Airport available runway monitoring and early warning system and method based on multi-source data fusion

Abstract

The airport available runway monitoring and early warning system and method based on multi-source data fusion are high in real-time performance, good in forecasting effect and large in monitoring range. The system comprises a laser wind finding radar (1), a CFD numerical simulation module (2), a mesoscale weather forecast module (3), a machine learning model (4) and a system server (5). The method comprises the following steps that: a system server performs data assimilation by combining real-time wind shear detection data measured by a laser wind measurement radar and forecast data of a current time point output by a mesoscale weather forecast module, processes the assimilated data by applying a CFD numerical simulation module, and calculates the real-time wind shear detection data by combining terrain and building parameters near an airport; according to the method, real-time sub-meter gridding three-dimensional wind field data is generated, and a sub-meter three-dimensional wind field structure is reconstructed, so that future wind shear of an airport is forecast and early warned, and then available runway prediction is carried out. The method is applied to the technical field of wind field prediction and judgment.

Description

Technical Field

[0001] The present invention relates to the technical field of wind field prediction and judgment, and in particular to an airport available runway monitoring and early warning system and method based on multi-source data fusion formed by combining four technologies. Background Art

[0002] Wind shear refers to the change in the wind vector in the horizontal and vertical directions. Wind shear over an airport can suddenly change the flight attitude of an aircraft, causing turbulence, go-arounds, crashes, etc. Traditional airport wind shear monitoring methods mainly use various ground sensors to monitor wind shear, which makes it impossible to warn of high-altitude wind shear. As a new type of wind field detection equipment using laser as a detection method, laser wind radar has high detection accuracy and spatial resolution. It can scan wind fields at different altitudes above the airport and accurately monitor high-altitude wind shear. However, due to the limitations of its detection distance and scanning method, it can only obtain limited wind field information within its scanning area and cannot form large-scale three-dimensional grid data. It has insufficient monitoring capabilities for large-scale wind shear systems far away from the airport and does not have the function of forecasting. Therefore, in recent years, a method combining CFD numerical simulation and mesoscale meteorological forecast model (WRF) has gradually been applied to airport wind shear monitoring. However, due to the large amount of CFD calculation, it can only establish a three-dimensional wind field sample database in advance, and then retrieve the sample with the highest matching degree according to the current situation. As the current three-dimensional wind field, it is not a real-time calculation, which also limits its ability to carry out subsequent development and utilization of three-dimensional wind field data. Summary of the invention

[0003] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide an airport available runway monitoring and early warning system and method based on multi-source data fusion with strong real-time performance, good forecasting effect and large monitoring range.

[0004] The technical solution adopted by the airport available runway monitoring and early warning system based on multi-source data fusion of the present invention is a system for monitoring and early warning of airport available runways based on multi-source data fusion, comprising: A laser wind radar used to detect wind fields within a set range and at a set height, providing the system with high-precision, real-time wind shear detection data; CFD numerical simulation module used to generate high-precision gridded wind field data through simulation calculation; A mesoscale meteorological forecast module for simulating future weather changes over a large area; Machine learning models for improving CFD calculation speed and predicting available runways at airports; and A system server, wherein the system server combines the real-time wind shear detection data measured by the laser wind measuring radar and the forecast data output by the mesoscale meteorological forecast module, assimilates the real-time detection data with the forecast data at the current time point output by the mesoscale meteorological forecast module, and then uses the CFD numerical simulation module to process the assimilated data, combines the terrain and building parameters near the airport, generates real-time sub-meter gridded three-dimensional wind field data, and reconstructs the sub-meter three-dimensional wind field structure, thereby forecasting and warning the future wind shear of the airport, and then predicting the available runway.

