An intelligent identification method of airport area wind shear based on multi-source wind field data

Abstract

The invention discloses an intelligent identification method of wind shear in an airport area based on multi-source wind field data, comprising the following steps: 1) generating a three-dimensional refined wind field in an airport area based on the multi-source wind field data; 1-1: constructing a numerical simulation model of a refined wind field in an airport area; 1-2: constructing a simulation wind field database based on a CFD model; 1-3: calculating the matching degree between the current measured multi-source wind field data and each group of CFD data, and finding the CFD simulation data with the highest matching degree with the measured multi-source wind field data; 2) fusing the wind field data based on CFD technology and WRF mode to obtain a three-dimensional wind field; 3) numerically correcting the fused three-dimensional wind field based on the current measured data to obtain digital wind field data of the airport area without any difference in all weather conditions; 4) intelligently identifying wind shear based on a convolutional neural network to obtain the location and shear intensity of the outgoing wind shear, and marking the wind shear in a three-dimensional wind field data map in combination with the refined three-dimensional wind field data.

Description

Technical Field

[0001] The present invention relates to the technical field of wind shear identification, and in particular to an intelligent identification method for wind shear in an airport area based on multi-source wind field data. Background Art

[0002] For aviation, flight safety is the first factor. In actual flight activities, there are many factors that affect flight safety, among which low-altitude wind shear is one of the most important factors. Since the aircraft is at low altitude during takeoff and landing, and the aircraft flies at a slow speed, if it encounters low-altitude wind shear at this time, it is difficult for the pilot to respond in time, and serious flight accidents are very likely to occur. Therefore, low-altitude wind shear is also called the "invisible aircraft killer". Timely identification of wind shear and issuing early warning information is an important part of flight safety assurance. Wind shear identification and early warning, two core technologies are designed, one is wind field monitoring, that is, wind field information acquisition; the other is wind shear identification algorithm, that is, wind shear information mining based on acquired data. With the advancement of technology, a variety of wind field detection methods have been developed, and each wind field detection method has developed a corresponding wind shear identification algorithm.

[0003] As early as the 1970s, the United States began to monitor wind shear caused by downbursts by building multiple wind stations at airports, which was called the Low-Level Wind Shear Alert System (LLWAS). The original LLWAS-I consisted of 6 wind stations, 1 deployed in the middle of the runway and the other 5 deployed around the runway. Wind shear was detected by detecting the wind speed difference between the central wind station and other wind stations. However, due to the large spacing between the 6 wind stations, when a smaller-scale micro-downburst occurs between two wind stations, it is difficult to detect. Therefore, LLWAS-I cannot detect micro-downbursts, and there is a high possibility of issuing false alarms. The improved LLWAS-II upgraded the software algorithm to improve the accuracy of detecting micro-downbursts and reduce the number of false alarms. At present, LLWAS-II has been completely eliminated and replaced by LLWAS-III, which can install up to 32 wind measurement sensors and adopt a new triangular area detection algorithm to improve the accuracy of warnings while reducing the false alarm rate. LLWAS-III divides wind sensors into groups of three or two, and calculates the wind field and its changes based on the wind speed and direction detected by the sensors and the relative positions of the sensors. If the wind field is divergent, the micro-downburst model established based on the symmetry assumption is used to calculate the change in wind speed and determine the outflow center of the downburst. However, the monitoring height of the LLWAS system is limited, and it can only detect horizontal wind shear near the surface, and the monitoring effect is affected by the number and location of surface wind stations.

