A severe convective gale nowcasting method and system based on upstream actual wind speed

Abstract

The present invention belongs to the field of gale prediction and production, and specifically provides a method and system for the near-forecasting of severe convective gale based on the upstream actual wind speed. Based on continuous multi-scene radar observation, the optical flow field (v) of the combined reflectivity radar echo in different time dimensions is calculated by the optical flow method, and the radar echo observation of 6 minutes is mapped to the radar observation of 10 minutes respectively, and the optical flow fields of different time dimensions are fused by dynamic weights, and finally the radar echo extrapolation of 10 minutes interval in the next 2 hours is obtained. Through the algorithm of XGBOOST-CNN, the mapped live radar observation of 10 minutes interval is combined with the ground 10-minute level gale live station monitoring, and the terrain information and underlying surface information are integrated to establish a small sample echo feature-wind speed inversion model based on machine learning technology, and obtain a real-time wind speed inversion model. An innovative grid-based near-forecasting technology for severe convective gale is proposed. Thunderstorm gale products for 10 minutes in the next 2 hours can be provided.

Description

Technical Field

[0001] The present invention relates to the technical field of gale prediction, and more specifically, to a method and system for predicting severe convective gale based on upstream actual wind speed. Background Art

[0002] Severe convection can trigger a series of disasters, including short-term strong winds. For such strong winds, the only effective means of observation is ground station observation. But the problem is that when the ground station observes, the strong wind event has already occurred, and it is impossible to provide early warning. This kind of strong convection and strong wind phenomenon is accompanied by the movement of strong convective clouds. The current monitoring methods for strong convective clouds are diverse. The most common one is based on Doppler weather radar, but the intensity of radar echoes is not a simple linear relationship with wind speed, and the formation process and movement speed of this convection are very fast, which cannot be obtained from numerical weather forecasts (NWP) that require a lot of calculations. Therefore, a method that can quickly establish an accurate nonlinear relationship is urgently needed.

[0003] In recent years, with the advancement of science and technology, artificial intelligence, as a black box technology based on big data statistics, has begun to develop rapidly in the computer field and gradually spread to all walks of life. At present, many methods based on artificial intelligence have been applied to short-term forecasting and have achieved many results. However, the current short-term forecasting technology for radar focuses more on radar echoes or precipitation, and rarely involves short-term strong wind weather phenomena. How to apply artificial intelligence to strong wind forecasting has broad prospects for practical engineering applications. Summary of the invention

[0004] The present invention aims to solve the technical problem that it is difficult to predict strong winds in the prior art.

[0005] The present invention provides a method for nowcasting severe convective gale based on upstream actual wind speed, comprising the following steps:

[0006] S1, based on continuous multi-scene radar observations, the optical flow field of the combined reflectivity radar echo in different time dimensions is calculated by the optical flow method;

[0007] S2, respectively mapping the radar echo observation at the first time to the radar observation at the second time to obtain the real-time radar observation, and fusing the optical flow fields of different time dimensions through dynamic weights to obtain the radar echo extrapolation at intervals of the second time length in a future time period;

[0008] S3, through the XGBOOST-CNN algorithm, the mapped live radar observations and ground gale live station monitoring information are integrated with terrain information and underlying surface information to establish a real-time wind speed inversion model based on machine learning;

[0009] S4, applying the real-time wind speed inversion model to the radar echo extrapolation result, and finally obtaining a wind speed extrapolation product predicted at intervals of a second time length in a future time period.

[0010] Preferably, the first time is 6 minutes, the second time is 10 minutes, and the time period is two hours.

[0011] Preferably, the S1 specifically includes:

[0012] Based on the multi-elevation 3D radar base data reading method, the polar coordinates are converted into the longitude and latitude coordinate system, and re-interpolated to a 1km grid resolution to obtain radar echo data at multiple elevation angles [5°, 15°, 24°, 34°, 43°, 60°];

[0013] The re-interpolated radar echo data of multiple elevation angles are matched one by one according to the longitude and latitude grids, and the highest reflectivity value at the corresponding longitude and latitude is found. The final maximum reflectivity value in the two-dimensional space is the combined reflectivity (CR).

[0014] Then, the dense inverse search algorithm is used to calculate the radar images of two consecutive scenes, and finally the optical flow field, that is, the velocity field, of the entire scene radar echo motion is obtained.