[0005] The method for predicting the available runways of an airport using the above-mentioned airport available runway monitoring and early warning system based on multi-source data fusion comprises the following steps: a. Deployment and data acquisition of laser wind radar: Deploy the laser wind radar at key locations in the airport and nearby areas, perform system calibration and performance testing on the laser wind radar to ensure its stable operation and accurate capture of atmospheric information, thereby providing real-time, high-resolution real-time wind field scanning data to provide basic data support for subsequent operations; b. Calculation of future wind field data: Use the mesoscale weather forecast module to simulate the future weather change trend of the airport area to obtain forecast data for a period of time in the future, then use the inversion algorithm and the CFD numerical simulation module to process the forecast data, combine the terrain and building parameters near the airport, simulate the atmospheric state, and generate sub-meter gridded three-dimensional wind field data of the airport at future times; c. Real-time wind field data calculation: After receiving and parsing the real-time wind field scanning data sent by the laser wind measuring radar and the forecast data output by the mesoscale meteorological forecast module, the system server performs data assimilation on the real-time wind field scanning data and the forecast data at the current time point output by the mesoscale meteorological forecast module, and then uses the CFD numerical simulation module to process the assimilated data, and combines the terrain and building parameters near the airport to generate real-time sub-meter gridded three-dimensional wind field data; d. In the process of calculating real-time wind field data, only the CFD numerical simulation module is used to calculate the sub-meter gridded three-dimensional wind field data at key time points, and then the calculation results at key time points are combined with the data obtained by real-time detection of the laser wind radar to drive the graph neural network calculation model, thereby quickly obtaining real-time sub-meter gridded three-dimensional wind field data; e. Wind shear detection: Extract the wind field data of the aircraft sliding down to the ground from the three-dimensional wind field data at the current and selected future different times, and calculate the wind shear data on the glide path, thereby realizing wind shear warning and forecast; f. Runway availability judgment: Combine the current and selected future three-dimensional wind field data at different times, as well as external data collected by the airport, and use the random forest model in machine learning to judge the current and future runway availability; g. Early warning system construction: Deploy an early warning module to automatically trigger early warning notifications when potential dangerous wind field conditions are detected, ensuring that each user terminal can take timely response measures.

[0006] In the step b, the specific steps of using the inversion algorithm and the CFD numerical simulation module to process the forecast data are as follows: standardizing the forecast data output by the mesoscale meteorological forecast module, including time alignment, spatial interpolation and format conversion, and unifying it to a grid coordinate system compatible with the CFD numerical simulation module; optimizing the low-resolution forecast data using a variational inversion algorithm, and fusing the forecast data with the statistical characteristics of the historical lidar observation data by minimizing the objective function to generate a high-resolution initial wind field input; Combined with the terrain and building parameters near the airport, the atmospheric state is simulated and calculated, and the specific steps of generating sub-meter gridded three-dimensional wind field data of the airport at a future moment are as follows: the inverted high-resolution wind field data is imported into the CFD numerical simulation module as the initial condition, and the airport three-dimensional terrain model and building geometric parameters are loaded at the same time, and the boundary conditions are set as the dynamic pressure and velocity coupling field; based on the Reynolds-averaged Navier-Stokes equations and the k-ε turbulence model, combined with the adaptive grid refinement technology, the airport area wind field is transiently simulated and calculated, and the three-dimensional wind speed components u, v, w and turbulent kinetic energy distribution of the sub-meter grid nodes are solved; the sub-meter gridded three-dimensional wind field data finally generated is stored in the spatiotemporal database of the system server for subsequent wind shear warning and runway prediction modules to call.

[0007] In step c, the specific steps of assimilating the real-time wind field scanning data with the forecast data are as follows: using the Newton relaxation data assimilation method to fuse the real-time wind field scanning data into the WRF model of the mesoscale meteorological forecast module, and the assimilation principle momentum equation is expressed as: , , where u i is the component of vector u in direction i, t is time, u j is the component of vector u in the j direction, v is the dynamic viscosity coefficient, x i is the coordinate component in the i direction, x j is the coordinate component in the j direction, ρ is the density, p is the pressure, f is the additional forcing term added according to the Nudging data assimilation method, and u cis the detection speed, u is the WRF prediction speed, τ u is the Newton relaxation coefficient, W(x,y,z) is the weight function; The specific steps of using the CFD numerical simulation module to process the real-time wind field scanning data, combining the terrain and building parameters near the airport, and then generating sub-meter gridded three-dimensional wind field data are as follows: integrating the real-time wind field scanning data obtained by the laser wind measuring radar with the meteorological data output by the mesoscale meteorological forecast module to form a unified data format; adapting the terrain and building parameters to the grid to ensure that the grid can accurately reflect the actual geographical and architectural environment of the airport and its surroundings; setting the boundary conditions around the airport based on the real-time wind field scanning data; using the CFD numerical simulation module to solve the three-dimensional Navier-Stokes equations, combining the terrain and building parameters, and calculating the parameters of each grid point.

