Airport sudden gale early warning method, system application, electronic equipment and medium

Abstract

The invention provides an airport sudden gale early warning method, system application, electronic equipment and a medium, and relates to the technical field of weather early warning. The method comprises the following steps: acquiring numerical ensemble forecast data, front tuyere real-time data and airport apron wind measurement station real-time data according to an airport position, and determining characteristic variables based on the numerical ensemble forecast data, the front tuyere real-time data and the airport apron wind measurement station real-time data; inputting the characteristic variable into a field-level early warning model to obtain a strong wind early warning probability output by the field-level early warning model; inputting the airport apron wind measurement station real-time data and the strong wind early warning probability into an airport apron level early warning model to obtain gridding wind field data output by the airport apron level early warning model; and based on the grid wind field data, visual airport sudden strong wind early warning is carried out, the visual airport sudden strong wind early warning comprises a strong wind thermodynamic diagram, a strong wind high-risk position and a strong wind high-risk wind zone, and early warning can be carried out on the airport sudden strong wind economically and feasibly.

Description

Technical Field

[0001] The present invention relates to the technical field of weather warning, and particularly to an airport sudden strong wind warning method, system application, electronic device and medium. Background Art

[0002] With the rapid development of the civil aviation industry, the safety and efficiency of airport operations have put forward higher requirements for the accuracy and timeliness of meteorological services. Airport sudden strong wind, as a high-risk meteorological phenomenon, poses a major threat to aircraft takeoff and landing, apron operations and ground facility safety. Traditional airport meteorological services mainly rely on global numerical prediction models such as ECMWF, GFS, GRAPES, etc. and regional meteorological observation equipment such as Doppler weather radars to provide wind field prediction and warning.

[0003] However, the horizontal resolution of global numerical prediction models is usually 9 - 25 km, making it difficult to resolve local wind field changes caused by micro-topographies such as large obstacles and building complexes within 5 km around the airport; although Doppler weather radars can monitor severe convective weather, due to the minimum elevation angle limit of usually ≥0.5°, there is a detection blind area for the wind field below 10 m near the ground, resulting in insufficient ability to capture sudden wind field disturbances. In view of the above problems, there is an urgent need in the industry for an economically feasible airport sudden strong wind warning solution that can integrate multi-source data. Summary of the Invention

[0004] In view of the problems existing in the prior art, the present invention provides an airport sudden strong wind warning method, system application, electronic device and medium.

[0005] The present invention provides an airport sudden strong wind warning method, including: Obtaining numerical ensemble forecast data, real-time data of the front air inlet and real-time data of the apron wind measurement station according to the airport location, and determining characteristic variables based on the numerical ensemble forecast data, the real-time data of the front air inlet and the real-time data of the apron wind measurement station; Inputting the characteristic variables into the field-level warning model to obtain the strong wind warning probability output by the field-level warning model; wherein, the field-level warning model is trained based on sample characteristic variables and corresponding sample strong wind warning probabilities; Inputting the real-time data of the apron wind measurement station and the strong wind warning probability into the apron-level warning model to obtain the grid wind field data output by the apron-level warning model; wherein, the apron-level warning model is trained based on sample apron wind measurement station data, sample strong wind warning probabilities and sample grid wind field data; Conducting airport sudden strong wind warning based on the grid wind field data.

[0006] An airport sudden strong wind warning method provided by the present invention, before training based on sample feature variables and corresponding sample strong wind warning probabilities, the method further includes: Obtain the historical data of the area around the apron and the historical records of sudden strong wind events; Based on the set triggering conditions, determine the strong wind warning probability of each sudden strong wind event in the historical records of sudden strong wind events to obtain the sample strong wind warning probability; determine the target historical data corresponding to each sudden strong wind event in the historical records of sudden strong wind events, and determine the sample feature variables based on the target historical data.

[0007] An airport sudden strong wind warning method provided by the present invention, before determining the strong wind warning probability of each sudden strong wind event in the historical records of sudden strong wind events based on the set triggering conditions, the method further includes: Obtain the historical data of the apron anemometer station; Determine the target historical data of the apron anemometer station within different lead time windows corresponding to each sudden strong wind event in the historical records of sudden strong wind events; Based on the target historical data of the apron anemometer station, set the triggering conditions for each strong wind warning probability.

[0008] An airport sudden strong wind warning method provided by the present invention, obtaining the grid wind field data output by the apron-level warning model, includes: When it is determined that the strong wind warning probability is within the set warning range, the apron-level warning model outputs grid wind field data with a first grid resolution based on the set dynamic resolution switching mechanism; When it is determined that the strong wind warning probability is outside the set warning range, the apron-level warning model outputs grid wind field data with a second grid resolution based on the set dynamic resolution switching mechanism; Wherein, the first grid resolution is greater than the second grid resolution.

[0009] An airport sudden strong wind warning method provided by the present invention, obtaining the grid wind field data output by the apron-level warning model, includes: When it is determined that the strong wind warning probability is within the set warning range, the apron-level warning model outputs grid wind field data based on the first iteration period; When it is determined that the strong wind warning probability is outside the set warning range, the apron-level warning model outputs grid wind field data based on the second iteration period; Wherein, the first iteration period is less than the second iteration period.

[0010] An airport sudden strong wind warning method provided by the present invention, obtaining the grid wind field data output by the apron-level warning model, includes: The apron-level early warning model outputs grid wind field data based on a time decay factor and a dynamically adjusted range; wherein, the dynamically adjusted range is dynamically adjusted according to the azimuth angle.

[0011] According to an airport sudden strong wind early warning method provided by the present invention, after training to obtain a field-level early warning model, the method further includes: Obtain the observation data of the apron wind measurement station, and determine the feedback characteristic variables based on the observation data of the apron wind measurement station; Input the feedback characteristic variables into the field-level early warning model to obtain the feedback strong wind early warning probability output by the field-level early warning model; Determine the strong wind early warning probability output by the field-level early warning model corresponding to the observation data of the apron wind measurement station, and determine the loss function between the strong wind early warning probability and the feedback strong wind early warning probability; Update the parameters of the field-level early warning model by minimizing the loss function.

