data storage format, analysis period naming and data description to obtain the large-scale electric power meteorological model training data set.

[0065] See also Figure 4 As shown, an embodiment of the present invention provides a device for constructing a large-scale electric power meteorological model training data set, comprising: An acquisition module is used to acquire data for generating a large-scale power meteorological model training data set; the data includes global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data for many years; An analysis module is used to construct an analysis model for generating a training data set; input the data for generating the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data; The generation module is used to form a large power meteorological model training data set based on the power meteorological reanalysis data.

[0066] In a specific embodiment, the acquisition module is specifically configured as follows: Obtain global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data for many years, and preprocess them to obtain data for generating a large power meteorological model training data set.

[0067] In a specific embodiment, the step of obtaining the preprocessing module specifically includes: Preprocess the global meteorological raw reanalysis data and convert the format into GRIB format; Preprocess the static data to make the spatial resolution and projection coordinates of the static data consistent with the analysis model; The multi-source heterogeneous historical observation data of electric power meteorology are preprocessed and converted into the format required for assimilation of the reanalysis model.

[0068] In a specific embodiment, in the step of acquiring global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data, the acquisition module collects ERA5 data from the European Center, FNL data from the United States, and CMA-RA data from China as global meteorological original reanalysis data; Obtain one of SRTM, GTOPO, GlobCover, and MODIS as static data; Multi-source heterogeneous power meteorological historical observation data include: data from ground meteorological stations, sounding stations, weather radars, meteorological satellite data, wind and light measurement data from new energy sites, online micrometeorological monitoring data from transmission lines, and wind measurement radars and wind profiler data for power grid facilities and equipment.

[0069] In a specific embodiment, the analysis module is specifically configured as follows: Obtain the configuration of the boundary layer scheme, cloud microphysics scheme, land surface process scheme, and simulation integration time step of the analytical model; Construct a power meteorological gain matrix of spatiotemporal four-dimensional approximation to obtain continuous analysis windows and cycle strategies; Based on the acquired analysis model's boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and circulation strategy, as well as the constructed power meteorological gain matrix of empty four-dimensional approximation, analysis and simulation are carried out to obtain power meteorological reanalysis data.

[0070] In a specific embodiment, in the step of obtaining the boundary layer scheme, cloud microphysics scheme, land surface process scheme and configuration of the simulation integration time step of the analysis model, the boundary layer scheme adopts the MYNN scheme; The cloud microphysics solution uses the WDM6 or Morrison 2-Moment solution; The land surface process scheme adopts Noah LSM or CLM4; The simulation integration time step was 18–30 s.

[0071] In a specific embodiment, the steps of constructing a spatiotemporal four-dimensional approximation power meteorological gain matrix and obtaining a continuous analysis window and a cycle strategy specifically include: The constructed power meteorological gain matrix is: dX / dt=F(X)+G(X_obs-X)(1) Where X represents the model state, F(X) is the physical and dynamic terms of the original analysis model, X_obs is the preprocessed power meteorological observation value, and G is the power meteorological gain matrix; The duration of the continuous analysis window is 3 hours; the cyclic strategy adopts a sliding window method, and the simulation state is saved at the end of each window as the initial value of the next round of numerical integration.

[0072] In a specific embodiment, the boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and cycle strategy of the obtained analysis model, and the constructed empty four-dimensional approximation power meteorological gain matrix are used to carry out analysis and simulation, and in the step of obtaining power meteorological reanalysis data, the analysis and simulation task is divided into several cycles according to the cold and hot starts of the analysis model. Each cycle is determined to be one day, the analysis model is cold started, and the sliding window within each cycle is set to the hot start of the analysis model. Each cycle is regarded as a complete simulation analysis; each cycle is simulated separately, or multiple cycles are simulated in parallel.

[0073] In a specific embodiment, the step of forming a large power meteorological model training data set based on the power meteorological reanalysis data specifically includes: Conduct simulation effect verification and deviation correction on power meteorological reanalysis data to obtain corrected power meteorological large model training data; The corrected large-scale electric power meteorological model training data is segmented and stored according to the preset data storage format, analysis period naming and data description to obtain the large-scale electric power meteorological model training data set.

[0074] See also Figure 5 As shown, an embodiment of the present invention provides an electronic device 100 for implementing a method for constructing a large-scale power meteorological model training data set; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0075] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the method for constructing a large-scale power meteorological model training data set described in the embodiment by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data (such as audio data) created based on the use of the electronic device 100. In addition, the memory 101 can include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device.