[0008] In the step d, the specific steps of using the CFD numerical simulation module to calculate the sub-meter gridded three-dimensional wind field data at the key time point are as follows: first, according to the output data of the mesoscale meteorological forecast module, combined with the airport operation requirements and historical wind shear event analysis, the key time point for key simulation is determined; then the real-time wind field data obtained by the laser wind radar is integrated with the future meteorological forecast data provided by the mesoscale meteorological forecast module, which data includes wind speed, wind direction, temperature, humidity, and air pressure; then the terrain data and building parameters of the airport and its surroundings are imported to generate a sub-meter three-dimensional grid; the forecast data is used as the boundary condition, and the CFD numerical simulation module is used to solve the sub-meter gridded three-dimensional wind field data at the key time point; By combining the calculation results of key time points with the data obtained by real-time detection of the laser wind radar, the graph neural network calculation model is driven to obtain real-time sub-meter gridded three-dimensional wind field data. The specific steps are as follows: In the wind field calculation, the wind field data is regarded as a graph structure, in which the nodes represent the positions in space and the edges represent the relationship between the positions. The graph convolutional neural network of the graph neural network learns the representation of the nodes by aggregating the features of the nodes and their neighbors, thereby calculating the dynamic changes of the wind field. The calculation formula of the graph convolutional neural network is: , in: is the node v in the The feature vector of the layer, N(v) is the set of neighbor nodes of node v, c ij is the normalized weight between node i and node j, d v and d j are the degrees of node i and node j respectively, It is is the weight matrix of the layer, and σ is the activation function.

[0009] In the step e, the wind field data of the aircraft sliding down to the ground is extracted from the three-dimensional wind field data at different moments in the current and selected future, and the wind shear data on the glide path is calculated, and the specific steps of realizing wind shear warning and forecast are: from the three-dimensional wind field data at different moments in the current and selected future, the wind speed and wind direction on the glide path are extracted, and then a time series of wind field parameters for each grid point on the glide path is generated, and a threshold value of wind shear intensity is set according to historical data and flight safety standards. When the wind shear intensity exceeds the set threshold, an early warning is triggered.

[0010] In the step f, the specific steps of using the random forest model in machine learning to judge the current runway availability and the future runway availability are as follows: extracting the sub-meter three-dimensional wind field data, airport observation data, mesoscale meteorological forecast data and external message data at the current and future moments from the system server to perform multi-source data fusion and feature construction, converting the input data into a feature vector acceptable to the model through a preprocessing pipeline, inputting the trained random forest model, generating the runway availability prediction results at the current and future moments, and calculating the confidence of the prediction results; The mathematical model of random forest can be expressed as: , in, is the predicted value, K is the number of decision trees, is the predicted value of the kth decision tree, is the parameter of the kth decision tree, and x is the input feature vector.

[0011] In step g, the warning module is a dynamic warning system based on multi-level thresholds, and the multi-level warning thresholds correspond to different wind shear intensities, turbulent kinetic energy and runway availability risk levels; when the warning module detects potential dangerous wind field conditions, the warning module automatically triggers a warning notification to each user terminal.

[0012] In step f, the external data includes observation data, forecast data and related message data.