[0004] With the application of Doppler weather radar in meteorological services, wind shear algorithms have developed rapidly. In 1985, foreign scholars proposed a difference filtering synthetic shear algorithm that synthesizes two-dimensional wind shear by calculating radial and tangential wind shears along the weather radar respectively. It can be used to identify wind shear caused by cold fronts, shear lines, convergence lines, cyclones, etc. At the same time, some scholars have improved the existing mesocyclone identification algorithm and proposed a gust front identification algorithm based on radial velocity convergence lines. In 1989, some scholars added a double verification of the continuity of the vertical wind field to the algorithm, realizing the identification of wind shear caused by gust fronts. In 2013, some scholars used network fusion technology to identify the location of wind shear by distinguishing the differences between adjacent grid points. In 2000, domestic scholars proposed a method of calculating synthetic wind shear using the least squares method, which successfully detected wet wind shear in airport airspace. In 2013, domestic scholars used the bidirectional gradient and convergence line identification algorithm to accurately locate the front of the gust front and successfully identify the wind shear caused by the gust front. Therefore, after years of development and improvement, the technology of identifying wet wind shear caused by gust fronts and downbursts based on Doppler weather radar has become relatively mature, but there are still many problems in detecting dry wind shear caused by aircraft wake vortex, radiation inversion, terrain obstacles, etc.

[0005] With the continuous maturity of laser wind radar technology, the wind shear monitoring algorithm of laser wind radar has developed rapidly. At present, the mainstream algorithms for laser wind radar monitoring wind shear are the IRIS algorithm of Vaisala Company in Finland and the Rainbow algorithm of Germany. These two algorithms have been widely used in many large international airports in the world with good results. The research team of Hong Kong Observatory in China was the first to carry out the research on the wind shear monitoring algorithm of laser wind radar, and innovatively proposed the glide path scanning method and the corresponding single slope detection algorithm, which were applied to the business support of Hong Kong Airport. At the same time, the research team of Hong Kong Observatory proposed the wind shear identification algorithm based on F factor and eddy current dissipation rate in 2012 and 2014 respectively. Civil Aviation University of China has carried out a large number of wind shear identification algorithm research. First, the wind shear warning algorithm of the Hong Kong Observatory research team was improved, and the glide path double slope detection algorithm and the small-scale wind shear detection algorithm with modified F factor were proposed; at the same time, for the wind field data under the PPI scanning mode of laser wind radar, a gradient search wind shear warning algorithm based on adaptive multi-scale was proposed to improve the wind shear warning rate. Laser wind radar has obvious advantages over traditional wind measurement technology. However, since the laser will be attenuated by air molecules, aerosol particles, cloud and rain particles when transmitted in the atmosphere, the detection performance of laser wind radar will be affected by weather conditions and is not suitable for the detection of humid wind fields.

[0006] In recent years, with the development of artificial intelligence technology, many wind shear intelligent recognition algorithms have been proposed. The Civil Aviation University of China team has done a lot of work in this regard: in 2012, a wind shear type recognition algorithm based on image processing methods was proposed. Compared with the traditional wind shear recognition algorithm, this algorithm ignores meteorological factors and completely transforms it into an image processing problem, realizing the type recognition of low-altitude wind shear; in 2013, a support vector machine parameter optimization method based on a predatory search strategy genetic algorithm was proposed for wind shear recognition; in the same year, a method for identifying the type of low-altitude wind shear was proposed using texture features extracted based on wavelet transform and BP neural network; in 2014, a probability method was proposed to solve the problem that the standard support vector machine cannot provide a posteriori probability in lidar wind shear image recognition. Support vector machine recognition method; in the same year, based on the shape characteristic relationship between the sample images of microburst, low-level jet, headwind and crosswind low-level wind shear, it was proposed to apply the feature extraction technology of wavelet invariant moment to wind shear recognition; in 2015, a texture feature fusion method combining rotation-invariant non-subsampled Contourlet transform and Weber descriptor was proposed; in the same year, for the laser radar scanning images of four different low-level wind shears, microburst, low-level jet, headwind and crosswind, a recognition method based on the combination of shape features and texture features was proposed; in 2019, a multi-layer feature extraction and adaptive fusion algorithm based on deep convolutional neural network was proposed. Although a variety of wind shear recognition algorithms have been proposed, each algorithm requires specific applicable conditions, and the recognition effect under different conditions varies greatly, and the actual effect remains to be tested.

[0007] In summary, the existing technology theoretically cannot achieve extremely ideal results through the shear identification algorithm of a single device, and cannot solve the wind shear identification based on multi-source wind field data, including the detection of dry or wet wind shear caused by different reasons. Summary of the invention

[0008] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides a comprehensive intelligent identification method for wind shear in the airport area based on multi-source wind field data. The invention can fuse multiple wind measurement results to obtain a three-dimensional refined wind field in the protection area, and then mine wind shear information from the three-dimensional wind field, and finally obtain a three-dimensional wind field data map marked with wind shear identification results.

