Gust short-term weather forecast method

Abstract

The invention discloses a gust short-term weather forecasting method. The method comprises the following steps: S1, acquiring a downscaling forecasting result of a wind field; s2, acquiring a gust actual condition extrapolation forecast result; s3, acquiring a deep learning gust extrapolation forecast result; and S4, obtaining an extended orthogonal empirical decomposition gust extrapolation forecast result. And S5, comprehensively applying the forecast results of the steps S1-S4 according to requirements to obtain a final gust short-term weather forecast. According to the method, gust short-term and imminent forecast with the frequency of 1km and the resolution ratio can be realized 10 minutes by 10 minutes in the future 2 hours.

Description

Technical Field

[0001] The invention belongs to the field of gust weather forecasting, and in particular relates to a gust short-term weather forecasting method. Background Art

[0002] Gust short-term weather forecast is of great significance for specific activities or industries such as aviation, navigation, construction, outdoor activities, etc. Specifically, it is: Aviation safety: For pilots and air traffic controllers, it is very important to know the gust conditions in a timely and accurate manner, which can help them make correct flight decisions and avoid danger to aircraft encountering strong gusts; Navigation safety: For ships and crew members, knowing the occurrence of gusts at sea in advance can help to adopt corresponding navigation strategies to avoid serious impact or damage to ships; Construction safety: At the construction site, gusts may pose a threat to high-altitude operations, hoisting operations, etc., so early warning of the occurrence of gusts can help to take safety measures to ensure the safety of construction personnel and equipment; Outdoor activity safety: When holding outdoor sports events, camping, mountaineering and other activities, warning of the occurrence of gusts can help organizers adjust activity plans and ensure the safety of participants; Agricultural production: For agricultural production, timely understanding of the gust conditions can help farmers take corresponding protective measures to protect crops and facilities from loss. However, for these specific activities or industries that require timely response, the current gust short-term weather forecast has a low level of precision and cannot provide more accurate gust short-term weather forecast services. Summary of the invention

[0003] In order to address the shortcomings of the existing technology, the present invention uses four different methods, comprehensively uses deep learning extrapolation, statistical method extrapolation, wind field downscaling diagnostic model, optical flow method extrapolation, data assimilation and other methods, combined with site wind speed observation, radar reflectivity observation, fusion of real-time grid products, and numerical models to achieve a short-term forecast of gusts with a frequency of 10 minutes per 1km resolution for the next 2 hours.

[0004] In order to achieve the above object, a method for short-term gust weather forecasting according to an embodiment of the present invention comprises the following steps:

[0005] S1. Obtain wind field downscaling forecast results;

[0006] S2. Obtaining the extrapolated forecast result of the actual gust situation;

[0007] S3. Obtain deep learning gust extrapolation forecast results;

[0008] S4. Obtain the extended orthogonal empirical decomposition gust extrapolation forecast result.

[0009] Furthermore, the method further comprises step S5, comprehensively applying the forecast results of steps S1-S4 according to needs to obtain a final gust short-term weather forecast.

[0010] Further, step S1, obtaining the wind field downscaling forecast result, specifically:

[0011] S11. Download the forecast data of ground and atmospheric meteorological elements for the next 2 hours;

[0012] S12, converting the data into a WPS intermediate file, and generating an intermediate file recognizable by WRF in combination with the terrain data;

[0013] S13, using the WRF mode, reading the intermediate file and generating a data file that can be recognized by the CALWRF program;

[0014] S14. Use CALWRF to generate a data file that can be recognized by CALMET;

[0015] S15. Use CALMET model to perform wind field downscaling diagnosis;

[0016] S16. Output the final wind field downscaling forecast result.

[0017] Further, step S2, obtaining the actual extrapolation forecast result of the gust, specifically:

[0018] S21. Download the numerical forecast products, station wind speed observation data and radar reflectivity data of the near time;

[0019] S22, using interpolation to unify different data times;

[0020] S23, calculating the radar reflectivity motion vector based on the optical flow method;

[0021] S24, based on multi-scale variational analysis, the station wind speed observation, radar reflectivity motion vector and numerical forecast are integrated;

[0022] S25. Downscaling diagnosis is performed on the fused data based on the CALMET model to obtain a near-realistic near-surface wind field that complies with the law of conservation of mass;

[0023] S26. Use the near-surface wind field to adjust the wind field predicted by the numerical model;

[0024] S27. Output the final wind field extrapolation forecast.