[0012] The present invention also provides an airport sudden strong wind early warning device, including: A characteristic variable determination module, configured to obtain numerical set forecast data, real-time data of a front air vent, and real-time data of an apron wind measurement station according to the airport location, and determine characteristic variables based on the numerical set forecast data, the real-time data of the front air vent, and the real-time data of the apron wind measurement station; A field-level early warning level determination module, configured to input the characteristic variables into a field-level early warning model to obtain the strong wind early warning probability output by the field-level early warning model; wherein, the field-level early warning model is trained based on sample characteristic variables and corresponding sample strong wind early warning probabilities; An apron wind field data determination module, configured to input the real-time data of the apron wind measurement station and the strong wind early warning probability into an apron-level early warning model to obtain the grid wind field data output by the apron-level early warning model; wherein, the apron-level early warning model is trained based on sample apron wind measurement station data, sample strong wind early warning probabilities, and sample grid wind field data; An apron-level early warning level determination module, configured to perform visual airport sudden strong wind early warning based on the grid wind field data, and the visual airport sudden strong wind early warning includes a strong wind heat map, a strong wind high-risk position, and a strong wind high-risk wind belt.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, it implements the airport sudden strong wind early warning method as described in any one of the above.

[0014] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the airport sudden strong wind warning method described in any one of the above is implemented.

[0015] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the airport sudden strong wind warning method described in any one of the above is implemented.

[0016] The airport sudden strong wind warning method, system application, electronic device and medium provided by the present invention determine characteristic variables based on the numerical ensemble forecast data, the real-time data of the front air inlet and the real-time data of the apron anemometer station, and input the characteristic variables into the field-level warning model to obtain the output strong wind warning probability, which can provide a macroscopic meteorological trend. Input the real-time data of the apron anemometer station and the strong wind warning probability into the apron-level warning model to obtain the grid wind field data output by refined microscopic inversion, which can economically and feasibly conduct airport sudden strong wind warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 is one of the flow schematic diagrams of the airport sudden strong wind warning method provided by the present invention.

[0019] Figure 2 is one of the application schematic diagrams of the airport sudden strong wind warning method provided by the present invention.

[0020] Figure 3 is the second of the application schematic diagrams of the airport sudden strong wind warning method provided by the present invention.

[0021] Figure 4 is the third of the application schematic diagrams of the airport sudden strong wind warning method provided by the present invention.

[0022] Figure 5 is the fourth of the application schematic diagrams of the airport sudden strong wind warning method provided by the present invention.

[0023] Figure 6 is the fifth of the application schematic diagrams of the airport sudden strong wind warning method provided by the present invention.

[0024] Figure 7 is the second of the flow schematic diagrams of the airport sudden strong wind warning method provided by the present invention.

[0025] Figure 8 It is the third schematic flow chart of the airport sudden strong wind warning method provided by the present invention.

[0026] Figure 9 It is the fourth schematic flow chart of the airport sudden strong wind warning method provided by the present invention.

[0027] Figure 10 It is the schematic structural diagram of the airport sudden strong wind warning device provided by the present invention.

[0028] Figure 11 It is the schematic structural diagram of the electronic device provided by the present invention. Detailed implementation manners

[0029] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without making creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0030] The following combines Figures 1-11 to describe the airport sudden strong wind warning method, system application, electronic device and medium of the present invention.

[0031] Figure 1 It is one of the schematic flow charts of the airport sudden strong wind warning method provided by the present invention. As Figure 1 shown, the method includes the following steps: Step 101: Obtain numerical ensemble forecast data, real-time data of the front air inlet and real-time data of the apron anemometer station according to the airport location, and determine characteristic variables based on the numerical ensemble forecast data, the real-time data of the front air inlet and the real-time data of the apron anemometer station.

[0032] Numerical ensemble forecast, also known as ensemble forecast, refers to obtaining a set of numerical forecast initial values with a certain probability density function distribution characteristic within a certain initial value error range through a certain mathematical method, and then integrating each initial value with a numerical model to obtain a set of forecast results. The numerical ensemble forecast data can be the probability density distribution information of the future weather state estimated from this set of forecast results, such as data like ensemble mean, probability, dispersion, extreme value, etc.

[0033] The front air vents refer to some areas where the wind force changes are highly correlated with sudden strong winds on the apron. The real-time data of the front air vents refers to the wind speed and wind direction data collected in real time by wind force monitoring devices installed at the front air vents. The real-time data of the front air vents can be the wind speed and wind direction data collected within the previous 1 hour, or the wind speed and wind direction data collected within the previous 0.5 hour, etc. The specific time window can be set according to actual needs and is not further limited in this embodiment.

[0034] The apron wind measurement stations refer to wind speed and wind direction observation stations with various densities deployed in the apron area. They can be of a sparse deployment density, or multiple wind measurement stations obtained by increasing the deployment density compared to the sparse wind measurement stations in the traditional apron. The real-time data of the apron wind measurement stations refers to the data related to wind force changes collected in real time by the apron wind measurement stations. The setting of the specific time window for the real-time data of the apron wind measurement stations is basically the same as that for the real-time data of the front air vents and will not be elaborated here. Exemplarily, the real-time data of the apron wind measurement stations can be the data obtained by making observations once every minute.

[0035] In one embodiment, obtaining the real-time data of the apron wind measurement stations includes: determining the areas with significant wind force changes based on the building parameters in the apron, and obtaining the real-time data of the apron wind measurement stations based on the areas with significant wind force changes.

[0036] Specifically, the building parameters may include the layout, height, shape, etc. of the buildings in the apron. Exemplarily, based on the building parameters in the apron, the areas that may cause the wind force in the apron to increase (such as the bottleneck effect) or decrease can be identified to determine the areas with significant wind force changes. Apron wind measurement stations are deployed in the areas with significant wind force changes to comprehensively cover the dynamic of the apron wind field, and collect and determine the wind field characteristic data related to wind force changes such as the spatial position of the wind force value, real-time wind speed, wind direction, maximum wind speed, average wind speed, and wind speed standard deviation at each position of the airport.

[0037] The characteristic variables refer to the quantitative indicators extracted or constructed from the numerical ensemble prediction data, the real-time data of the front air vents, and the real-time data of the apron wind measurement stations that have a potential association with the high wind probability. Exemplarily, the numerical ensemble prediction data, the real-time data of the front air vents, and the real-time data of the apron wind measurement stations can be split into wind force value vectors according to the wind direction, and a series of characteristic variables such as the mean value, quantile, and dispersion degree can be derived through feature engineering. In this way, through the characteristic variables, the statistical characteristics and change trends of meteorological data can be comprehensively captured, the average level and distribution of meteorological elements can be reflected, and the dispersion degree and volatility of meteorological data can be revealed, providing a rich information basis for the field-level early warning model.

[0038] Step 102: Input the feature variables into the field-level early warning model to obtain the gale early warning probability output by the field-level early warning model. Among them, the field-level early warning model is trained based on sample feature variables and corresponding sample gale early warning probabilities.