[0009] Technical solution: In order to achieve the above invention purpose, the present invention provides an intelligent identification method of airport area wind shear based on multi-source wind field data, comprising the following steps:

[0010] Step 1: Generate a three-dimensional refined wind field in the airport area based on multi-source wind field data;

[0011] Step 2: Based on CFD technology and WRF model, wind field data is integrated to obtain a three-dimensional wind field; WRF refers to the weather forecast model, the mesoscale weather forecast model;

[0012] Step 3: Based on the current measured data, the fused three-dimensional wind field is numerically corrected to obtain the digital wind field data of the airport area without any difference in all weather conditions;

[0013] Step 4: Based on the convolutional neural network, the wind shear is intelligently identified to obtain the location and shear intensity of the wind shear, and the wind shear is marked in the three-dimensional wind field data map in combination with the refined three-dimensional wind field data.

[0014] Furthermore, in the present invention: the generation of a three-dimensional refined wind field further includes the following steps:

[0015] Step 1-1: Construct a refined wind field numerical simulation model for the airport area;

[0016] Step 1-2: Construct a simulated wind farm database based on the CFD model;

[0017] Step 1-3: Calculate the matching degree between the current measured multi-source wind field data and each group of CFD data, and find the CFD simulation data with the highest matching degree with the measured multi-source wind field data;

[0018] Step 1-4: Numerical correction of CFD simulation data based on current measured multi-source wind field data;

[0019] Step 1-5: Obtain the current three-dimensional refined wind field after data reconstruction.

[0020] Furthermore, in the present invention: the wind field data fusion further comprises the following steps:

[0021] Step 2-1: Export WRF data and process it;

[0022] Step 2-2: Generate CFD refined network;

[0023] Step 2-3: Exchange data between WRF and CFD;

[0024] Step 2-4: Finally output WRF data and process it.

[0025] Furthermore, in the present invention: the numerical correction further comprises the following steps:

[0026] Step 3-1: For areas with reliable measured wind field data, the measured data are mainly used;

[0027] Step 3-2: For areas without measured wind field data, supplement them with CFD simulation data under the current meteorological conditions;

[0028] Step 3-3: When there are two or more wind measurement results at the same location, a weighted calculation is performed. The weight of each wind measurement device is determined by the weather conditions at the time, the wind measurement principle of each device, and the observation location;

[0029] Step 3-4: Areas with measured wind field data have different weights of influence on wind field data within a certain range around them. The specific influence range and influence weight need to be determined in combination with sensitivity experiments.

[0030] Furthermore, in the present invention: the intelligent identification of wind shear based on convolutional neural network also includes the following steps:

[0031] Step 4-1: Construct a wind shear recognition convolutional neural network structure and establish a wind field plane image set with wind shear characteristics;

[0032] Step 4-2: Extract image features and train wind shear recognition convolutional neural network node parameters;

[0033] Step 4-3: Construct a plane scanning scheme for regional refined three-dimensional wind field data;

[0034] Step 4-4: Output the location and shear intensity of wind shear;

[0035] Step 4-5: Mark the shear position and shear intensity in the 3D wind field data map.

[0036] The present invention also proposes an intelligent identification method for airport area wind shear based on multi-source wind field data, comprising the following steps:

[0037] Step 1: Each wind measuring device measures the airport area separately;

[0038] Step 2: Fusion of measurement results based on priority;

[0039] Step 3: Obtain wind shear warning information based on the fusion results.

[0040] Furthermore, in the present invention: the wind measuring equipment includes a wind direction and speed meter, a Doppler weather radar, a wind profiler and a laser wind measuring radar.