[0013] The beneficial effect of the present invention is as follows: the present invention combines laser wind measuring radar, CFD numerical simulation, mesoscale weather forecast model (WRF) and machine learning to build an airport wind shear monitoring and early warning system. By combining these four technologies, a sub-meter airport wind shear monitoring and early warning and available runway prediction system based on multi-source data fusion is designed. Not only can the wind shear over the airport be detected in real time, but also the sub-meter three-dimensional wind field structure can be reconstructed, so as to forecast and warn the future wind shear of the airport, and then realize the function of available runway prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a simplified structural block diagram of the system of the present invention; Figure 2 is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0015] like Figure 1 As shown in FIG. 1 , an airport available runway monitoring and early warning system based on multi-source data fusion includes: A laser wind measuring radar 1 for detecting wind fields within a set range and at a set height, and providing the system with high-precision, real-time wind shear detection data; CFD numerical simulation module 2 for forming high-precision gridded wind field data through simulation calculation; Mesoscale meteorological forecast module 3 for simulating future weather changes over a large area; Machine learning models for improving CFD calculation speed and predicting available runways at airports4; and The system server 5 combines the real-time wind shear detection data measured by the laser wind measuring radar 1 and the forecast data output by the mesoscale meteorological forecast module 3, assimilates the real-time detection data with the forecast data of the current time point output by the mesoscale meteorological forecast module 3, and then uses the CFD numerical simulation module 2 to process the assimilated data, combines the terrain and building parameters near the airport, generates real-time sub-meter gridded three-dimensional wind field data, and reconstructs the sub-meter three-dimensional wind field structure, so as to forecast and warn the future wind shear of the airport, and then predict the available runway.

[0016] The method for predicting the available runways of an airport using the above-mentioned airport available runway monitoring and early warning system based on multi-source data fusion comprises the following steps: a. Deployment and data acquisition of laser wind radar: Deploy the laser wind radar 1 at key locations in the airport and nearby areas, perform system calibration and performance testing on the laser wind radar 1 to ensure its stable operation and accurate capture of atmospheric information, thereby providing real-time, high-resolution real-time wind field scanning data to provide basic data support for subsequent operations; b. Calculation of future wind field data: Use the mesoscale weather forecast module 3 to simulate the future weather change trend of the airport area to obtain forecast data for a period of time in the future, then use the inversion algorithm and the CFD numerical simulation module 2 to process the forecast data, combine the terrain and building parameters near the airport, simulate the atmospheric state, and then generate sub-meter gridded three-dimensional wind field data of the airport at future times; c. Real-time wind field data calculation: After receiving and parsing the real-time wind field scanning data sent by the laser wind measuring radar 1 and the forecast data output by the mesoscale meteorological forecast module 3, the system server 5 performs data assimilation on the real-time wind field scanning data and the forecast data at the current time point output by the mesoscale meteorological forecast module 3, and then uses the CFD numerical simulation module 2 to process the assimilated data, and combines the terrain and building parameters near the airport to generate real-time sub-meter gridded three-dimensional wind field data; d. In the process of calculating real-time wind field data, only the CFD numerical simulation module 2 is used to calculate the sub-meter gridded three-dimensional wind field data at key time points, and then the calculation results at key time points are combined with the data obtained by real-time detection of the laser wind radar 1 to drive the graph neural network calculation model, thereby quickly obtaining real-time sub-meter gridded three-dimensional wind field data; e. Wind shear detection: Extract the wind field data of the aircraft sliding down to the ground from the three-dimensional wind field data at the current and selected future different times, and calculate the wind shear data on the glide path, thereby realizing wind shear warning and forecast; f. Judgment of available runways: Combined with the three-dimensional wind field data at different times in the current and selected future, as well as external data such as observation data, forecast data and related messages collected by the airport, the random forest model in machine learning is used to judge the current and future runway availability; g. Early warning system construction: Deploy an early warning module to automatically trigger early warning notifications when potential dangerous wind field conditions are detected, ensuring that each user terminal can take timely response measures.