[0039] The gale early warning probability refers to the probability of gale occurrence. Exemplarily, the gale early warning probability can be the probability of gale occurrence within a future time window. Compared with the gale early warning that directly outputs a single deterministic result based on numerical weather prediction, the field-level early warning model can quantify uncertainty and improve the prediction accuracy of local sudden gale events by outputting the gale early warning probability.

[0040] Exemplarily, the gale early warning probability output by the field-level early warning model can be the hourly gale early warning probabilities for the next 24 hours or the next 7 days and other relatively long periods in the entire airport area.

[0041] Step 103: Input the real-time data of the apron anemometer station and the gale early warning probability into the apron-level early warning model to obtain the grid wind field data output by the apron-level early warning model. Among them, the apron-level early warning model is trained based on sample apron anemometer station data, sample gale early warning probabilities, and sample grid wind field data. The grid wind field data, also known as grid meteorological data, wind field grid data, etc., refers to the structured wind field distribution data with spatial continuity and integrity.

[0042] Exemplarily, the grid wind field data output by the apron-level early warning model can be the grid wind field data for each hour or each moment in the next 3 hours.

[0043] Exemplarily, the real-time data of the apron anemometer station can be characterized to obtain grid wind field input data, and the grid wind field data for a future period can be predicted and output based on the grid wind field input data. The method of characterization can be wind field inversion using mathematical statistics methods, etc.

[0044] In this way, through in-depth analysis of the real-time data of the apron anemometer station by characterization, the discrete real-time data of the apron anemometer station and the gale early warning probability are transformed into continuous wind field information covering the entire apron area, which can determine the wind field distribution data with higher spatial resolution while retaining the characteristics of the original data, facilitating the provision of data in the corresponding data format and accuracy for subsequent processing.

[0045] Step 104: Conduct visual airport sudden gale early warning based on the grid wind field data. The visual airport sudden gale early warning includes gale heat maps, high-risk gale positions, and high-risk gale belts.

[0046] The airport sudden strong wind warning method provided by the embodiment of the present invention determines characteristic variables based on the numerical ensemble forecast data, the real-time data of the front air inlet, and the real-time data of the apron anemometer station, and inputs the characteristic variables into the field-level warning model to obtain the output strong wind warning probability, which can provide a macroscopic meteorological trend. Inputting the real-time data of the apron anemometer station and the strong wind warning probability into the apron-level warning model, the grid wind field data output by its refined microscopic inversion can be obtained, and the airport sudden strong wind warning can be carried out economically and feasibly.

[0047] In one embodiment, the airport sudden strong wind warning is directly carried out by combining the strong wind warning probability output by the district-level warning model and the grid wind field data output by the apron-level warning model.

[0048] In one embodiment, anemometers are also arranged on the top of the terminal building to monitor the high-altitude wind field. Working in coordination with the apron anemometers on the apron ground, it can comprehensively cover the airport area and monitor the three-dimensional wind field characteristic data.

[0049] In one embodiment, a visual airport sudden strong wind warning is carried out based on the grid wind field data, including: determining the regional strong wind warning level based on the grid wind field data for a visual airport sudden strong wind warning.

[0050] Exemplarily, the complex grid wind field data can be converted into an intuitive color-layered dynamic wind field moving line, a strong wind heat map, etc. through visualization technology, so as to facilitate determining the regional strong wind warning level and visually presenting the wind speed distribution and change trend. Among them, the regional strong wind warning level refers to the strong wind warning level of each area of the airport. For example, before carrying out the airport sudden strong wind warning, the airport can be divided into 20 small areas first, so as to carry out the sudden strong wind warning for each area.

[0051] Specifically, the strong wind warning level can preferably be divided by different colors, such as blue, yellow, orange, red, etc.; it can also be divided by different numbers, such as level one, level two, level three, level four, etc. Among them, when dividing by different colors, the level can be represented by the warmth or coldness of the hue. Blue can represent a low level of strong wind warning, and red warm color can represent a high level of strong wind warning; when dividing by different numbers, level one can represent a low level of strong wind warning, and level four can represent a high level of strong wind warning. In this way, it is convenient for the staff to quickly judge and take countermeasures.

[0052] The dynamic wind field moving line can be a three-dimensional wind field moving line. Such as Figure 2As shown, in one embodiment, the strong wind probability forecast map uses a three-dimensional wind field movement line, which can clearly show the direction of the wind field through the three-dimensional wind field movement line map; wherein, the color of the movement line can be synchronized with the color of the strong wind warning level corresponding to its wind speed, so that when the wind speed is in a certain range, the movement line can be presented with the color effect of the corresponding strong wind warning level, so that the staff can more easily distinguish the sizes of different wind speeds, and determine the corresponding strong wind warning level accordingly, which can effectively improve the staff's analysis and response capabilities to wind field conditions.

[0053] like Figure 3 As shown, in one embodiment, the gale probability forecast map uses a gale heat map, and uses a gradient color system from cold to warm tones to represent different gale warning levels, so that as the wind speed value increases, the color of the gale heat map smoothly transitions from cold to warm, thereby intuitively presenting the distribution of wind speed strength. Figure 4 As shown, the border color of the warning information of sudden strong wind warning for the airport can be synchronized with the color of the strong wind warning level, for example Figure 3 Medium blue represents low-level alarm information. In this way, with the help of color differentiation, it is easier for staff to know the specific status of the alarm information more clearly and make corresponding judgments and handle it quickly.

[0054] In one embodiment, if Figure 5 As shown, after determining the high-risk position for high winds according to the regional high-wind warning level, the high-risk position for high winds can be marked and displayed with a striking color that is different from the normal position. In this way, the staff can more quickly determine the high-risk position for high winds with the help of the distinct color difference; in another embodiment, the high-risk position for high winds can be further associated with the high-risk wind belt for high winds, so as to focus on the wind field conditions around the high-risk areas for high winds such as the high-risk position for high winds and the high-risk wind belt for high winds, so as to take effective countermeasures in time and respond to emergencies.

[0055] like Figure 6 As shown, the warning information may include two types. One is a remote area strong wind warning, which displays the strong wind warning for the remote area in the next 12 hours, and the wind speed value is represented by the color corresponding to the strong wind warning level; the other is a strong wind warning for the airport area, and the wind speed value is also represented by the color corresponding to the strong wind warning level.