[0041] Further, in the present invention: the wind direction and anemometer are set on the top of the wind pole, and the wind pole is set around the airport area; the Doppler weather radar is used for wet wind field detection; the wind profiler is used for dry wind field detection; the laser wind radar is used for slideway detection in the airport area. Construct a CFD wind field simulation model. The types and quantities of wind measuring instruments that need to be deployed at multiple points. Accurate wind measuring instruments are required on site. Due to the advantages of laser wind radar, laser wind radar is mainly used in combination with ground wind poles. When the station has other wind measuring instruments (wind profiler, Doppler weather radar, etc.), the measurement results can also be connected to this system in the required data format. It is necessary to minimize the number of instruments under the premise of accuracy and effectiveness of regional wind field monitoring. By obtaining the wind field information of the set detection node, the wind field information of some nodes is used as the initial value, and the wind field information of another part of the nodes is used as the target value. By bringing in the initial value, combined with the established CFD wind field simulation model, the wind field information of the target value is calculated, compared, the most sensitive position is found, and the corresponding instrument is placed. Obtain wind field data from multiple sources.

[0042] Beneficial effect: Compared with the prior art, the beneficial effect of the present invention is that the wind shear identification method proposed in the present invention is based on multi-source wind field data and a wind shear identification algorithm of a convolutional neural network, which can detect wind shear information under different meteorological conditions and realize accurate identification of wind shear under a variety of meteorological conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the overall process of the airport area wind shear intelligent identification method based on multi-source wind field data proposed by the present invention;

[0044] Figure 2 A schematic diagram of the process of generating a three-dimensional refined wind field in the present invention;

[0045] Figure 3 It is a schematic diagram of the flow of wind field data fusion in the present invention;

[0046] Figure 4 It is a schematic diagram of the process of intelligently identifying wind shear based on convolutional neural network in the present invention. DETAILED DESCRIPTION

[0047] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings:

[0048] The present invention may be implemented in many different forms and should not be considered limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0049] Due to the different wind measurement principles of different equipment, the detection accuracy of each wind measurement equipment under different meteorological conditions is different, resulting in different success rates of wind shear warning. In general, the wind direction and anemometer have the highest detection accuracy, but the wind direction and anemometer are usually installed on the top of a 10m high wind pole and can only detect single-point horizontal wind in the near-ground layer. The detection area and height are limited. To detect the wind field in a larger area, the number of stations needs to be increased. Due to its limited detection height, it can only be used to monitor low-altitude wind shear caused by downbursts. The tracer of Doppler weather radar is cloud and rain particles, so it can only detect areas with clouds and rain. It has a certain effect on wet low-altitude wind shear accompanied by downbursts, but its working blind area is large and the distance resolution is poor. It is difficult to aim at the aircraft glide path and runway area for observation, and the detection effect of small and micro-scale wind shear such as micro-downbursts is poor; and it is limited by the tracer itself and does not have the ability to detect dry wind shear. The tracer of wind profiler is atmospheric turbulence, which is suitable for dry wind field detection. However, a single wind profiler can only detect the vertical wind field information in the overhead area. If the detection range is to be expanded, multiple network detections are required, which is economical and difficult to implement. In addition, its detection results are easily interfered by ground clutter, the reliability of the near-ground wind field is low, and there are detection blind spots, which is not conducive to identifying low-altitude wind shear. The tracer of laser wind radar is aerosol particles, which is suitable for dry wind field detection. It has unique advantages such as strong anti-interference ability, small blind spots, high data spatial resolution, and the ability to quickly obtain refined low-altitude three-dimensional wind field information. It has been widely used in aviation safety assurance work. Since the laser will be attenuated by air molecules, aerosol particles, and cloud and rain particles when transmitting in the atmosphere, the detection performance of laser wind radar will be affected by weather conditions. It is not suitable for the detection of wet wind fields, but it is the best detection method for dry wind fields, especially for wind field detection and wind shear monitoring in key areas such as glideslopes and runways.

[0050] like Figure 1 FIG. 1 is a schematic diagram of the overall process of an intelligent identification method for wind shear in an airport area based on multi-source wind field data proposed by the present invention. The method comprises the following steps:

[0051] Step 1: Generate a three-dimensional refined wind field in the airport area based on multi-source wind field data;

[0052] Step 2: Based on CFD technology and WRF mode, wind field data is integrated to obtain a three-dimensional wind field. The computational fluid dynamics (CFD) model is a discrete solution to the fluid dynamics control equations by numerical methods. It is used to simulate the near-ground wind field and can fully reproduce the flow of air under complex terrain. It is the development trend of wind field analysis technology. According to the geographical model of the business support area, CFD numerical simulation is carried out to obtain CFD wind field simulation results with different directions, different wind speeds, and different gust characteristics. The selection of the grid size of the business support area needs to be determined in combination with the terrain conditions to achieve the minimum calculation amount and the best simulation results.