[0025] Further, step S3, obtaining the deep learning gust extrapolation forecast result, specifically:

[0026] S31, based on historical radar reflectivity and site wind speed observations, the model parameters are trained using the PhyDnet method and stored locally;

[0027] S32. Download the numerical forecast products, station wind speed observation data and radar reflectivity data of the near time;

[0028] S33, using interpolation method to unify different data times;

[0029] S34, read the pre-trained model parameter file, use the site wind speed observation and radar reflectivity as input, and generate a site wind speed forecast for the next 2 hours at a frequency of 10 minutes;

[0030] S35. Use multiscale variational analysis to fuse site wind speed forecasts with model forecasts;

[0031] S36. Output the final wind field extrapolation forecast.

[0032] Further, step S4, obtaining the extended orthogonal empirical decomposition gust extrapolation forecast result, specifically:

[0033] S41, download historical HRCLDAS data and site wind speed observations, use the extended orthogonal empirical decomposition method to train the model parameters, and store them locally;

[0034] S42, downloading the HRCLDAS multi-source fusion near-real-time wind field data and numerical forecast data and site wind speed observation data at the near time;

[0035] S43, using time interpolation, interpolating the HRCLDAS and numerical forecast data to the time of the observed data;

[0036] S44, based on multi-scale variational analysis, the interpolation results of the station wind speed observations are used for correction to obtain near-real-world grid wind speed data;

[0037] S45, loading the pre-trained model parameters, inputting the obtained data, and performing wind speed forecasting;

[0038] S46. Output the final wind field extrapolation forecast.

[0039] The beneficial effects of the present invention are:

[0040] The present invention implements the latest deep learning gust forecasting method, and innovatively designs a gust statistical forecasting method based on empirical orthogonal decomposition, a gust observation extrapolation method based on optical flow method and wind field diagnostic downscaling model, and realizes gust diagnostic downscaling refined forecasting for traditional numerical forecast products. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of calling a code file for obtaining a wind field downscaling forecast result in a method for short-term gust weather forecasting of the present invention;

[0042] Figure 2 It is a schematic diagram of calling a code file for obtaining a gust real-time extrapolation forecast result according to an embodiment of the present invention;

[0043] Figure 3 It is a schematic diagram of calling a code file for obtaining a deep learning gust extrapolation forecast result according to an embodiment of the present invention;

[0044] Figure 4 It is a schematic diagram of a code file call for obtaining an extended orthogonal empirical decomposition gust extrapolation forecast result according to an embodiment of the present invention;

[0045] Figure 5 is a flow chart of an extended orthogonal empirical decomposition method according to an embodiment of the present invention;

[0046] Figure 6 The present invention is a technical roadmap for a short-term gust weather forecast system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the present invention more clear, further description is given below in conjunction with the accompanying drawings and embodiments.

[0048] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0049] As an example of implementation of the present invention, as shown in the accompanying drawings, the present invention provides a method for short-term gust weather forecasting, which includes the following steps:

[0050] S1. Obtain wind field downscaling forecast results;

[0051] CALMET is a wind field diagnostic model based on the constraint of the law of conservation of mass. It sets specific constraints and iteratively adjusts the grid wind field to obtain a downscaled wind field that is relatively consistent with more detailed underlying surface characteristics.

[0052] The "CALMET downscaling" method uses the CALMET wind field diagnostic model to perform downscaling analysis on the 3km resolution numerical forecast products updated hourly, and obtains the 1km resolution gust downscaling forecast for the next 2 hours and 10 minutes in the western Guangdong region. This method mainly solves the problem of wind field-terrain incompatibility caused by simply using interpolation methods to downscale numerical forecast products. At the same time, it also makes certain corrections to the wind field numerical forecast results based on more detailed and realistic terrain data. The detailed calculation process of the method is as follows:

[0053] (1) Download the forecast data of ground and atmospheric meteorological elements for the next 2 hours

[0054] (2) Convert the data into WPS intermediate files and combine them with terrain data to generate intermediate files that can be recognized by WRF.

[0055] (3) Using the WRF mode, read the intermediate file and generate a data file that can be recognized by the CALWRF program.