[0017] In the step b, the specific steps of using the inversion algorithm and the CFD numerical simulation module 2 to process the forecast data are as follows: standardizing the forecast data output by the mesoscale meteorological forecast module 3, including time alignment, spatial interpolation and format conversion, and unifying it to a grid coordinate system compatible with the CFD numerical simulation module 2; optimizing the low-resolution forecast data using a variational inversion algorithm, and fusing the forecast data with the statistical characteristics of the historical lidar observation data by minimizing the objective function (including the weighted sum of squares of the background field error and the observation error), to generate a high-resolution initial wind field input; Combined with the terrain and building parameters near the airport, the atmospheric state is simulated and calculated, and the specific steps of generating the sub-meter gridded three-dimensional wind field data of the airport at the future moment are as follows: the inverted high-resolution wind field data is imported into the CFD numerical simulation module 2 as the initial condition, and the airport three-dimensional terrain model (DEM) and building geometric parameters (including height, surface roughness and layout topology) are loaded at the same time, and the boundary conditions are set as dynamic pressure and velocity coupling fields; based on the Reynolds-averaged Navier-Stokes equations (RANS) and the k-ε turbulence model, combined with the adaptive mesh refinement technology (AMR), the airport area wind field is transiently simulated and calculated, and the three-dimensional wind speed components u, v, w and turbulent kinetic energy distribution of the sub-meter grid nodes are solved; the sub-meter gridded three-dimensional wind field data (including timestamp, spatial coordinates and wind speed vector) finally generated are stored in the spatiotemporal database of the system server 5 for subsequent wind shear warning and runway prediction modules to call.

[0018] In step c, the specific steps of assimilating the real-time wind field scanning data with the forecast data are as follows: using the Newton relaxation data assimilation method to fuse the real-time wind field scanning data into the WRF model of the mesoscale meteorological forecast module 3, and the assimilation principle momentum equation is expressed as: , , where u i is the component of vector u in direction i, t is time, u j is the component of vector u in the j direction, v is the dynamic viscosity coefficient, x i is the coordinate component in the i direction, x j is the coordinate component in the j direction, ρ is the density, p is the pressure, f is the additional forcing term added according to the Nudging data assimilation method, and u c is the detection speed, u is the WRF prediction speed, τ u is the Newton relaxation coefficient, W(x,y,z) is the weight function; The specific steps of using the CFD numerical simulation module 2 to process the real-time wind field scanning data, combining the terrain and building parameters near the airport, and then generating sub-meter gridded three-dimensional wind field data are as follows: integrating the real-time wind field scanning data obtained by the laser wind measurement radar 1 with the meteorological data output by the mesoscale meteorological forecast module 3 to form a unified data format; adapting the terrain and building parameters to the grid to ensure that the grid can accurately reflect the actual geographical and architectural environment of the airport and its surroundings; setting the boundary conditions around the airport according to the real-time wind field scanning data, including wind speed, wind direction, air pressure, etc.; using the CFD numerical simulation module 2 to solve the three-dimensional Navier-Stokes equations, combining the terrain and building parameters, and calculating the wind speed, wind direction and other parameters of each grid point.

[0019] In the step d, the specific steps of using the CFD numerical simulation module 2 to calculate the sub-meter gridded three-dimensional wind field data at the key time point are as follows: first, according to the output data of the mesoscale meteorological forecast module 3, combined with the airport operation requirements and historical wind shear event analysis, the key time points for key simulation are determined, and these time points usually include the current time point, the time point when wind shear may occur in the future, and the peak period of flight take-off and landing at the airport, etc.; then the real-time wind field data obtained by the laser wind radar 1 is integrated with the future meteorological forecast data provided by the mesoscale meteorological forecast module 3, and these data include wind speed, wind direction, temperature, humidity, and air pressure; then the terrain data (such as elevation, slope, etc.) and building parameters (such as height, shape, distribution, etc.) of the airport and its surroundings are imported to generate a sub-meter three-dimensional grid; the forecast data is used as the boundary condition, and the CFD numerical simulation module 2 is used to solve the sub-meter gridded three-dimensional wind field data at the key time point; By combining the calculation results of key time points with the data obtained by real-time detection of the laser wind radar 1, the graph neural network calculation model is driven to obtain real-time sub-meter gridded three-dimensional wind field data. The specific steps are as follows: in the wind field calculation, the wind field data is regarded as a graph structure, in which the nodes represent the positions in space and the edges represent the relationship between the positions. The graph convolutional neural network of the graph neural network learns the representation of the nodes by aggregating the features of the nodes and their neighbors, thereby calculating the dynamic changes of the wind field. The calculation formula of the graph convolutional neural network is: , in: is the node v in the The feature vector of the layer, N(v) is the set of neighbor nodes of node v, c ij is the normalized weight between node i and node j, d v and d j are the degrees of node i and node j respectively, It is is the weight matrix of the layer, and σ is the activation function.