[0053] Step 3: Based on the current measured data, the fused three-dimensional wind field is numerically corrected to obtain the digital wind field data of the airport area without any difference in all weather conditions;

[0054] Step 4: Based on the convolutional neural network, the wind shear is intelligently identified to obtain the location and shear intensity of the wind shear, and the wind shear is marked in the three-dimensional wind field data map in combination with the refined three-dimensional wind field data.

[0055] Wherein, the generation of a three-dimensional refined wind field further includes the following steps:

[0056] Step 1-1: Construct a refined wind field numerical simulation model for the airport area;

[0057] Among them, the refined wind field numerical simulation model of the airport area also includes collecting the topographic and geomorphic characteristics of the airport area and obtaining the historical wind field data of the airport area, and performing characteristic analysis on the historical wind field of the airport area to generate model boundary conditions, and combining the topographic and geomorphic characteristics of the airport area to obtain the refined wind field numerical simulation model of the airport area.

[0058] Step 1-2: Construct a simulated wind farm database based on the CFD model;

[0059] Step 1-3: Calculate the matching degree between the current measured multi-source wind field data and each group of CFD data, and find the CFD simulation data with the highest matching degree with the measured multi-source wind field data;

[0060] Step 1-4: Numerical correction of CFD simulation data based on current measured multi-source wind field data;

[0061] Step 1-5: Obtain the current three-dimensional refined wind field after data reconstruction.

[0062] Among them, the wind field data fusion algorithm is used to process direct wind measurement results obtained by various means, which are inconsistent in time and space and have different measurement accuracy. After fusion, the digital wind field data of the airport area with no difference in all weather conditions is obtained for further wind shear identification. In this embodiment, the measured wind field data is combined with the wind field simulation results of the CFD model and the mesoscale WRF model to perform wind field data fusion, and the three-dimensional wind field obtained by fusion is numerically corrected based on the current measured data.

[0063] Specifically, the wind field data fusion further includes the following steps:

[0064] Step 2-1: Export WRF data and process it;

[0065] The step further includes:

[0066] Step 2-1-1: WRF simulates the coarse-resolution wind field at a specific airport;

[0067] Step 2-1-2: Process the wind field data of the side boundary by interpolation;

[0068] Step 2-1-3: Polynomial fitting of the function of the lateral boundary wind field data changing with height.

[0069] Step 2-2: Generate CFD refined network;

[0070] The step further includes:

[0071] Step 2-2-1: Generate a high-resolution grid based on STL format elevation data;

[0072] Step 2-2-2: SnappyHexMesh generates the calculation network and imports it into CFD.

[0073] Step 2-3: Exchange data between WRF and CFD;

[0074] The step further includes:

[0075] Step 2-3-1: Write the simulated side boundary function into the initial CFD file;

[0076] Step 2-3-2: Use the groovy function library to assign values ​​to the side boundary grid.

[0077] Step 2-4: Finally output WRF data and process it.

[0078] The step further includes:

[0079] Step 2-4-1: Monitor WRF output and inflow direction;

[0080] Step 2-4-2: WRF and CFD exchange data at each step;

[0081] Step 2-4-3: CFD fine grid numerical calculation and output results.

[0082] Specifically, data assimilation is used to perform numerical correction, and the numerical correction also includes the following steps:

[0083] Step 3-1: For areas with reliable measured wind field data, the measured data are mainly used;

[0084] Step 3-2: For areas without measured wind field data, supplement them with CFD simulation data under the current meteorological conditions;

[0085] Step 3-3: When there are two or more wind measurement results at the same location, a weighted calculation is performed. The weight of each wind measurement device is determined by the weather conditions at the time, the wind measurement principle of each device, and the observation location;

[0086] Step 3-4: Areas with measured wind field data have different weights of influence on wind field data within a certain range around them. The specific influence range and influence weight need to be determined in combination with sensitivity experiments.