[0076] The at least one processor 102 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 gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100 and connects various parts of the entire electronic device 100 using various interfaces and lines.

[0077] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a method for constructing a large-scale power meteorological model training data set. The processor 102 can execute the plurality of instructions to implement: Obtain data to generate a large-scale power meteorological model training data set; Constructing an analysis model for generating a training data set; inputting the data for generating the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data; A large power meteorological model training dataset is formed based on the power meteorological reanalysis data.

[0078] If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).

[0079] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, 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 magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] 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 flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts 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, 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 flowcharts and / or block diagrams. 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.

[0081] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 The function specified in one or more boxes.

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for constructing a large-scale power meteorological model training data set, characterized in that: include: Acquire data for generating a large-scale power meteorological model training data set; the data includes global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data for many years; Constructing an analysis model for generating a training data set; inputting the data for generating the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data; A large power meteorological model training dataset is formed based on the power meteorological reanalysis data.

2. The method for constructing a large-scale electric power meteorological model training data set according to claim 1, characterized in that: The step of obtaining data for generating a large electric power meteorological model training data set specifically includes: Obtain global meteorological original reanalysis data, static data, and multi-source heterogeneous historical power meteorological observation data for many years, and preprocess them to obtain data for generating a large power meteorological model training data set.

3. The method for constructing a large-scale electric power meteorological model training data set according to claim 2, characterized in that: The steps of performing pre-processing specifically include: Preprocess the global meteorological raw reanalysis data and convert the format into the universal binary format GRIB; Preprocess the static data to make the spatial resolution and projection coordinates of the static data consistent with the analysis model; The multi-source heterogeneous historical observation data of electric power meteorology are preprocessed and converted into the format required for assimilation of the reanalysis model.

4. The method for constructing a large-scale electric power meteorological model training data set according to claim 1, characterized in that: The constructing generates an analysis model for a training data set; The step of inputting the data generated from the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data specifically includes: Obtain the configuration of the boundary layer scheme, cloud microphysics scheme, land surface process scheme, and simulation integration time step of the analytical model; Construct a power meteorological gain matrix of spatiotemporal four-dimensional approximation to obtain continuous analysis windows and cycle strategies; Based on the acquired analysis model's boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and circulation strategy, as well as the constructed power meteorological gain matrix of empty four-dimensional approximation, analysis and simulation are carried out to obtain power meteorological reanalysis data.

5. The method for constructing a large-scale electric power meteorological model training data set according to claim 4, characterized in that: In the step of obtaining the boundary layer scheme, cloud microphysics scheme, land surface process scheme and configuration of the simulation integration time step of the analysis model, the boundary layer scheme adopts the MYNN scheme; The cloud microphysics solution uses the WDM6 or Morrison 2-Moment solution; The land surface process scheme adopts Noah LSM or CLM4; The simulation integration time step was 18–30 s.

6. The method for constructing a large-scale electric power meteorological model training data set according to claim 4, characterized in that: The steps of constructing a power meteorological gain matrix of a spatiotemporal four-dimensional approximation and obtaining a continuous analysis window and a cycle strategy specifically include: The constructed power meteorological gain matrix is: (1) in, Represents the model state, are the physical and dynamic terms of the original analysis model, is the pre-processed power meteorological observation value, is the power meteorological gain matrix; The duration of the continuous analysis window is 3 hours; the cyclic strategy adopts a sliding window method, and the simulation state is saved at the end of each window as the initial value of the next round of numerical integration.

7. The method for constructing a large-scale electric power meteorological model training data set according to claim 4, characterized in that: The boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and cycle strategy of the obtained analysis model, and the constructed power meteorological gain matrix of the empty four-dimensional approximation are used to carry out analysis and simulation, and in the step of obtaining power meteorological reanalysis data, the analysis and simulation task is divided into several cycles according to the cold and hot starts of the analysis model. Each cycle is determined to be one day, the analysis model is cold started, and the sliding window within each cycle is set to the hot start of the analysis model. Each cycle is regarded as a complete simulation analysis; each cycle is simulated separately, or multiple cycles are simulated in parallel.

8. The method for constructing a large-scale electric power meteorological model training data set according to claim 1, characterized in that: The step of forming a large power meteorological model training data set based on the power meteorological reanalysis data specifically includes: Conduct simulation effect verification and deviation correction on power meteorological reanalysis data to obtain corrected power meteorological large model training data; The corrected large-scale electric power meteorological model training data is segmented and stored according to the preset data storage format, analysis period naming and data description to obtain the large-scale electric power meteorological model training data set.