[0056] (4) Use CALWRF to generate data files that can be recognized by CALMET.

[0057] (5) Use the CALMET model to perform wind field downscaling diagnosis.

[0058] (6) Output the final wind field downscaling forecast results.

[0059] Code structure: The method code is mainly divided into two functional modules:

[0060] (1) Module 1

[0061] This part of the program is stored in / home / wind_gust_algorithm / calmet_downscale / run / , and is mainly composed of 4 python code files, namely run_calmet_downscaling.py, DataInterface.py, Algorithm.py, and winter.py. Among them, run_calmet_downscaling.py is the method execution program (main program), which is responsible for calling other programs. DataInterface.py is responsible for downloading the required numerical model products, writing data, etc. Algorithm.py mainly includes functions such as model product prediction time interpolation, spatial interpolation program, and module 2 call. Winter.py is responsible for storing data formats that can be recognized by WPS and is called by DataInterface.py.

[0062] (2) Module 2

[0063] This part is mainly composed of the compiled WRF model pre-processor, CALWRF conversion tool, and CALMET model body. The code is stored in the / home / wind_gust_algorithm / calmet directory. Among them, / calmet / WPS is the WRF pre-processor, which is responsible for reading the model forecast data and underlying surface data, and generating intermediate data files that can be recognized by WRF. / calmet / WRF stores the WRF model program, which is responsible for converting the data generated by the WPS program into the WRF model initial field; / calmet / bin / calwrf is a data format conversion program, which is responsible for reading the WRF model initial field data and converting it into 3D.dat and 2D.dat format files that can be recognized by the CALMET model; / calmet / bin / calmet is the CALMET model program, which is responsible for reading the data output by calwrf and performing wind field downscaling diagnosis to obtain a 1km wind field that conforms to the fine terrain characteristics and satisfies the mass conservation. The code file call is as shown in the attachment. Figure 1 shown.

[0064] S2. Obtaining the extrapolated forecast result of the actual gust situation;

[0065] This method is based on radar reflectivity data, station wind speed observation data, and numerical model forecast data. It uses multi-scale variational data assimilation, optical flow method, and CALMET model to fuse multiple observation data with forecast data to obtain a relatively accurate and refined ground near-real-time wind field. Finally, the near-real-time wind field is used to adjust the forecast data to quickly obtain a 1km resolution gust forecast product with a frequency of 2 hours and 10 minutes in the future in western Guangdong. The detailed calculation process of the method is as follows:

[0066] (1) Download the numerical forecast products, station wind speed observation data and radar reflectivity data for the near future.

[0067] (2) Use interpolation to unify the time of different data.

[0068] (3) Calculate the radar reflectivity motion vector based on the optical flow method. The optical flow method used is the Variational EchoTracking method, and the detailed description is given in Germann and Zawadzki (2002).

[0069] (4) Based on multi-scale variational analysis, the station wind speed observation, radar reflectivity motion vector and numerical forecast are integrated. Multi-scale variational analysis is a data assimilation method based on multiple grids and variational methods. It accelerates the assimilation process by considering grids with different resolutions and integrates data information of different scales. Its formula is as follows:

[0070]

[0071] y 1 =y o -h 1 (x b )

[0072] y k =y k-1 -h k-1 (x k-1 )k=2,....,n

[0073]

[0074] (5) Based on the CALMET model, the fused data is downscaled and diagnosed to obtain a near-realistic near-surface wind field that conforms to the law of conservation of mass.

[0075] (6) The wind field near the ground is used to adjust the wind field predicted by the numerical model. The adjustment formula is as follows: i The forecast time is τ, which is 6 hours.

[0076] v adj =w t v obs +(1-w t )v fcst

[0077]

[0078] (7) Output the final wind field extrapolation forecast.

[0079] Code structure: The method code is mainly divided into four functional modules:

[0080] (1) Module 1

[0081] This part of the program is stored in / home / wind_gust_algorithm / obs_extrapolation / src / , which is mainly responsible for calling each module, downloading and reading data. The module consists of 4 python files, namely run_obs_extrapolation_main.py, AlgorithmLib.py, DataInterface.py, and winter.py. Among them, run_calmet_downscaling.py is the method execution program (main program), which is responsible for calling other programs. DataInterface.py is responsible for downloading the required numerical model products, writing data, etc. Algorithm.py mainly includes functions such as model product prediction time interpolation, spatial interpolation program, and calling other modules. winter.py is responsible for storing data formats that can be recognized by WPS and is called by DataInterface.py.