[0087] Wherein, the intelligent identification of wind shear based on convolutional neural network also includes the following steps:

[0088] Step 4-1: Construct a wind shear recognition convolutional neural network structure and establish a wind field plane image set with wind shear characteristics;

[0089] Step 4-2: Extract image features and train wind shear recognition convolutional neural network node parameters;

[0090] Step 4-3: Construct a plane scanning scheme for regional refined three-dimensional wind field data;

[0091] Step 4-4: Output the location and shear intensity of wind shear;

[0092] Step 4-5: Mark the shear position and shear intensity in the 3D wind field data map.

[0093] The present invention also proposes an intelligent identification method for wind shear in an airport area based on multi-source wind field data, the method comprising the following steps:

[0094] Step 1: Each wind measuring device measures the airport area respectively; wherein the wind measuring equipment includes a wind direction and speed meter, a Doppler weather radar, a wind profiler and a laser wind measuring radar.

[0095] Specifically, the wind direction and anemometer are mounted on the top of a wind pole, and the wind pole is arranged around the airport area. In this embodiment, the height of the wind pole is not less than 10m; the Doppler weather radar is used for wet wind field detection; the wind profiler is used for dry wind field detection; the laser wind radar is used for slideway detection in the airport area; the measurement results are fused based on the priority; and the fusion warning of wind shear identification results of multiple devices is carried out. The fusion is reflected in the release of wind shear identification results in different spatial locations and under different weather conditions.

[0096] Specifically, the wind direction and anemometer have the highest detection accuracy, so its measurement results are set as the first priority for release. However, the wind direction and anemometer can only be used to detect single-point horizontal wind in the near-ground layer, and the detection area and height are limited. Therefore, it can only be used to monitor low-altitude wind shear caused by downbursts, etc. In order to detect wind fields in larger areas, the number of sites to be built needs to be increased.

[0097] The tracers of Doppler weather radar are cloud and rain particles, so it can only detect areas with clouds and rain, that is, it is used to detect wet wind fields, and the prediction results of wet low-altitude wind shear accompanied by downbursts are relatively accurate. However, the working blind area of ​​Doppler weather radar is large, the distance resolution is poor, and it is difficult to aim at the aircraft glide path and runway area for observation. The detection effect of small and micro-scale wind shear such as micro-downbursts is poor, and it is limited by the tracer itself and does not have the ability to detect dry wind shear. Therefore, for wind shear warning information in airport areas outside the glide path, under wet conditions, that is, cloudy, rainy, snowy, and foggy weather, the prediction results of Doppler weather radar are released as the first priority.

[0098] The tracer of wind profiler is atmospheric turbulence, which is suitable for dry wind field detection. However, a single wind profiler can only detect vertical wind field information in the overhead area, and the detection range is small. If the detection range needs to be expanded, multiple networked detections are required, which is economical and difficult to implement in practice. In addition, the detection results of wind profiler are easily interfered by ground clutter, the reliability of near-ground wind field is low, and there are detection blind spots, which is not conducive to identifying low-altitude wind shear. Therefore, the detection results of wind profiler are supplementary information and have the lowest priority.

[0099] The tracer of laser wind radar is aerosol particles, which is suitable for dry wind field detection. It has the unique advantages of strong anti-interference ability, small blind area, high data spatial resolution, and the ability to quickly obtain refined low-altitude three-dimensional wind field information. It has been widely used in aviation safety assurance. Since the laser will be attenuated by air molecules, aerosol particles, and cloud and rain particles when it is transmitted in the atmosphere, the detection performance of laser wind radar will be affected by weather conditions and is not suitable for the detection of wet wind fields. However, it is the best detection method for dry wind fields, especially for wind field detection and wind shear monitoring in key areas such as glide paths and runways. Therefore, for wind shear warning information in the glide path area, laser wind radar is issued as the first priority.

[0100] The fusion warning here refers to the comprehensive warning of the identification results of multiple devices, that is, it is not a warning based on the identification results of a single device. The fusion is reflected in the method of issuing wind shear identification results in different spatial locations and different weather conditions.