[0015] Preferably, the mapping in S2 is a time scale remapping, specifically including: 00→00, 06-12→10, 18-24→20, 30→30, 36-42→40, 48-54→50;

[0016] Among them, the number before the arrow is the radar observation batch, and the number after the arrow is the ground station observation batch.

[0017] Preferably, S2 specifically includes: taking into account the positions of multi-scene radar echoes in the past, calculating the optical flow field at a certain historical moment and a nearby moment, and calculating the optical flow field of different forms at each extrapolated time by introducing influence weights, so as to obtain the radar extrapolation result for the next 2 hours and 10 minutes.

[0018] Preferably, the S3 specifically includes:

[0019] S31, find the radar echo situation within a 5km×5km range above the ground station. For each observation station, a 5×5 grid of radar echoes and the corresponding 10-minute maximum wind speed are used as samples. From the samples, remove the points where the sum of the 5×5 radar echo samples is less than 20, and remove the samples where the maximum wind record exceeds 45m / s. All radar observations and ground station observations that meet the above requirements are used as samples;

[0020] S32, the data is cut into five equal parts and put into the XGBOOST and CNN models. For each model algorithm, 4 parts are selected as training sets and 1 part is selected as verification set. The models are trained and verified to obtain 5 models, namely {XGBOOST1, XGBOOST2, XGBOOST3, XGBOOST4, XGBOOST5, CNN1, CNN2, CNN3, CNN4, CNN5}. The maximum value of the model is obtained, and the real-time wind speed inversion model is MAX (MODEL{10}).

[0021] Preferably, S4 specifically includes: cutting the radar data (1000×1000) of all times of the extrapolation time into small blocks of 5×5, inputting into the real-time wind speed inversion model MAX (MODEL{10}) fitted by historical data at the current time, inverting only the area with radar echo within the 5×5 range, and finally outputting the wind speed product of echo inversion;

[0022] For the next start time, the model needs to refit the new relationship as the new model MAX(MODEL{10})new.

[0023] The present invention also provides a severe convective gale nowcasting system based on upstream actual wind speed, characterized in that the detection and extraction system is used to implement a severe convective gale nowcasting method based on upstream actual wind speed, comprising:

[0024] Radar preprocessing module, used to calculate the optical flow field of combined reflectivity radar echoes in different time dimensions based on continuous multi-scene radar observations by optical flow method;

[0025] The optical flow interpolation and extrapolation module is used to map the radar echo observation at the first time to the radar observation at the second time to obtain the real-time radar observation, and fuse the optical flow fields of different time dimensions through dynamic weights to obtain the radar echo extrapolation at intervals of the second time length in the future time period;

[0026] The machine learning module is used to use the XGBOOST-CNN algorithm to map the live radar observations and ground gale live station monitoring information, and integrate the terrain information and underlying surface information to establish a real-time wind speed inversion model based on machine learning;

[0027] The product generation module is used to apply the real-time wind speed inversion model to the radar echo extrapolation result, and finally obtain the wind speed extrapolation product predicted at the interval of the second time length in the future time period.

[0028] The present invention also provides an electronic device, comprising a memory and a processor, wherein the processor is used to implement the steps of a method for predicting severe convective gale based on upstream actual wind speed when executing a computer management program stored in the memory.

[0029] The present invention also provides a computer-readable storage medium on which a computer management program is stored. When the computer management program is executed by a processor, the steps of a method for predicting severe convective gale based on upstream actual wind speed are implemented.

[0030] Beneficial effect: The present invention provides a method and system for the near-term prediction of severe convective gale based on the upstream actual wind speed. Based on continuous multi-scene radar observations, the optical flow field (v) of the combined reflectivity radar echo in different time dimensions is calculated by the optical flow method, and the 6-minute radar echo observations are respectively mapped to the 10-minute radar observations, and the optical flow fields in different time dimensions are fused through dynamic weights, and finally the radar echo extrapolation of the next 2 hours and 10-minute intervals is obtained. Through the XGBOOST-CNN algorithm, the mapped 10-minute interval real-time radar observations are compared with the ground 10-minute level gale real-time station monitoring, and the terrain information and underlying surface information are integrated to establish a small sample echo feature-wind speed inversion model based on machine learning technology to obtain a real-time wind speed inversion model. The real-time wind speed inversion model is applied to the radar echo extrapolation results to finally obtain the wind speed extrapolation product of the next 2 hours and 10-minute intervals. Based on real-time radar echo detection data and strong wind monitoring information from meteorological stations (automatic stations, national stations), an innovative grid-based nowcasting technology for strong convective winds is proposed through the combined application of real-time machine learning and optical flow methods. It can provide thunderstorm strong wind products for every 10 minutes in the next 2 hours. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A flow chart of a severe convective gale nowcasting method based on upstream actual wind speed provided by the present invention;