[0082] (2) Module 2

[0083] This part of the program is stored in / home / wind_gust_algorithm / obs_extrapolation / src / . It is mainly responsible for using the optical flow method to extrapolate the radar reflectivity and obtain the reflectivity movement vector. The module contains two program files, vet.py and _vet.c. The two programs jointly implement the Variational Echo Tracking method and are finally called by Algorithm.py of module 1.

[0084] (3) Module 3

[0085] This part of the program is stored in / home / wind_gust_algorithm / obs_extrapolation / src / , which is mainly responsible for multi-scale variational analysis and fusion of different data. The module contains a program file, stmas_lib.f90, which implements the multi-scale variational analysis method and is finally called by Algorithm.py of module 1.

[0086] (4) Module 4

[0087] This part is mainly composed of the compiled WRF model pre-processor, CALWRF conversion tool, and CALMET model body. The code is stored in the / home / wind_gust_algorithm / calmet directory. Among them, / calmet / WPS is the WRF pre-processor, which is responsible for reading the model forecast data and underlying surface data, and generating intermediate data files that can be recognized by WRF. / calmet / WRF stores the WRF model program, which is responsible for converting the data generated by the WPS program into the WRF model initial field; / calmet / bin / calwrf is a data format conversion program, which is responsible for reading the WRF model initial field data and converting it into 3D.dat and 2D.dat format files that can be recognized by the CALMET model; / calmet / bin / calmet is the CALMET model program, which is responsible for reading the data output by calwrf and performing wind field downscaling diagnosis to obtain a 1km wind field that conforms to the fine terrain characteristics and satisfies the mass conservation. The code file call is as shown in the attachment. Figure 2 shown.

[0088] S3. Obtain deep learning gust extrapolation forecast results;

[0089] This method is based on radar reflectivity data, site wind speed observation data, and numerical model forecast data. PhyDnet is used to establish a forecast model to achieve short-term site wind speed forecasts, and multi-scale variational analysis is used to integrate the interpolation results of the corresponding numerical forecasts to quickly obtain a 1km resolution gust forecast product for the next 2 hours and 10 minutes in the western Guangdong region. The detailed calculation process of the method is as follows:

[0090] (1) Based on historical radar reflectivity and site wind speed observations, the model parameters are trained using the PhyDnet method and stored locally. PhyDnet is a deep learning model used for radar echo extrapolation forecasting. The same technical means can also be applied to gust extrapolation forecasting. The specific implementation process of this method is as follows:

[0091] ① Identification of severe convective gusts

[0092] Collect historical multi-year reflectivity factor networking data and minute-level instantaneous average wind observations in the study area; use weather radar observation data in Guangdong Province for quality control and networking to form a weather radar sample data set with a 6-minute interval and a sequence length of 3 hours. Use the instantaneous average wind observations with a resolution of 5 minutes on the ground and resample to a time resolution of 6 minutes, and then grid the instantaneous average wind observations to form an instantaneous average wind sample data set that matches the weather radar observations. Use the radar networking data and the instantaneous average wind data on the ground 1 hour before the match as input factors, and use the instantaneous average wind observations of the next 2 hours as the true value field to establish an instantaneous average wind forecast data set; split the instantaneous average wind forecast data set into training set, validation set, and test set at a ratio of 8:1:1.

[0093] ② Establishment of severe convective gust nowcasting model

[0094] Based on the PhyDnet spatiotemporal convolutional grid model, an instantaneous average wind forecast model is constructed to perform instantaneous average wind forecast. The technical route is as follows:

[0095] a. The PhyDnet spatiotemporal convolutional network improves the long-term forecast effect by splitting the motion of the echo into advection motion terms and nonlinear disturbance terms, and introduces an attention mechanism. The constructed sequence training set is used to train the gale forecast model, and the validation set is used to test the model effect. Based on the test results, the model structure and loss function are continuously adjusted to continuously optimize the model.

[0096] b. On this basis, the historical observation data of instantaneous maximum wind and average wind in the study area are continuously collected, and a relationship model between instantaneous average wind and maximum wind is established based on statistical methods.