9. A device for constructing a large-scale training data set for a power meteorological model, characterized in that: include: An acquisition module is used to acquire data for generating a large-scale power meteorological model training data set; the data includes global meteorological original reanalysis data, static data, and multi-source heterogeneous power meteorological historical observation data for many years; An analysis module is used to construct an analysis model for generating a training data set; input the data for generating the large power meteorological model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data; The generation module is used to form a large power meteorological model training data set based on the power meteorological reanalysis data.

10. The device for constructing a large-scale electric power meteorological model training data set according to claim 9, characterized in that: The step of obtaining data for generating a large electric power meteorological model training data set specifically includes: Obtain global meteorological original reanalysis data, static data, and multi-source heterogeneous historical power meteorological observation data for many years, and preprocess them to obtain data for generating a large power meteorological model training data set.

11. The device for constructing a large-scale electric power meteorological model training data set according to claim 10, characterized in that: The steps of performing pre-processing specifically include: Preprocess the global meteorological raw reanalysis data and convert the format into the universal binary format GRIB; Preprocess the static data to make the spatial resolution and projection coordinates of the static data consistent with the analysis model; The multi-source heterogeneous historical observation data of electric power meteorology are preprocessed and converted into the format required for assimilation of the reanalysis model.

12. The device for constructing a large-scale electric power meteorological model training data set according to claim 9, characterized in that: The steps of constructing an analysis model for generating a training data set; inputting the data of the generated power meteorological large model training data set into the analysis model for analysis and simulation to obtain power meteorological reanalysis data specifically include: Obtain the configuration of the boundary layer scheme, cloud microphysics scheme, land surface process scheme, and simulation integration time step of the analytical model; Construct a power meteorological gain matrix of spatiotemporal four-dimensional approximation to obtain continuous analysis windows and cycle strategies; Based on the acquired analysis model's boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and circulation strategy, as well as the constructed power meteorological gain matrix of empty four-dimensional approximation, analysis and simulation are carried out to obtain power meteorological reanalysis data.

13. The device for constructing a large-scale electric power meteorological model training data set according to claim 12, characterized in that: In the step of obtaining the boundary layer scheme, cloud microphysics scheme, land surface process scheme and configuration of the simulation integration time step of the analysis model, the boundary layer scheme adopts the MYNN scheme; The cloud microphysics solution uses the WDM6 or Morrison 2-Moment solution; The land surface process scheme adopts Noah LSM or CLM4; The simulation integration time step was 18–30 s.

14. The apparatus for constructing a large-scale electric power meteorological model training data set according to claim 12, characterized in that: The steps of constructing a power meteorological gain matrix of a spatiotemporal four-dimensional approximation and obtaining a continuous analysis window and a cycle strategy specifically include: The constructed power meteorological gain matrix is: (1) in, Represents the model state, are the physical and dynamic terms of the original analysis model, is the pre-processed power meteorological observation value, is the power meteorological gain matrix; The duration of the continuous analysis window is 3 hours; the cyclic strategy adopts a sliding window method, and the simulation state is saved at the end of each window as the initial value of the next round of numerical integration.

15. The device for constructing a large-scale electric power meteorological model training data set according to claim 12, characterized in that: The boundary layer scheme, cloud microphysics scheme, land surface process scheme, simulation integration time step, continuous analysis window and cycle strategy of the obtained analysis model, and the constructed power meteorological gain matrix of the empty four-dimensional approximation are used to carry out analysis and simulation, and in the step of obtaining power meteorological reanalysis data, the analysis and simulation task is divided into several cycles according to the cold and hot starts of the analysis model. Each cycle is determined to be one day, the analysis model is cold started, and the sliding window within each cycle is set to the hot start of the analysis model. Each cycle is regarded as a complete simulation analysis; each cycle is simulated separately, or multiple cycles are simulated in parallel.

16. The device for constructing a large-scale electric power meteorological model training data set according to claim 9, characterized in that: The step of forming a large power meteorological model training data set based on the power meteorological reanalysis data specifically includes: Conduct simulation effect verification and deviation correction on power meteorological reanalysis data to obtain corrected power meteorological large model training data; The corrected large-scale electric power meteorological model training data is segmented and stored according to the preset data storage format, analysis period naming and data description to obtain the large-scale electric power meteorological model training data set.

17. An electronic device, characterized in that: It comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement a method for constructing a large-scale power meteorological model training data set as described in any one of claims 1 to 8.

18. A computer-readable storage medium, characterized in that The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by the processor, it implements the method for constructing a large-scale power meteorological model training data set according to any one of claims 1 to 8.

Images