[0020] In the step e, the wind field data of the aircraft sliding down to the ground is extracted from the three-dimensional wind field data at different moments in the current and selected future, and the wind shear data on the glide path is calculated, and the specific steps of realizing wind shear warning and forecast are: from the three-dimensional wind field data at different moments in the current and selected future, the wind speed and wind direction on the glide path are extracted, and then a time series of wind field parameters for each grid point on the glide path is generated, and a threshold value of wind shear intensity is set according to historical data and flight safety standards. When the wind shear intensity exceeds the set threshold, an early warning is triggered.

[0021] In the step f, the specific steps of using the random forest model in machine learning to judge the current runway availability and the future runway availability are as follows: extracting the sub-meter-level three-dimensional wind field data (including wind speed, wind direction, turbulent kinetic energy), airport observation data (runway surface temperature, visibility, precipitation intensity), mesoscale meteorological forecast data (pressure gradient, temperature field) and external message data (flight scheduling, runway maintenance status) at the current and future times from the system server 5 to perform multi-source data fusion and feature construction, converting the input data into a feature vector acceptable to the model through a preprocessing pipeline, inputting the trained random forest model, generating the runway availability prediction results at the current and future times, and calculating the confidence of the prediction results; The mathematical model of random forest can be expressed as: , in, is the predicted value, K is the number of decision trees, is the predicted value of the kth decision tree, is the parameter of the kth decision tree, and x is the input feature vector.

[0022] In step g, the warning module is a dynamic warning system based on multi-level thresholds, and the multi-level warning thresholds correspond to different wind shear intensities, turbulent kinetic energy and runway availability risk levels; when the warning module detects potential dangerous wind field conditions, the warning module automatically triggers a warning notification to each user terminal.

[0023] The present invention will improve the monitoring and early warning capabilities of airport wind shear, prevent the impact of wind shear on aircraft take-off and landing, and provide a more scientific and accurate decision-making basis for airports through available runway prediction. The present invention can be promoted and used in domestic military and civil airports, effectively improving the safety level and operating efficiency of airports, and has huge market prospects.

[0024] Finally, it should be emphasized that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An airport available runway monitoring and early warning system based on multi-source data fusion, characterized by: It includes A laser wind radar (1) for detecting wind fields within a set range and at a set height, and providing the system with high-precision, real-time wind shear detection data; A CFD numerical simulation module (2) for generating high-precision gridded wind field data through simulation calculation; Mesoscale weather forecast module for simulating future weather changes over large areas (3); Machine learning models for improving CFD calculation speed and predicting airport runway availability (4); and A system server (5), wherein the system server (5) combines the real-time wind shear detection data measured by the laser wind radar (1) and the forecast data output by the mesoscale meteorological forecast module (3), performs data assimilation on the real-time detection data and the forecast data at the current time point output by the mesoscale meteorological forecast module (3), and then uses the CFD numerical simulation module (2) to process the assimilated data, combines the terrain and building parameters near the airport, generates real-time sub-meter gridded three-dimensional wind field data, and reconstructs the sub-meter three-dimensional wind field structure, thereby forecasting and warning the future wind shear of the airport, and further predicting the available runway.