[0097] c. Use a variety of scoring methods (such as TS and HSS) to evaluate the instantaneous average wind forecast results of the deep learning model on the test set, taking into account the accuracy and speed of the model, and obtain the optimal deep learning model for instantaneous average wind forecast.

[0098] (2) Download the numerical forecast products, station wind speed observation data and radar reflectivity data for the near future.

[0099] (3) Use interpolation to unify the time of different data.

[0100] (4) Read the pre-trained model parameter file, use the site wind speed observation and radar reflectivity as input, and generate a site wind speed forecast every 10 minutes for the next 2 hours.

[0101] (5) Use multi-scale variational analysis to integrate site wind speed forecasts with model forecasts.

[0102] (6) Output the final wind field extrapolation forecast.

[0103] Code structure: The method code is mainly divided into three functional modules:

[0104] (1) Module 1

[0105] This part of the program is stored in / home / wind_gust_algorithm / GDWRInterface, which is mainly responsible for calling each module. This module contains a python file, run_GDW_wind_main.py, which is the main program of the whole method.

[0106] (2) Module 2

[0107] The module code is stored in / home / wind_gust_algorithm / GDWRInterface / models. It is the core program of the PhyDnet model and contains a code file phydnet_att.py. This module is responsible for reading the model pre-training parameters stored in GDWRInterface / weight, encoding and decoding data, loading related deep learning libraries, and generating wind speed forecast results.

[0108] (3) Module 3

[0109] This part of the program is stored in / home / wind_gust_algorithm / GDWRInterface / utils / , which is mainly responsible for downloading forecast data and various types of observation data, as well as time-space interpolation and multi-scale variational analysis. The module contains two program files, __init__.py and stmas_lib.f90. Among them, __init__.py implements the downloading and time-space interpolation of forecast data and various types of observation data, and stmas_lib.f90 is the main program for multi-scale variational analysis. The code file call is as follows Figure 3 shown.

[0110] S4. Obtain the extended orthogonal empirical decomposition gust extrapolation forecast result.

[0111] This method is based on HRCLDAS multi-source fusion data, numerical forecast products, and station wind speed observations. It uses the extended orthogonal empirical decomposition method to establish a forecast model, realize short-term forecast of grid point wind speed, and quickly obtain a 1km resolution gust forecast product with a frequency of 1 hour and 10 minutes in the future in western Guangdong. The detailed calculation process of the method is as follows:

[0112] (1) Download historical HRCLDAS data and site wind speed observations, use the extended orthogonal empirical decomposition method to train the model parameters, and store them locally. The construction of the extended orthogonal empirical decomposition method wind speed forecast model is mainly divided into the following three steps:

[0113] ①Construction of wind speed covariance matrix

[0114] Unlike scalar fields, wind fields are expressed as binary vectors, usually expressed as meridional wind components and zonal wind components. The covariance of a vector is not only related to the magnitude of the vector, but also to the angle, so it needs to be expressed through a vector. This project uses the cosine of the two components of the wind field as the definition of the wind field covariance, and the specific formula is as follows:

[0115] f=v a 2 +v b 2 -2v a v b ·cosγ

[0116] This scheme takes into account the respective modes of the two vectors and their angles, so that the subsequent typical feature extraction of the wind field propagation process based on this takes into account both wind speed and wind direction, and has a higher degree of discrimination.

[0117] ②Simulation of the excitation and propagation process of gusts based on the extended empirical orthogonal decomposition method

[0118] The wind speed field covariance matrix is ​​generated by the time window sliding method:

[0119]

[0120] The excitation and propagation processes of typical gusts are solved through matrix eigenvalues ​​and eigenvectors.

[0121]

[0122] This method can ensure that different typical processes are orthogonal, and the first several modes can explain most of the variance, that is, the typical process is representative.

[0123] ③Build a forecast model

[0124] On the basis of the extended empirical orthogonal decomposition of the gridded wind speed field, the similarity between the wind field evolution process and each typical process in the current window period can be analyzed, and the path of subsequent gust evolution can be obtained based on the similarity time coefficients of different typical processes and the historical data. The prediction of the time coefficient is realized by the long short-term memory network (Long Short-Term Memory, LSTM). The flowchart of the extended orthogonal empirical decomposition method is shown in the attached figure. Figure 5 shown.