[0032] Figure 2 A schematic diagram of the hardware structure of a possible electronic device provided by the present invention;

[0033] Figure 3 A schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention;

[0034] Figure 4 A framework diagram of a severe convective gale nowcasting system based on upstream actual wind speed provided by the present invention;

[0035] Figure 5 A comparison chart of the optical flow method extrapolation and actual situation in the patent provided by the present invention;

[0036] Figure 6 A comparison chart of the site results of the model inversion provided by the present invention and the actual observation results. DETAILED DESCRIPTION

[0037] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0038] Figure 1 The present invention provides a severe convective gale nowcasting method based on upstream actual wind speed, comprising the following steps:

[0039] S1, based on continuous multi-scene radar observations, the optical flow field of the combined reflectivity radar echo in different time dimensions is calculated by the optical flow method;

[0040] S2, respectively mapping the radar echo observation at the first time to the radar observation at the second time to obtain the real-time radar observation, and fusing the optical flow fields of different time dimensions through dynamic weights to obtain the radar echo extrapolation at intervals of the second time length in a future time period;

[0041] S3, through the XGBOOST-CNN algorithm, the mapped live radar observations and ground gale live station monitoring information are integrated with terrain information and underlying surface information to establish a real-time wind speed inversion model based on machine learning;

[0042] S4, applying the real-time wind speed inversion model to the radar echo extrapolation result, and finally obtaining a wind speed extrapolation product predicted at intervals of a second time length in a future time period.

[0043] This scheme is based on continuous multi-scene radar observations. The optical flow method is used to calculate the optical flow field (v) of the combined reflectivity radar echo in different time dimensions, and the 6-minute radar echo observations are mapped to the 10-minute radar observations, and the optical flow fields in different time dimensions are fused through dynamic weights, and finally the radar echo extrapolation for the next 2 hours and 10-minute intervals is obtained. Through the XGBOOST-CNN algorithm, the mapped 10-minute interval live radar observations are combined with the ground 10-minute level gale live station monitoring, and the terrain information and underlying surface information are integrated to establish a small sample 'echo feature-wind speed' inversion model based on machine learning technology to obtain a real-time wind speed inversion model. The real-time wind speed inversion model is applied to the radar echo extrapolation results to finally obtain the wind speed extrapolation product for the next 2 hours and 10-minute intervals.

[0044] In a specific implementation scenario, data processing is first performed through a radar preprocessing module, which specifically includes:

[0045] Reading of multi-elevation 3D radar base data: The radar data storage format is in the form of polar coordinates. The data needs to be converted into coordinates, and the polar coordinates are converted into longitude and latitude coordinate systems, and re-interpolated to a grid with a resolution of 1km to obtain radar echo data with multiple elevation angles [5°, 15°, 24°, 34°, 43°, 60°].

[0046] Combined reflectivity synthesis: The re-interpolated multi-elevation radar echo data are matched one by one according to the longitude and latitude grids, and the highest reflectivity value at the corresponding longitude and latitude is found. The final maximum reflectivity value in the two-dimensional space is the combined reflectivity (CR).

[0047] Then the optical flow method is the dense inverse search algorithm, which calculates two consecutive radar images and finally obtains the optical flow field of the entire radar echo motion, that is, I i+delta -I i =V 0 (Velocity field).

[0048] In a preferred solution, the mapping in S2 is a time scale remapping: the radar observation time is 6 minutes per scene. In order to establish a machine learning relationship with the 10-minute ground station in the later stage, it is necessary to use the optical flow method to perform time scale interpolation on the radar. The corresponding relationship is as follows (the batch before the arrow is the radar observation batch, and the batch after the arrow is the ground station observation batch), 00→00, 06~12→10, 18~24→20, 30→30, 36~42→40, 48~54→50.