[0056] In addition, the corresponding wind protection measures and operation procedures can be automatically added according to the wind warning level and area type to carry out sudden wind warning at the airport, realizing intelligent assistance. For example, when a red warning appears in a certain area, it can send a message to the target staff to adjust the parking position of the expected incoming aircraft to the sheltered area and to tie up the parked aircraft, and at the same time reinforce the surrounding equipment and facilities, and suspend high-altitude operations when necessary, so as to improve the efficiency and safety of apron operations.

[0057] In one embodiment, different types of airport sudden strong wind warnings can be issued for different departments. For example, the strong wind probability forecast map can be timely and accurately conveyed to key decision-making departments such as the airport operation control center and the flight area management department through various distribution channels such as the internal information platform of the airport, e-mail, and text messages for airport sudden strong wind warnings, providing a macro decision-making basis; the specific grid wind field data and the strong wind probability forecast map can be pushed in real time to each operating unit on the apron, such as the aircraft maintenance department, the ground service support department, and the fuel support department, through various distribution channels such as the apron operation management system of the airport and mobile terminal applications, for airport sudden strong wind warnings, facilitating early preparation for wind prevention, optimizing operation arrangements, and reducing the impact of sudden strong winds on airport operations.

[0058] Based on any of the above embodiments, before training based on the sample feature variables and the corresponding sample strong wind warning probabilities, the method further includes: Obtaining historical data of the area around the apron and historical records of sudden strong wind events; Determining the strong wind warning probability of each sudden strong wind event in the historical records of sudden strong wind events based on a set trigger condition to obtain the sample strong wind warning probability; determining the target historical data corresponding to each sudden strong wind event in the historical records of sudden strong wind events, and determining the sample feature variables based on the target historical data.

[0059] Specifically, the area around the apron can be an area within a set range centered on the apron. The specific set range can be set according to specific needs and is not further limited in this embodiment. Exemplarily, the area around the apron can be an area within about 100 km around the airport apron. The setting of the specific time window of the historical data of the area around the apron can also be set according to specific needs and is not further limited in this embodiment. Exemplarily, the historical data of the area around the apron can be wind field observation data related to wind force changes such as wind speed and wind direction collected in the past 10 years.

[0060] The historical records of sudden strong wind events refer to the relevant records of sudden strong wind events on the apron over the years. Exemplarily, the historical records of sudden strong wind events can be determined through documents such as the work logs of the apron operation management department.

[0061] Exemplarily, based on the time information of the sudden strong wind events in the historical records of sudden strong wind events, the corresponding historical data of the apron wind measurement station can be obtained from the historical data of the area around the apron to obtain the target historical data; and the target historical data is split into wind force value vectors according to the wind direction, and the corresponding sample feature variables are derived through feature engineering, and the sample feature variables and the sample strong wind warning probabilities are corresponding, so as to train a field-level warning model based on the sample feature variables and the sample strong wind warning probabilities.

[0062] In one embodiment, real-time data of the front air outlet is obtained, and the method further includes: performing correlation analysis on the historical data and the historical records of sudden gale events to determine the front air outlet, and obtaining real-time data of the front air outlet.

[0063] In one embodiment, the historical data and the historical records of sudden strong wind events can be correlated and analyzed by mathematical statistics methods, and the wind field characteristics and areas that meet the requirements of the correlation with the sudden strong wind events on the apron can be identified to determine the front air outlets; then wind monitoring equipment is deployed at the front air outlets to collect real-time data of the front air outlets such as wind speed and wind direction data. Among them, the mathematical statistics method can be correlation analysis, cluster analysis, etc. Exemplarily, by analyzing the historical data of the area around the apron for the past 10 years and 50 sudden strong wind events, it is found that special points such as mountain passes in a certain mountainous area are prone to form a "narrow channel effect", resulting in frequent surges in wind speed before the event, and the mountain passes in the mountainous area are identified as the front air outlets.

[0064] In one embodiment, correlation analysis is performed on the historical data and the historical records of sudden gale events to determine a sudden gale warning threshold; The real-time wind speed is determined based on the real-time data of the front air outlet, and a sudden gale warning is issued for the airport when the real-time wind speed exceeds the sudden gale warning threshold.

[0065] The sudden gale warning threshold corresponds to the warning window, and the specific sudden gale warning threshold and warning window duration can be set according to actual needs, and this embodiment does not further limit this. In one embodiment, the sudden gale warning threshold is set to 15m / s, and when the real-time wind speed exceeds the threshold, the airport sudden gale warning is issued 2 hours in advance.

[0066] In this embodiment, by combining the real-time data of the front air outlet with the set sudden gale warning threshold, the sudden gale warning threshold corresponds to the warning window, which can provide a time window before the apron gale occurs, buying time for the airport's wind protection measures. In addition, it can be used in parallel as a supplement to the dual-level coupled warning through the field-level warning model and the apron-level warning model, improving the reliability of the sudden gale warning for the airport.

[0067] Based on any of the above embodiments, before determining the gale warning probability of each sudden gale event in the sudden gale event historical record based on the set trigger condition, the method further includes: Obtain historical data of the apron wind measuring station; Determine the historical data of the target apron wind measuring station within different advance time windows corresponding to each sudden high wind event in the sudden high wind event historical records; The triggering condition of each high wind warning probability is set based on the historical data of the target apron wind measuring station.

[0068] Specifically, the historical data of the apron anemometer station refers to the historical data related to the wind force change collected by the apron anemometer station. The specific time window of the historical data of the apron anemometer station can be set according to actual needs, and no further limitation is made in this embodiment. Exemplarily, the historical data of the apron anemometer station can be the data related to the wind force change collected within 3 years.

[0069] Exemplarily, the time information of the sudden strong wind event can be determined. According to this time information, the historical data of the apron anemometer station with different advance time windows is obtained as the target historical data of the apron anemometer station. The wind field characteristics such as the maximum wind speed, average wind speed, and wind speed standard deviation are analyzed based on the target historical data of the apron anemometer station, and the triggering conditions for different strong wind warning probabilities are set.

[0070] Among them, the triggering condition can be a triggering threshold or a triggering range, etc. The advance time window can be 1 hour or 2 hours, etc.

[0071] In one embodiment, the building layout of the apron is analyzed, and it is determined that the channel between the terminal building and the apron is prone to form a wind force concentration area. Five anemometer stations are arranged in this area, 3-year data is collected, and the optimal threshold for the orange warning is calculated to be a wind speed of 28.5 m / s (1 hour in advance, hit rate 90%, false alarm rate 10%).

[0072] In this embodiment, the advance time window analysis can capture the wind field characteristics before the strong wind occurs. Through the associated processing of different advance time windows, the most predictive wind field characteristics can be screened out, the hit rate of the strong wind warning probability can be improved, and the false alarm rate of the strong wind warning probability can be reduced.