[0125] (2) Download the HRCLDAS multi-source fusion near-real-time wind field data, numerical forecast data and site wind speed observation data at the nearest time.

[0126] (3) Using time interpolation, HRCLDAS and numerical forecast data are interpolated to the time of the observed data.

[0127] (4) Based on multi-scale variational analysis, the interpolation results of ② are corrected using station wind speed observations to obtain near-real-world grid wind speed data.

[0128] (5) Load the pre-trained model parameters, input the data obtained in step ③, and make wind speed forecasts.

[0129] (6) Output the final wind field extrapolation forecast.

[0130] Code structure: The method code is mainly divided into two functional modules:

[0131] (1) Module 1

[0132] This part of the program is stored in / home / wind_gust_algorithm / hrcldas_extrapolation / src, which is mainly responsible for data download, reading and writing, spatiotemporal interpolation and calling module 2. This module contains 3 python codes, namely: run_hrcldas_extrapolation_main.py, AlgorithmLib.py, and DataInterface.py. Among them, run_hrcldas_extrapolation_main.py is the main program of the method, DataInterface.py is responsible for data download and reading and writing, and Algorithm.py is responsible for calling module 2, reading pre-trained model parameters and data spatiotemporal interpolation.

[0133] (2) Module 2

[0134] This part of the program is stored in / home / wind_gust_algorithm / hrcldas_extrapolation / src / , which is mainly responsible for multi-scale variational analysis and fusion of different data. The module contains a program file, stmas_lib.f90, which implements the multi-scale variational analysis method and is ultimately called by Algorithm.py in module 1. The code file call is as shown in the attached file. Figure 4 shown.

[0135] Output file description

[0136] The output results of the four methods are all in netcdf format. The variables related to wind speed short-term forecast in the file are described as follows:

[0137] lon: longitude information; lat: latitude information; time: forecast time; u10_fcst: zonal wind forecast, three-dimensional array (time, latitude, longitude); v10_fcst: radial wind forecast, three-dimensional array (time, latitude, longitude).

[0138] The method also includes step S5, comprehensively applying the forecast results of steps S1-S4 according to needs to obtain a final gust short-term weather forecast.

[0139] This method implements the latest deep learning gust forecasting method, and innovatively designs a gust statistical forecasting method based on empirical orthogonal decomposition, a gust observation extrapolation method based on optical flow method and wind field diagnostic downscaling model, and realizes gust diagnostic downscaling forecast for traditional numerical forecast products. Taking into account the characteristics of different methods (such as basic data download time, model calculation time, forecast accuracy, data capacity, etc.), the specific implementation process of the method is optimized, and finally the current 4 gust short-term forecasting methods and their implementation methods are obtained.

[0140] All methods are deployed on the 117 server, in the directory / home / wind_gust_algorithm, and can be scheduled through the method scheduling platform. The four short-term gust forecast products obtained by the four methods can be used in a comprehensive manner (such as weighted summation of results, extreme value warning, etc.) to improve the gust forecast service capabilities in western Guangdong.

[0141] As another implementation aspect, the invention research content is:

[0142] (1) Based on meteorological station data, the mesoscale numerical model (WRF), the microscale diagnostic model (CALMET), and the data assimilation algorithm, a wind field dataset with a resolution of 1 km from 2018 to 2020 was developed.

[0143] (2) Based on the 1 km resolution wind field dataset, historical gale cases were selected, and a gale event relationship model was established grid by grid point, and a gust warning model based on historical wind field data was constructed.

[0144] (3) Based on the gust warning model, the optimized CALMET mode, the short-term wind field extrapolation algorithm, and the determination of high wind areas based on radar echo extrapolation, a comprehensive analysis is conducted to obtain a high wind statistics-extrapolation fusion short-term forecast product.

[0145] As another implementation aspect, the invention technical route is:

[0146] (1) Site observation data collation

[0147] The hourly temperature, wind field, and relative humidity data of ground meteorological observation stations in Guangdong Province from 2018 to 2020 were extracted, and quality control was carried out using climate extreme value test, temporal continuity test, and spatial continuity test.