[0049] I i+06 -I i =V

[0050]

[0051] Multi-time (i.e. different time) dimension dynamic weighted optical flow method: The motion of radar echo itself is complex and not a simple linear motion. The optical flow field calculated from two consecutive adjacent time radar echoes cannot describe the overall motion of the radar. It mainly depicts the movement direction of the strong value area in a short time. For the overall convective cloud cluster, this evaluation of the movement trend is inaccurate, and it is necessary to consider the echo motion trajectory from a larger time dimension. In the embodiment of the present application, the position of multi-scene radar echoes in the past is considered, and the optical flow field at a certain moment in history and the nearby moment is calculated. By introducing influence weights, the optical flow field of different forms at each extrapolated time is calculated, thereby obtaining the radar extrapolation results for the next 2 hours and 10 minutes. The specific calculation formula is shown in Table 1 below:

[0052] Table 1 Calculation formula

[0053]

[0054] Where V is the calculated optical flow velocity field, I is the radar echo, t is the current time, t+n is the future forecast time, and tn is the historical observation time.

[0055] The preferred solution to establish a real-time wind speed inversion model based on machine learning specifically includes:

[0056] S31, sample collection: based on all ground stations, find the radar echo within 5km×5km above the ground station. Since the spatial resolution of radar echo is 1km, for each observation site, there is a 5×5 grid of radar echo and the corresponding 10-minute maximum wind speed as samples. From the samples, remove the points where the sum of 5×5 radar echo samples is less than 20, and remove the samples where the maximum wind record exceeds 45m / s. All radar and ground station observations that meet the above requirements are taken as samples.

[0057] S32, based on XGBOOST / CNN, a small sample "echo feature-wind speed" inversion model is established. The data is cut into five equal parts and put into the XGBOOST and CNN models. For each model algorithm, 4 parts are selected as training sets and 1 part is used as verification set. The model is trained and verified to obtain 5 models, namely {XGBOOST1, XGBOOST2, XGBOOST3, XGBOOST4, XGBOOST5, CNN1, CNN2, CNN3, CNN4, CNN5}. The maximum value of the model is obtained, that is, the obtained model is MAX(MODEL{10}).

[0058] Product generation module: The radar data (1000×1000) of all extrapolated times are cut into 5×5 small blocks and input into the model MAX(MODEL{10}) fitted by historical data at the current time. Only the areas with radar echoes within the 5×5 range are inverted. The final data output is the wind speed product of echo inversion. For the next reporting time, the model needs to be refitted with a new relationship as the new model MAX(MODEL{10})new.

[0059] The following is a test and comparative analysis of the algorithm, comparing the optical flow field extrapolation method based on different time dimensions of the embodiment of the present invention with the traditional optical flow method, and inverting the wind speed in the process and comparing it with the observation.

[0060] The optical flow method of the application embodiment of the present invention is compared with other optical flow methods, and the final results are shown in Table 2 below:

[0061] Table 2 Comparison of different optical flow methods

[0062]

[0063] It can be seen that the mean absolute error and root mean square error of the optical flow method of the application embodiment of the present invention are both the lowest, which has an unexpected effect.

[0064] like Figure 5 and Figure 6 As shown, it can be seen that the actual situation is very close to the extrapolation of the optical flow method in the embodiment of the present invention, which shows that the prediction effect of the approaching strong wind is good and consistent with the actual situation. The hit rate is very high, which shows that this method has a good hit rate for strong wind areas. The hit rate calculation formula is:

[0065] Figure 6 The hit rates are shown.

[0066] This scheme has established a short-term forecasting technology for severe convective gale based on upstream actual wind speed and real-time radar detection, which has shown good results in forecasting radar echo extrapolation and forecasting gale impact areas, and can be used to forecast short-term severe convective gale in some local areas, providing a relatively simple and quick method for short-term gale forecasting.