[0073] Based on any of the above embodiments, the grid wind field data output by the apron-level warning model is obtained, including: When it is determined that the strong wind warning probability is within the set warning range, the apron-level warning model outputs grid wind field data with a first grid resolution based on the set dynamic resolution switching mechanism; When it is determined that the strong wind warning probability is outside the set warning range, the apron-level warning model outputs grid wind field data with a second grid resolution based on the set dynamic resolution switching mechanism; Among them, the first grid resolution is greater than the second grid resolution.

[0074] Specifically, the set warning range refers to the warning probability range for which refined grid wind field data needs to be provided. The specific warning probability range can be set according to actual needs by self-statistical historical data, and no further limitation is made in this embodiment. Exemplarily, the warning probability range can be more than 40%.

[0075] Exemplarily, when the probability of strong wind warning does not exceed 40%, the apron-level warning model outputs gridded wind field data with a grid resolution of 100 meters. When the probability of strong wind warning exceeds 40%, the apron-level warning model can automatically switch the grid resolution to 50 meters and output gridded wind field data with a grid resolution of 50 meters.

[0076] In this embodiment, by automatically increasing the grid resolution when the probability of strong wind warning is relatively high, the spatial details of wind field prediction can be increased, so as to more accurately capture the influence of local terrain such as buildings on the apron on the wind field, and improve the reliability of airport sudden strong wind warning based on gridded wind field data.

[0077] Further, when the apron-level warning model outputs gridded wind field data with a first grid resolution, it can output gridded wind field data with the same grid resolution for the entire apron, or output gridded wind field data with a first refined grid resolution for the ordinary area of the apron and gridded wind field data with a second refined grid resolution for the key area of the apron. Among them, the first refined grid resolution is greater than the second grid resolution, and the second refined grid resolution is greater than the first refined grid resolution. Exemplarily, the key areas may include runway ends, taxiway intersections, high-risk aprons, etc., and the grid resolution at positions such as runway ends, taxiway intersections, and high-risk aprons can be further refined to 25m.

[0078] Key areas such as runway ends and taxiway intersections are the core areas for aircraft takeoff, landing, and taxiing. In this embodiment, by further refining the grid resolution of the key areas, the guarantee of flight safety can be further increased through fine wind field prediction.

[0079] After inputting the real-time data of apron wind measurement stations and the probability of strong wind warning into the apron-level warning model, the grid point density can be increased through spatial interpolation, the area covered by a single grid can be reduced, and gridded wind field data can be obtained, thereby improving the sensitivity to wind field changes.

[0080] In order to improve the accuracy and stability of wind field prediction in the encrypted area, in one embodiment, obtaining the gridded wind field data output by the apron-level warning model includes: the apron-level warning model outputs gridded wind field data based on a time decay factor. In this way, when performing spatial interpolation processing, the time decay factor can ensure that the latest observation data obtains a higher weight, facilitating the synchronous update of the gridded wind field data output by the apron-level warning model with the real-time risk, and reducing the risk of warning delay or misjudgment caused by the lag of the output gridded wind field data.

[0081] Specifically, the time decay factor can adopt the form of an exponential function, such that the weight decays exponentially with the increase of the time difference, so as to reduce the weight of historical data and ensure that the weight of the latest data is higher. The formula of this algorithm is: Wherein, is the estimated wind speed value at the target position ,, is the th measured wind speed value at the th observation point, is the spatial weight coefficient of the i-th observation point, is the time decay factor, is the decay rate constant, is the target time,

[0082] Wherein, The value of can be determined by itself according to the decay intensity of the weight with the time difference. Exemplarily, ; Satisfies ; And The unit of can be minutes.

[0083] In another embodiment, the grid wind field data output by the apron-level early warning model is obtained, including: the apron-level early warning model outputs grid wind field data based on dynamically adjusted range; wherein, the dynamically adjusted range is dynamically adjusted according to the azimuth angle. In this way, when performing spatial interpolation processing, the dynamically adjusted range can better adapt to the spatial correlation under different wind direction conditions, improving the accuracy and reliability of spatial interpolation.

[0084] Exemplarily, the spatial covariance function can adopt an exponential model, and the range a in the exponential model is dynamically adjusted according to the azimuth angle. When the east wind is dominant, a = 800m in the east-west direction and a = 300m in the north-south direction, which can reflect the wind direction anisotropy.

[0085] Wherein, is the wind speed correlation between two points with a spatial distance of h meters, is the variance of the wind speed data, is the range.

[0086] The adjustment rules for other wind directions are basically the same as those when the east wind is dominant, and will not be elaborated in this embodiment.

[0087] In this embodiment, by improving the spatial interpolation process in coordination with the time decay factor and the dynamic adjustment range, it is possible to better process spatio-temporal data, especially suitable for the spatial interpolation process of data with time correlation and wind direction anisotropy.

[0088] Based on any of the above embodiments, the gridded wind field data output by the apron-level warning model is obtained, including: When it is determined that the high wind warning probability is within the set warning range, the apron-level warning model outputs the gridded wind field data based on the first iteration period; When it is determined that the high wind warning probability is outside the set warning range, the apron-level warning model outputs the gridded wind field data based on the second iteration period; Wherein, the first iteration period is less than the second iteration period.

[0089] Specifically, the set warning range in this embodiment is basically the same as that in the foregoing embodiments, and will not be elaborated here.

[0090] Exemplarily, when the high wind warning probability does not exceed 40%, the apron-level warning model can output the gridded wind field data every 10 minutes, and when the high wind warning probability exceeds 40%, the apron-level warning model can output the gridded wind field data every 1 minute.

[0091] In this embodiment, by automatically shortening the iteration period when the high wind warning probability is relatively high, the gridded wind field data can be updated more frequently, facilitating the timely capture of the rapid changes in the wind field and providing more powerful support for the airport to respond to sudden high winds.

[0092] In one embodiment, it may be that when the probability of high wind occurrence at any moment within a set future time window is within the set warning range, the apron-level warning model outputs the gridded wind field data with the first grid resolution based on the set dynamic resolution switching mechanism, and / or the apron-level warning model outputs the gridded wind field data based on the first iteration period to ensure the provision of high-precision and high-timeliness airport sudden high wind warnings at critical moments.

[0093] The airport sudden high wind warning method provided by the embodiments of the present invention, based on the dynamic resolution switching mechanism and the automatic adjustment of the iteration period, can improve the reliability of the airport sudden high wind warning while, during non-high-risk periods, keeping the apron-level warning model at a lower grid resolution and a higher iteration period, reducing unnecessary computational loads, optimizing the allocation of computing resources, and improving the economic feasibility of warning against airport sudden high winds.