[0148] (2) WRF model basic model dataset production

[0149] Regional climate simulation is carried out based on the WRF model. During the simulation, the FDDA assimilation algorithm is used to perform hourly data assimilation correction on the model data, and the final output is a 10-minute basic model data set.

[0150] (3) CALMET diagnostic mode code optimization and upgrade

[0151] When carrying out (2), the CALMET diagnostic model was simultaneously transformed and upgraded. By optimizing the code structure and setting the parallel algorithm, the optimized CALMET diagnostic model (referred to as CALMET-OpenMP) was obtained.

[0152] (4) Preparation of refined wind field diagnostic data set

[0153] Based on the WRF 10-minute basic data set and CALMET-OpenMP, a downscaling diagnostic simulation was carried out to obtain a refined wind field diagnostic data set with a 10-minute and 1km resolution for the whole Guangdong Province.

[0154] (5) Production of a typical gale case diagnosis fusion dataset

[0155] Based on the refined wind field diagnosis dataset obtained in (1) and (4) and the wind field observations of ground meteorological stations, typical strong wind cases are selected and fused and corrected again through the multi-scale variational assimilation algorithm to obtain the final typical strong wind case diagnosis fusion dataset.

[0156] (6) Research and development of high wind statistical forecast model

[0157] Based on the typical gale case diagnosis fusion data set obtained in (5), a gale statistical forecast model is established grid by grid. The model can eventually diagnose and forecast the wind field of grid points within a specific range and infer the possible wind speed based on the grid wind field data.

[0158] (7) Development of high wind statistical forecast products

[0159] Based on the CALMET-OpenMP obtained in (3), the real-time wind field with a resolution of 1 km near the ground is diagnosed using the meteorological observations every ten minutes as input. The diagnosed wind field is substituted into the high wind statistical forecast model obtained in (6), and the real-time high wind statistical forecast is carried out to obtain the high wind forecast product A.

[0160] (8) Research on downscaling of local model forecasts

[0161] Based on CALMET-OpenMP and the local model forecast data of Guangdong Meteorological Bureau, downscaling simulation was carried out to obtain hourly refined wind field diagnosis and forecast data.

[0162] (9) Mesoscale wind forecast time encryption

[0163] Based on the time interpolation algorithm, the local mesoscale model forecast data of the Guangdong Provincial Meteorological Bureau is time-encrypted to obtain a 10-minute mesoscale wind field forecast.

[0164] (10) Development of short-term extrapolated forecast products for strong winds

[0165] Based on the hourly refined wind field diagnostic forecast obtained in (8), the 10-minute mesoscale wind field forecast obtained in (9) is used as the tendency item to carry out the 10-minute refined wind field extrapolation forecast to obtain the strong wind forecast product B.

[0166] (11) Development of extrapolated forecast products for severe convection and strong wind areas

[0167] The 10-minute mesoscale wind field forecast obtained in (9) is used as the tendency item, combined with the optical flow method, radar echo extrapolation is carried out based on the real-time radar reflectivity data, and the scope of the strong wind area is determined according to the extrapolation results to obtain the strong wind forecast product C.

[0168] (12) Production of fusion products of strong wind statistics and extrapolated short-term forecast

[0169] The strong wind forecast product A obtained in (7), the strong wind forecast product B obtained in (10), and the strong wind forecast product C obtained in (11) are integrated to obtain the final strong wind statistics-extrapolation fusion short-term forecast product D, which is the final business product output.

[0170] Key technical points: (1) Development of basic model data set based on WRF model: It is necessary to carry out 3-year climate simulation based on WRF model and ground meteorological station data. It requires a lot of manpower and computer time. (2) CALMET model optimization and transformation: It is necessary to disassemble, optimize and parallelize the CALMET model code. It requires a lot of working time. (3) Production of strong wind case diagnosis fusion data set "This step needs to be based on the premise that (1) and (2) are completed. First, the improved CALMET is used to perform downscaling diagnostic simulation on the WRF model results, and then multi-scale variation is used for secondary fusion correction. (4) Development of grid strong wind statistical forecast model: The data used needs to be based on (3), and the forecast model is currently an unknown. (5) Research on mesoscale model time encryption algorithm: It is necessary to develop a mesoscale numerical model time encryption algorithm to encrypt the hourly forecast output into Ten minutes at a time. (6) Research on refined short-term wind field extrapolation algorithm based on mesoscale wind field tendency forecast: It needs to be established on the basis of (2) and (5), and research on short-term wind field extrapolation algorithm should be carried out. (7) Research on radar reflectivity extrapolation algorithm based on mesoscale wind field tendency forecast and determination of high wind area: It needs to be established on the basis of (5), and research on short-term wind field extrapolation algorithm should be carried out. (8) Real-time grid wind field data production and high wind statistical forecast based on CALMET and actual meteorological observation data: It needs to be established on the basis of (2) and (4).