[0067] like Figure 4 As shown, the embodiment of the present invention further provides a severe convective gale nowcasting system based on the upstream actual wind speed, and the detection and extraction system is used to implement the severe convective gale nowcasting method based on the upstream actual wind speed, including:

[0068] Radar preprocessing module, used to calculate the optical flow field of combined reflectivity radar echoes in different time dimensions based on continuous multi-scene radar observations by optical flow method;

[0069] The optical flow interpolation and extrapolation module is used to map the radar echo observation at the first time to the radar observation at the second time to obtain the real-time radar observation, and fuse the optical flow fields of different time dimensions through dynamic weights to obtain the radar echo extrapolation at intervals of the second time length in the future time period;

[0070] The machine learning module is used to use the XGBOOST-CNN algorithm to map the live radar observations and ground gale live station monitoring information, and integrate the terrain information and underlying surface information to establish a real-time wind speed inversion model based on machine learning;

[0071] The product generation module is used to apply the real-time wind speed inversion model to the radar echo extrapolation result, and finally obtain the wind speed extrapolation product predicted at the interval of the second time length in the future time period.

[0072] See also Figure 2 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 2As shown, an embodiment of the present invention provides an electronic device, including a memory 1310, a processor 1320, and a computer program 1311 stored in the memory 1310 and executable on the processor 1320. When the processor 1320 executes the computer program 1311, the following steps are implemented: S1, based on continuous multi-scene radar observation, the optical flow field of the combined reflectivity radar echo in different time dimensions is calculated by the optical flow method;

[0073] S2, respectively mapping the radar echo observation at the first time to the radar observation at the second time to obtain the real-time radar observation, and fusing the optical flow fields of different time dimensions through dynamic weights to obtain the radar echo extrapolation at intervals of the second time length in a future time period;

[0074] S3, through the XGBOOST-CNN algorithm, the mapped live radar observations and ground gale live station monitoring information are integrated with terrain information and underlying surface information to establish a real-time wind speed inversion model based on machine learning;

[0075] S4, applying the real-time wind speed inversion model to the radar echo extrapolation result, and finally obtaining a wind speed extrapolation product predicted at intervals of a second time length in a future time period.

[0076] See also Figure 3 Schematic diagram of an embodiment of a computer-readable storage medium provided by the present invention. Figure 3 As shown, this embodiment provides a computer-readable storage medium 1400 on which a computer program 1411 is stored. When the computer program 1411 is executed by a processor, the following steps are implemented: S1, based on continuous multi-scene radar observation, calculating the optical flow field of the combined reflectivity radar echo in different time dimensions by the optical flow method;

[0077] S2, respectively mapping the radar echo observation at the first time to the radar observation at the second time to obtain the real-time radar observation, and fusing the optical flow fields of different time dimensions through dynamic weights to obtain the radar echo extrapolation at intervals of the second time length in a future time period;

[0078] S3, through the XGBOOST-CNN algorithm, the mapped live radar observations and ground gale live station monitoring information are integrated with terrain information and underlying surface information to establish a real-time wind speed inversion model based on machine learning;

[0079] S4, applying the real-time wind speed inversion model to the radar echo extrapolation result, and finally obtaining a wind speed extrapolation product predicted at intervals of a second time length in a future time period.

[0080] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0081] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0082] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0083] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0084] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0085] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0086] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A severe convective gale nowcasting method based on upstream actual wind speed, characterized in that: The following steps are involved: S1, based on continuous multi-scene radar observations, the optical flow field of the combined reflectivity radar echo in different time dimensions is calculated by the optical flow method; S2, respectively mapping the radar echo observation at the first time to the radar observation at the second time to obtain the real-time radar observation, and fusing the optical flow fields of different time dimensions through dynamic weights to obtain the radar echo extrapolation at intervals of the second time length in a future time period; S3, through the XGBOOST-CNN algorithm, the mapped live radar observations and ground gale live station monitoring information are integrated with terrain information and underlying surface information to establish a real-time wind speed inversion model based on machine learning; S4, applying the real-time wind speed inversion model to the radar echo extrapolation result, and finally obtaining a wind speed extrapolation product predicted at intervals of a second time length in a future time period.

2. The severe convective gale nowcasting method based on upstream actual wind speed according to claim 1, characterized in that: The first time is 6 minutes, the second time is 10 minutes, and the time period is two hours.