[0094] Based on any of the above embodiments, after training the field-level warning model, the method further includes: Obtain the observation data of the apron anemometer station, and determine the feedback characteristic variables based on the observation data of the apron anemometer station; Input the feedback characteristic variables into the field-level early warning model to obtain the feedback gale early warning probability output by the field-level early warning model; Determine the gale early warning probability output by the field-level early warning model corresponding to the observation data of the apron anemometer station, and determine the loss function between the gale early warning probability and the feedback gale early warning probability; Update the parameters of the field-level early warning model by minimizing the loss function.

[0095] Specifically, the observation data of the apron anemometer station refers to the measured data of the apron anemometer station collected every day, which is used to construct a closed-loop learning system for the field-level early warning model and perform online parameter update for the field-level early warning model.

[0096] Exemplarily, an incremental learning pipeline can be constructed to transmit the observation data of the apron anemometer station back. After processing the transmitted data such as cleaning, denoising, and standardization, determine the feedback characteristic variables and perform iterative optimization training on the field-level model.

[0097] The working principle and technical effect of determining the feedback characteristic variables through the observation data of the apron anemometer station are basically the same as those of determining the sample characteristic variables based on the target historical data, and will not be elaborated here.

[0098] Exemplarily, a momentum SGD (Stochastic Gradient Descent) optimizer can be used to update the parameters of the field-level early warning model with a learning rate η = 0.01 and a batch size B = 256. In this way, by introducing a momentum term, the convergence speed of the field-level early warning model can be accelerated, avoiding the field-level early warning model falling into a local optimal solution, and improving the generalization ability of the field-level early warning model.

[0099] SHAP (SHapley Additive exPlanations) values can quantify the contribution of each feature to the model prediction. Specifically, through SHAP value analysis, the key features for predicting wind field changes can be identified, so as to dynamically adjust the weights of type parameters such as mean wind, gust wind, surrounding wind, and thunderstorm weather of the input features. For example, before the arrival of a thunderstorm, the thunderstorm proximity index plays an important role in predicting wind field changes, and its weight is dynamically adjusted and increased by 15% - 20%. This dynamic adjustment mechanism enables the model to better adapt to the wind field characteristics under different meteorological conditions and improve the prediction accuracy.

[0100] In this embodiment, the parameters of the field-level early warning model are updated by obtaining the observation data of the apron anemometer station, and the operation of the apron-level early warning model is guided by the updated field-level early warning model, realizing a closed-loop feedback optimization mechanism, which can improve the reliability of airport sudden gale early warning.

[0101] In one embodiment, after obtaining the observation data of the apron anemometer station, the observation data of the apron anemometer station can be divided into a training data set and a test data set. After updating the parameters of the field-level early warning model with the training data set, the prediction performance of the model is evaluated based on the test data set through indicators such as accuracy rate and recall rate. According to the evaluation results, the optimization strategy is further adjusted.

[0102] To specifically illustrate the functions of the airport sudden gale early warning method provided in this embodiment, the following combines Figure 7 to provide a specific example.

[0103] As Figure 8 shown, perform correlation analysis on historical data and historical records of sudden gale events to determine the front tuyere, and obtain real-time data of the front tuyere; Obtain numerical ensemble forecast data, real-time data of the front tuyere, and real-time data of the apron anemometer station according to the airport location, and determine feature variables from the numerical ensemble forecast data, real-time data of the front tuyere, and real-time data of the apron anemometer station through feature engineering; Obtain historical data of the area around the apron and historical records of sudden gale events; obtain historical data of the apron anemometer station; determine the historical data of the target apron anemometer station within different lead time windows corresponding to each sudden gale event in the historical records of sudden gale events; set the triggering conditions for each gale warning probability based on the historical data of the target apron anemometer station; determine the gale warning probability of each sudden gale event in the historical records of sudden gale events based on the set triggering conditions to obtain the sample gale warning probability; determine the target historical data corresponding to each sudden gale event in the historical records of sudden gale events, and determine the sample feature variables based on the target historical data; Specifically, the ensemble mean (EM) and median of specific parameters such as the wind speed and precipitation of all members can be calculated to capture the dominant trend; the 10%, 25%, 75%, and 90% quantiles can be extracted to quantify the possibility of extreme events. For example, when the 90% quantile wind speed > 20 m / s, it indicates a high-risk area; the standard deviation (STD), range (Max - Min), etc. can be calculated to reflect the forecast uncertainty; the model bias is corrected for high-dispersion areas by combining "stochastic physical perturbations" (such as SPPT); sliding statistics such as 24-hour sliding variance and 72-hour trend slope are performed on the time series of ensemble members to capture the evolution law of weather systems. For example, when the 3-hour change rate of the pressure gradient > 2 hPa / km², it indicates the feature of frontal acceleration.

[0104] Input the feature variables into the field-level early warning model to obtain the gale warning probability output by the field-level early warning model; refer to Figure 7, the field-level early warning model is trained based on sample feature variables and corresponding sample strong wind early warning probabilities; and obtain the observation data of the apron anemometer station, conduct data comparison and analysis to facilitate incremental training or adjust the prediction time range. Specifically, the feedback feature variables can be determined based on the observation data of the apron anemometer station; input the feedback feature variables into the field-level early warning model to obtain the feedback strong wind early warning probability output by the field-level early warning model; determine the strong wind early warning probability output by the field-level early warning model corresponding to the observation data of the apron anemometer station, and determine the loss function between the strong wind early warning probability and the feedback strong wind early warning probability; update the parameters of the field-level early warning model by minimizing the loss function; evaluate the updated field-level early warning model based on the test data set and perform optimization feedback; see Figure 7 , the district-level early warning model can be an XGBoost model, and the output can be the strong wind early warning probability representing the entire apron area; Such as Figure 9 shown, fuse the real-time data of the apron anemometer station and the strong wind early warning probability, and then input it into the apron-level early warning model to perform feature processing to obtain the grid wind field input data, see Figure 7 , when it is determined that the strong wind early warning probability is within the set early warning range, enter the apron-level high-precision mode, and the apron-level early warning model outputs grid wind field data with a grid resolution of 50 meters based on the first iteration period and the set dynamic resolution switching mechanism; when it is determined that the strong wind early warning probability is outside the set early warning range, enter the apron-level normal mode, and the apron-level early warning model outputs grid wind field data with a grid resolution of 100 meters based on the second iteration period and the set dynamic resolution switching mechanism; the apron-level early warning model is trained based on sample apron anemometer station data, sample strong wind early warning probabilities and sample grid wind field data or obtained through machine learning hierarchical early warning; the apron-level early warning model outputs grid wind field data based on the time decay factor and the dynamic adjustment range; it can also be verified and feedback optimized based on a preset test set; within the set early warning range, the strong wind early warning probability can exceed 40%; Conduct airport sudden strong wind early warning based on the early warning visualization of the grid wind field data.