[0171] As another implementation aspect, the present invention also provides a gust short-term weather forecast / warning system, the technical route of which is as shown in the attached Figure 6 shown.

[0172] In the several embodiments provided in the present application, it should be understood that the disclosed client can be implemented in other ways. Among them, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of units or units can be electrical or other forms.

[0173] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0174] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A method for short-term gust weather forecasting, characterized in that: It includes the following steps: S1. Obtain wind field downscaling forecast results; S2. Obtaining the extrapolated forecast result of the actual gust situation; S3. Obtain deep learning gust extrapolation forecast results; S4. Obtain the extended orthogonal empirical decomposition gust extrapolation forecast result.

2. The gust short-term weather forecast method according to claim 1, characterized in that: The method further comprises step S5, comprehensively applying the forecast results of steps S1-S4 according to requirements to obtain a final gust short-term weather forecast.

3. The method for short-term gust weather forecasting according to claim 1, characterized in that: Step S1: Obtain wind field downscaling forecast results, specifically: S11. Download the forecast data of ground and atmospheric meteorological elements for the next 2 hours; S12, converting the data into a WPS intermediate file, and generating an intermediate file recognizable by WRF in combination with the terrain data; S13, using the WRF mode, reading the intermediate file and generating a data file that can be recognized by the CALWRF program; S14. Use CALWRF to generate a data file that can be recognized by CALMET; S15. Use CALMET model to perform wind field downscaling diagnosis; S16. Output the final wind field downscaling forecast result.

4. The method for short-term gust weather forecasting according to claim 1, characterized in that: Step S2: Obtaining the actual extrapolated forecast result of gusts, specifically: S21. Download the numerical forecast products, station wind speed observation data and radar reflectivity data of the near time; S22, using interpolation to unify different data times; S23, calculating the radar reflectivity motion vector based on the optical flow method; S24, based on multi-scale variational analysis, the station wind speed observation, radar reflectivity motion vector and numerical forecast are integrated; S25. Downscaling diagnosis is performed on the fused data based on the CALMET model to obtain a near-realistic near-surface wind field that complies with the law of conservation of mass; S26. Use the near-surface wind field to adjust the wind field predicted by the numerical model; S27. Output the final wind field extrapolation forecast.

5. The method for short-term gust weather forecasting according to claim 1, characterized in that: Step S3: Obtain the deep learning gust extrapolation forecast result, specifically: S31, based on historical radar reflectivity and site wind speed observations, the model parameters are trained using the PhyDnet method and stored locally; S32. Download the numerical forecast products, station wind speed observation data and radar reflectivity data of the near time; S33, using interpolation method to unify different data times; S34, read the pre-trained model parameter file, use the site wind speed observation and radar reflectivity as input, and generate a site wind speed forecast for the next 2 hours at a frequency of 10 minutes; S35. Use multiscale variational analysis to fuse site wind speed forecasts with model forecasts; S36. Output the final wind field extrapolation forecast.

6. The method for short-term gust weather forecasting according to claim 1, characterized in that: Step S4, obtaining the extended orthogonal empirical decomposition gust extrapolation forecast result, specifically: S41, download historical HRCLDAS data and site wind speed observations, use the extended orthogonal empirical decomposition method to train the model parameters, and store them locally; S42, downloading the HRCLDAS multi-source fusion near-real-time wind field data and numerical forecast data and site wind speed observation data at the near time; S43, using time interpolation, interpolating the HRCLDAS and numerical forecast data to the time of the observed data; S44, based on multi-scale variational analysis, the interpolation results of the station wind speed observations are used for correction to obtain near-real-world grid wind speed data; S45, loading the pre-trained model parameters, inputting the obtained data, and performing wind speed forecasting; S46. Output the final wind field extrapolation forecast.

Images