3. The severe convective gale nowcasting method based on upstream actual wind speed according to claim 1, characterized in that: The S1 specifically includes: Based on the multi-elevation 3D radar base data reading method, the polar coordinates are converted into the longitude and latitude coordinate system, and re-interpolated to a 1km grid resolution to obtain radar echo data at multiple elevation angles [5°, 15°, 24°, 34°, 43°, 60°]; The re-interpolated radar echo data of multiple elevation angles are matched one by one according to the longitude and latitude grids, and the highest reflectivity value at the corresponding longitude and latitude is found. The final maximum reflectivity value in the two-dimensional space is the combined reflectivity. Then, the radar images of two consecutive scenes are calculated through a dense inversion search algorithm, and finally the optical flow field, that is, the velocity field, of the entire radar echo motion is obtained.

4. The severe convective gale nowcasting method based on upstream actual wind speed according to claim 1, characterized in that: The mapping in S2 is a time scale remapping, specifically including: 00→00, 06~12→10, 18~24→20, 30→30, 36~42→40, 48~54→50; Among them, the number before the arrow is the radar observation batch, and the number after the arrow is the ground station observation batch.

5. The severe convective gale nowcasting method based on upstream actual wind speed according to claim 1, characterized in that: The S2 specifically includes: taking into account the positions of multi-scene radar echoes in the past, calculating the optical flow field at a certain historical moment and a nearby moment, and calculating the optical flow field of different forms at each extrapolated time by introducing influence weights, thereby obtaining the radar extrapolation result for the next 2 hours and 10 minutes.

6. The severe convective gale nowcasting method based on upstream actual wind speed according to claim 1, characterized in that: The S3 specifically includes: S31, search for radar echo conditions within a 5km×5km range above the ground station. For each observation station, a 5×5 grid of radar echoes and the corresponding 10-minute maximum wind speed are used as samples. From the samples, remove points where the sum of 5×5 radar echo samples is less than 20, and remove samples where the maximum wind speed record exceeds 45m / s. All radar observations and ground station observations that meet the above requirements are used as samples; S32, the data is cut into five equal parts and put into the XGBOOST and CNN models. For each model algorithm, 4 parts are selected as training sets and 1 part is selected as verification set. The models are trained and verified to obtain 5 models, namely {XGBOOST1, XGBOOST2, XGBOOST3, XGBOOST4, XGBOOST5, CNN1, CNN2, CNN3, CNN4, CNN5}. The maximum value of the model is obtained, and the real-time wind speed inversion model is MAX (MODEL{10}).

7. The severe convective gale nowcasting method based on upstream actual wind speed according to claim 6, characterized in that: S4 specifically includes: cutting the radar data of all times of the extrapolated time (1000×1000) into small blocks of 5×5, inputting them into the real-time wind speed inversion model MAX (MODEL{10}) fitted by historical data at the current time, inverting only the area with radar echo within the 5×5 range, and finally outputting the wind speed product of echo inversion; For the next start time, the model needs to refit the new relationship as the new model MAX(MODEL{10})new.

8. A severe convective wind nowcasting system based on upstream real-time wind speed, characterized in that: The severe convective gale nowcasting system is used to implement the severe convective gale nowcasting method based on upstream actual wind speed according to any one of claims 1 to 7, comprising: Radar preprocessing module, used to calculate the optical flow field of combined reflectivity radar echoes in different time dimensions based on continuous multi-scene radar observations by optical flow method; The optical flow interpolation and extrapolation module is used to map the radar echo observation at the first time to the radar observation at the second time to obtain the real-time radar observation, and fuse the optical flow fields of different time dimensions through dynamic weights to obtain the radar echo extrapolation at intervals of the second time length in the future time period; The machine learning module is used to use the XGBOOST-CNN algorithm to map the live radar observations and ground gale live station monitoring information, and integrate the terrain information and underlying surface information to establish a real-time wind speed inversion model based on machine learning; The product generation module is used to apply the real-time wind speed inversion model to the radar echo extrapolation result, and finally obtain the wind speed extrapolation product predicted at the interval of the second time length in the future time period.

9. An electronic device, characterized in that: It includes a memory and a processor, and the processor is used to implement the steps of the severe convective gale forecasting method based on the upstream actual wind speed as described in any one of claims 1 to 7 when executing the computer management program stored in the memory.

10. A computer-readable storage medium, characterized in that: A computer management program is stored thereon, and when the computer management program is executed by the processor, the steps of the severe convective gale forecasting method based on the upstream actual wind speed as described in any one of claims 1 to 7 are implemented.

Images