2. The method for predicting the available runways of an airport using the airport available runway monitoring and early warning system based on multi-source data fusion as claimed in claim 1 is characterized in that: The method comprises the following steps: a. Deployment and data acquisition of laser wind radar: deploy the laser wind radar (1) at key locations in the airport and adjacent areas, perform system calibration and performance testing on the laser wind radar (1) to ensure its stable operation and accurate capture of atmospheric information, thereby providing real-time, high-resolution real-time wind field scanning data to provide basic data support for subsequent operations; b. Calculation of future wind field data: using the mesoscale weather forecast module (3) to simulate the future weather change trend in the area where the airport is located, to obtain forecast data for a period of time in the future, and then using the inversion algorithm and the CFD numerical simulation module (2) to process the forecast data, combining the terrain and building parameters near the airport, to simulate and calculate the atmospheric state, and then generate sub-meter gridded three-dimensional wind field data of the airport at the future time; c. Real-time wind field data calculation: after receiving and parsing the real-time wind field scanning data sent by the laser wind measuring radar (1) and the forecast data output by the mesoscale meteorological forecast module (3), the system server (5) performs data assimilation on the real-time wind field scanning data and the forecast data at the current time point output by the mesoscale meteorological forecast module (3), and then uses the CFD numerical simulation module (2) to process the assimilated data, and combines the terrain and building parameters near the airport to generate real-time sub-meter gridded three-dimensional wind field data; d. In the process of calculating the real-time wind field data, only the CFD numerical simulation module (2) is used to calculate the sub-meter gridded three-dimensional wind field data at the key time point, and then the calculation results at the key time point are combined with the data obtained by the real-time detection of the laser wind radar (1) to drive the graph neural network calculation model, so as to quickly obtain the real-time sub-meter gridded three-dimensional wind field data; e. Wind shear detection: Extract the wind field data of the aircraft sliding down to the ground from the three-dimensional wind field data at the current and selected future different times, and calculate the wind shear data on the glide path, thereby realizing wind shear warning and forecast; f. Runway availability judgment: Combine the current and selected future three-dimensional wind field data at different times, as well as external data collected by the airport, and use the random forest model in machine learning to judge the current and future runway availability; g. Early warning system construction: Deploy an early warning module to automatically trigger early warning notifications when potential dangerous wind field conditions are detected, ensuring that each user terminal can take timely response measures.

3. The method according to claim 2, characterized in that In step b, the specific steps of using the inversion algorithm and the CFD numerical simulation module (2) to process the forecast data are as follows: standardizing the forecast data output by the mesoscale meteorological forecast module (3), including time alignment, spatial interpolation and format conversion, and unifying it into a grid coordinate system compatible with the CFD numerical simulation module (2); optimizing the low-resolution forecast data using a variational inversion algorithm, and fusing the forecast data with the statistical features of the historical lidar observation data by minimizing the objective function to generate a high-resolution initial wind field input; The specific steps of simulating and calculating the atmospheric state in combination with the terrain and building parameters near the airport, and then generating the sub-meter gridded three-dimensional wind field data of the airport at a future moment are as follows: importing the inverted high-resolution wind field data as the initial condition into the CFD numerical simulation module (2), loading the airport three-dimensional terrain model and building geometric parameters at the same time, and setting the boundary condition as the dynamic pressure and velocity coupling field; based on the Reynolds averaged Navier-Stokes equation and the k-ε turbulence model, combined with the adaptive grid refinement technology, performing transient simulation calculations on the airport area wind field, solving and obtaining the three-dimensional wind speed components u, v, w and turbulent kinetic energy distribution of the sub-meter grid nodes; and storing the finally generated sub-meter gridded three-dimensional wind field data in the spatiotemporal database of the system server (5) for subsequent wind shear warning and runway prediction modules to call.

4. The method according to claim 2, characterized in that: In step c, the specific steps of assimilating the real-time wind field scanning data with the forecast data are as follows: using the Newton relaxation data assimilation method to fuse the real-time wind field scanning data into the WRF model of the mesoscale meteorological forecast module (3), and the assimilation principle momentum equation is expressed as: ， ， where u i is the component of vector u in direction i, t is time, u j is the component of vector u in the j direction, v is the dynamic viscosity coefficient, x i is the coordinate component in the i direction, x j is the coordinate component in the j direction, ρ is the density, p is the pressure, f is the additional forcing term added according to the Nudging data assimilation method, and u c is the detection speed, u is the WRF prediction speed, τ u is the Newton relaxation coefficient, W(x,y,z) is the weight function; The specific steps of using the CFD numerical simulation module (2) to process the real-time wind field scanning data and combining it with the terrain and building parameters near the airport to generate sub-meter gridded three-dimensional wind field data are as follows: integrating the real-time wind field scanning data obtained by the laser wind radar (1) with the meteorological data output by the mesoscale meteorological forecast module (3) to form a unified data format; adapting the terrain and building parameters to the grid to ensure that the grid can accurately reflect the actual geographical and architectural environment of the airport and its surroundings; According to the real-time wind field scanning data, the boundary conditions around the airport are set; the CFD numerical simulation module (2) is used to solve the three-dimensional Navier-Stokes equation, and the parameters of each grid point are calculated by combining the terrain and building parameters.