[0105] The airport sudden strong wind early warning device provided by the present invention will be described below. The airport sudden strong wind early warning device described below can be mutually corresponded and referred to with the airport sudden strong wind early warning method described above.

[0106] Figure 10 is a schematic structural diagram of the airport sudden strong wind early warning device provided by the present invention. Such as Figure 10 shown, the device includes: The feature variable determination module 1010 is configured to obtain numerical ensemble forecast data, real-time data of the front air outlet, and real-time data of the apron anemometer station according to the airport location, and determine feature variables based on the numerical ensemble forecast data, the real-time data of the front air outlet, and the real-time data of the apron anemometer station; The field-level warning level determination module 1020 is configured to input the feature variables into the field-level warning model to obtain the high wind warning probability output by the field-level warning model; wherein, the field-level warning model is trained based on sample feature variables and corresponding sample high wind warning probabilities; The apron wind field data determination module 1030 is configured to input the real-time data of the apron anemometer station and the high wind warning probability into the apron-level warning model to obtain the grid-based wind field data output by the apron-level warning model; wherein, the apron-level warning model is trained based on sample apron anemometer station data, sample high wind warning probabilities, and sample grid-based wind field data; The apron-level warning level determination module 1040 is configured to perform visual airport sudden high wind warning based on the grid-based wind field data, and the visual airport sudden high wind warning includes high wind heat maps, high risk positions of high winds, and high risk wind belts.

[0107] Based on any of the above embodiments, the airport sudden high wind warning device further includes a first training data acquisition module, which is configured to: Obtain historical data of the apron surrounding area and historical records of sudden high wind events; Determine the high wind warning probability of each sudden high wind event in the historical records of sudden high wind events based on set trigger conditions to obtain the sample high wind warning probability; determine the target historical data corresponding to each sudden high wind event in the historical records of sudden high wind events, and determine the sample feature variables based on the target historical data.

[0108] Based on any of the above embodiments, the airport sudden high wind warning device further includes a trigger condition determination module, which is configured to: Obtain historical data of the apron anemometer station; Determine the target historical data of the apron anemometer station within different lead time windows corresponding to each sudden high wind event in the historical records of sudden high wind events; Set the trigger conditions for each high wind warning probability based on the target historical data of the apron anemometer station.

[0109] Based on any of the above embodiments, the wind field data determination module 1030 is specifically configured to: Determine that when the high wind warning probability is within the set warning range, the apron-level warning model outputs grid-based wind field data with a first grid resolution based on a set dynamic resolution switching mechanism; When it is determined that the high wind warning probability is outside the set warning range, the apron-level warning model outputs grid-based wind field data with a second grid resolution based on the set dynamic resolution switching mechanism; wherein, the first grid resolution is greater than the second grid resolution.

[0110] Based on any of the above embodiments, the apron wind field data determination module 1030 is specifically configured to: When it is determined that the high wind warning probability is within the set warning range, the apron-level warning model outputs grid-based wind field data based on the first iteration period; When it is determined that the high wind warning probability is outside the set warning range, the apron-level warning model outputs grid-based wind field data based on the second iteration period; wherein, the first iteration period is less than the second iteration period.

[0111] Based on any of the above embodiments, the apron wind field data determination module 1030 is specifically configured to: The apron-level warning model outputs grid-based wind field data based on a time decay factor and a dynamically adjusted range; wherein, the dynamically adjusted range is dynamically adjusted according to the azimuth angle.

[0112] Based on any of the above embodiments, the airport sudden high wind warning device further includes a feedback update module, which is used for: Obtain the observation data of the apron wind measurement station, and determine the feedback characteristic variables based on the observation data of the apron wind measurement station; Input the feedback characteristic variables into the field-level warning model to obtain the feedback high wind warning probability output by the field-level warning model; Determine the high wind warning probability output by the field-level warning model corresponding to the observation data of the apron wind measurement station, and determine the loss function between the high wind warning probability and the feedback high wind warning probability; Update the parameters of the field-level warning model by minimizing the loss function.

[0113] Figure 11 Illustrates a schematic structural diagram of an electronic device, such as Figure 11As shown in the figure, the electronic device may include: a processor 1110, a communications interface 1120, a memory 1130, and a communication bus 1140. Among them, the processor 1110, the communications interface 1120, and the memory 1130 complete communication with each other through the communication bus 1140. The processor 1110 may call the logical instructions in the memory 1130 to execute the airport sudden strong wind warning method, and the method includes: obtaining numerical ensemble forecast data, real-time data of the front tuyere, and real-time data of the apron anemometer station according to the airport location, and determining characteristic variables based on the numerical ensemble forecast data, the real-time data of the front tuyere, and the real-time data of the apron anemometer station; inputting the characteristic variables into the field-level warning model to obtain the strong wind warning probability output by the field-level warning model; wherein, the field-level warning model is trained based on sample characteristic variables and corresponding sample strong wind warning probabilities; inputting the real-time data of the apron anemometer station and the strong wind warning probability into the apron-level warning model to obtain the grid wind field data output by the apron-level warning model; wherein, the apron-level warning model is trained based on sample apron anemometer station data, sample strong wind warning probabilities, and sample grid wind field data; performing visual airport sudden strong wind warning based on the grid wind field data, and the visual airport sudden strong wind warning includes a strong wind heat map, a high-risk position of the strong wind, and a high-risk wind belt of the strong wind.

[0114] In addition, when the logical instructions in the above-mentioned memory 1130 are implemented in the form of software function units and sold or used as independent products, they may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, may be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0115] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the airport sudden strong wind warning method provided by the above-mentioned various methods. The method includes: obtaining numerical ensemble forecast data, real-time data of the front air inlet, and real-time data of the apron anemometer station according to the airport location, and determining characteristic variables based on the numerical ensemble forecast data, the real-time data of the front air inlet, and the real-time data of the apron anemometer station; inputting the characteristic variables into the field-level warning model to obtain the strong wind warning probability output by the field-level warning model; wherein, the field-level warning model is trained based on sample characteristic variables and corresponding sample strong wind warning probabilities; inputting the real-time data of the apron anemometer station and the strong wind warning probability into the apron-level warning model to obtain the grid wind field data output by the apron-level warning model; wherein, the apron-level warning model is trained based on sample apron anemometer station data, sample strong wind warning probabilities, and sample grid wind field data; performing visual airport sudden strong wind warning based on the grid wind field data, and the visual airport sudden strong wind warning includes a strong wind heat map, a high-risk position of strong wind, and a high-risk wind belt of strong wind.

[0116] On another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the airport sudden strong wind warning method provided by the above-mentioned various methods. The method includes: obtaining numerical ensemble forecast data, real-time data of the front air inlet, and real-time data of the apron anemometer station according to the airport location, and determining characteristic variables based on the numerical ensemble forecast data, the real-time data of the front air inlet, and the real-time data of the apron anemometer station; inputting the characteristic variables into the field-level warning model to obtain the strong wind warning probability output by the field-level warning model; wherein, the field-level warning model is trained based on sample characteristic variables and corresponding sample strong wind warning probabilities; inputting the real-time data of the apron anemometer station and the strong wind warning probability into the apron-level warning model to obtain the grid wind field data output by the apron-level warning model; wherein, the apron-level warning model is trained based on sample apron anemometer station data, sample strong wind warning probabilities, and sample grid wind field data; performing visual airport sudden strong wind warning based on the grid wind field data, and the visual airport sudden strong wind warning includes a strong wind heat map, a high-risk position of strong wind, and a high-risk wind belt of strong wind.

[0117] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0118] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An airport sudden strong wind warning method, characterized in that, Including: Obtain numerical ensemble forecast data, real-time data of the front air inlet, and real-time data of the apron anemometer station according to the airport location, and determine characteristic variables based on the numerical ensemble forecast data, the real-time data of the front air inlet, and the real-time data of the apron anemometer station; Input the characteristic variables into the field-level early warning model to obtain the high wind early warning probability output by the field-level early warning model; wherein, the field-level early warning model is trained based on sample characteristic variables and corresponding sample high wind early warning probabilities; Input the real-time data of the apron anemometer station and the high wind early warning probability into the apron-level early warning model to obtain the grid wind field data output by the apron-level early warning model; wherein, the apron-level early warning model is trained based on sample apron anemometer station data, sample high wind early warning probabilities, and sample grid wind field data; Conduct visual airport sudden high wind early warning based on the grid wind field data, and the visual airport sudden high wind early warning includes a high wind heat map, high risk positions of high winds, and high risk wind belts of high winds.

2. The airport sudden strong wind warning method according to claim 1, characterized in that, Before training based on sample characteristic variables and corresponding sample high wind early warning probabilities, the method further includes: Obtain historical data of the apron surrounding area and historical records of sudden high wind events; Determine the high wind early warning probability of each sudden high wind event in the historical records of sudden high wind events based on set triggering conditions to obtain the sample high wind early warning probabilities; determine the target historical data corresponding to each sudden high wind event in the historical records of sudden high wind events, and determine the sample characteristic variables based on the target historical data.

3. The airport sudden strong wind warning method according to claim 2, wherein Before determining the high wind early warning probability of each sudden high wind event in the historical records of sudden high wind events based on set triggering conditions, the method further includes: Obtain historical data of the apron anemometer station; Determine the target historical data of the apron anemometer station within different lead time windows corresponding to each sudden high wind event in the historical records of sudden high wind events; Set the triggering conditions for each high wind early warning probability based on the target historical data of the apron anemometer station.

4. The airport sudden strong wind warning method according to claim 1, characterized in that Obtaining the grid wind field data output by the apron-level early warning model includes: When it is determined that the high wind early warning probability is within the set early warning range, the apron-level early warning model outputs grid wind field data with a first grid resolution based on a set dynamic resolution switching mechanism; When it is determined that the high wind early warning probability is outside the set early warning range, the apron-level early warning model outputs grid wind field data with a second grid resolution based on a set dynamic resolution switching mechanism; Wherein, the first grid resolution is greater than the second grid resolution.

5. The airport sudden strong wind warning method according to claim 1 or 4, characterized in that, Obtaining the grid wind field data output by the apron-level early warning model includes: When it is determined that the high wind early warning probability is within the set early warning range, the apron-level early warning model outputs grid wind field data based on a first iteration period; When it is determined that the high wind early warning probability is outside the set early warning range, the apron-level early warning model outputs grid wind field data based on a second iteration period; Wherein, the first iteration period is less than the second iteration period.

6. The airport sudden strong wind warning method according to claim 5, characterized in that, Obtaining the grid wind field data output by the apron-level early warning model includes: The apron-level warning model outputs grid wind field data based on a time decay factor and a dynamically adjusted range; wherein, the dynamically adjusted range is dynamically adjusted according to the azimuth angle.

7. The airport sudden strong wind warning method according to claim 2 or 3, characterized in that, After training the field-level warning model, the method further includes: Obtaining the observation data of the apron wind measurement station and determining the feedback characteristic variables based on the observation data of the apron wind measurement station; Inputting the feedback characteristic variables into the field-level warning model to obtain the feedback gale warning probability output by the field-level warning model; Determining the gale warning probability output by the field-level warning model corresponding to the observation data of the apron wind measurement station, and determining the loss function between the gale warning probability and the feedback gale warning probability; Updating the parameters of the field-level warning model by minimizing the loss function.

8. An application of an airport sudden strong wind warning system, characterized in that, Including: A characteristic variable determination module, configured to obtain numerical ensemble forecast data, real-time data of the front air inlet, and real-time data of the apron wind measurement station according to the airport location, and determine characteristic variables based on the numerical ensemble forecast data, the real-time data of the front air inlet, and the real-time data of the apron wind measurement station; A field-level warning level determination module, configured to input the characteristic variables into the field-level warning model to obtain the gale warning probability output by the field-level warning model; wherein, the field-level warning model is trained based on sample characteristic variables and corresponding sample gale warning probabilities; An apron wind field data determination module, configured to input the real-time data of the apron wind measurement station and the gale warning probability into the apron-level warning model to obtain the grid wind field data output by the apron-level warning model; wherein, the apron-level warning model is trained based on sample apron wind measurement station data, sample gale warning probabilities, and sample grid wind field data; An apron-level warning level determination module, configured to perform a visual airport sudden gale warning based on the grid wind field data, and the visual airport sudden gale warning includes a gale heat map, a high-risk gale position, and a high-risk gale belt.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the airport sudden gale warning method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the airport sudden gale warning method according to any one of claims 1 to 7.

Images