5. The method according to claim 2, characterized in that: In the step d, the specific steps of using the CFD numerical simulation module (2) to calculate the sub-meter gridded three-dimensional wind field data at the key time point are as follows: first, based on the output data of the mesoscale meteorological forecast module (3), combined with the airport operation requirements and historical wind shear event analysis, the key time point for key simulation is determined; then, the real-time wind field data obtained by the laser wind radar (1) is integrated with the future meteorological forecast data provided by the mesoscale meteorological forecast module (3), which data includes wind speed, wind direction, temperature, humidity, and air pressure; then, the terrain data and building parameters of the airport and its surroundings are imported to generate a sub-meter gridded three-dimensional grid; the forecast data is used as a boundary condition, and the CFD numerical simulation module (2) is used to solve the sub-meter gridded three-dimensional wind field data at the key time point; By combining the calculation results at key time points with the data obtained by real-time detection of the laser wind radar (1), the graph neural network calculation model is driven to obtain real-time sub-meter gridded three-dimensional wind field data. The specific steps are as follows: in wind field calculation, the wind field data is regarded as a graph structure, in which nodes represent positions in space and edges represent the relationship between positions. The graph convolutional neural network of the graph neural network learns the representation of nodes by aggregating the features of nodes and their neighbors, thereby calculating the dynamic changes of the wind field. The calculation formula of the graph convolutional neural network is: ， in: is the node v in the The feature vector of the layer, N(v) is the set of neighbor nodes of node v, c ij is the normalized weight between node i and node j, d v and d j are the degrees of node i and node j respectively, It is is the weight matrix of the layer, and σ is the activation function.

6. The method according to claim 2, characterized in that In the step e, the wind field data of the aircraft sliding down to the ground is extracted from the three-dimensional wind field data at different moments in the current and selected future, and the wind shear data on the glide path is calculated, and the specific steps of realizing wind shear warning and forecast are: from the three-dimensional wind field data at different moments in the current and selected future, the wind speed and wind direction on the glide path are extracted, and then a time series of wind field parameters for each grid point on the glide path is generated, and a threshold value of wind shear intensity is set according to historical data and flight safety standards. When the wind shear intensity exceeds the set threshold, an early warning is triggered.

7. The method according to claim 2, characterized in that In step f, the specific steps of using the random forest model in machine learning to judge the current runway availability and the future runway availability are as follows: extracting sub-meter-level three-dimensional wind field data, airport observation data, mesoscale meteorological forecast data and external message data at the current and future times from the system server (5) to perform multi-source data fusion and feature construction, converting the input data into a feature vector acceptable to the model through a preprocessing pipeline, inputting the trained random forest model, generating the runway availability prediction results at the current and future times, and calculating the confidence of the prediction results; The mathematical model of random forest can be expressed as: ， in, is the predicted value, K is the number of decision trees, is the predicted value of the kth decision tree, is the parameter of the kth decision tree, and x is the input feature vector.

8. The method according to claim 2, characterized in that: In step g, the warning module is a dynamic warning system based on multi-level thresholds, and the multi-level warning thresholds correspond to different wind shear intensity, turbulent kinetic energy and runway availability risk levels respectively; When the early warning module detects a potentially dangerous wind field condition, the early warning module automatically triggers an early warning notification to each user terminal.

9. The method according to claim 2, characterized in that: In step f, the external data includes observation data, forecast data and related message data.

Images