[0101] The specific solution refers to the fusion warning of wind shear identification results of multiple devices. The steps are as follows: 1) Using the precise terrain information obtained by GIS, combined with the geomorphic characteristics of the business guarantee area (distinguishing between winter, summer, spring and autumn), a geographic information model of the business guarantee area containing information such as elevation and roughness is constructed; 2) Constructing a computational fluid dynamics (CFD) wind field simulation model: The computational fluid dynamics (CFD) model is a discrete solution to the fluid mechanics control equations solved by numerical methods. It is used to simulate the near-ground wind field and can fully reproduce the flow of air under complex terrain. It is the development trend of wind field analysis technology; according to the geographic model of the business guarantee area, CFD numerical simulation is carried out to obtain CFD wind field simulation results with different directions, different wind speeds and different gust characteristics; the selection of the grid size of the business guarantee area needs to be determined in combination with the terrain and geomorphic conditions to achieve the minimum calculation amount and the best simulation results;

[0102] 3) Observation point sensitivity analysis, determine the types and quantities of wind measuring instruments that need to be deployed at multiple points in the business support area according to the terrain characteristics; accurate wind measuring instruments are needed on site, and laser wind measuring radars combined with ground wind poles are used to obtain wind field information of set detection nodes, and the wind field information of some nodes is used as the initial value, and the wind field information of other nodes is used as the target value. By bringing in the initial value and combining it with the established CFD wind field simulation model, the wind field information of the target value is calculated and compared to find the most sensitive position and place the corresponding instrument.

[0103] 4) Establish a regional wind field monitoring system; according to the results of the sensitivity test in step 3), construct a multi-point wind field observation system, and based on the measured wind field observation results, use the CFD wind field simulation model established under geographical conditions, and use a table lookup method combined with an optimized interpolation method to obtain the wind field in any horizontal and vertical directions within the area.

[0104] According to specific business needs, safety warnings and safety time window information are provided for the flight takeoff and landing conditions of different aircraft flight conditions such as crosswind, low-altitude horizontal wind shear and vertical wind shear, and medium- and high-altitude horizontal wind shear and vertical wind shear.

[0105] The overall process of CFD simulation of wind farms under various terrains is divided into three parts: pre-processing, simulation calculation and post-processing. The pre-processing part includes the acquisition and processing of terrain data, ground and simulation area modeling and grid generation, which is a prerequisite for the CFD simulation of wind farms; the simulation calculation part includes boundary condition setting (inlet wind profile, outlet self-outflow, wall and symmetric boundary, etc.), turbulence model selection, discrete format selection, solver selection and simulation initialization, which is the main solution process of CFD simulation of wind farms: the post-processing part includes the extraction and analysis of simulation results, the purpose of which is to more intuitively conduct qualitative and quantitative tests on the rationality of the CFD simulation results of wind farms, and extract relevant data for wind field analysis on this basis.

[0106] When the station has other wind measuring instruments (wind profiler, Doppler weather radar, etc.), the measurement results can also be connected to this system in the required data format. The wind profiler detection results as auxiliary information are integrated with the results of other equipment.

[0107] Step 3: Obtain wind shear warning information based on the fusion results.

[0108] It should be noted that the above-described embodiments only express some implementation methods of the present invention, and the description thereof cannot be understood as limiting the scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements can be made without departing from the concept of the present invention, and these should all fall within the scope of protection of the present invention.

Claims

1. An intelligent identification method for wind shear in an airport area based on multi-source wind field data, characterized in that: The following steps are included: Step 1: Generate a three-dimensional refined wind field in the airport area based on multi-source wind field data; Step 2: Based on CFD technology and WRF model, wind field data is fused to obtain a three-dimensional wind field; Step 3: Based on the current measured data, the fused three-dimensional wind field is numerically corrected to obtain the digital wind field data of the airport area without any difference in all weather conditions; Step 4: Intelligently identify wind shear based on convolutional neural network to obtain the location and shear intensity of wind shear, and mark wind shear in the three-dimensional wind field data map in combination with refined three-dimensional wind field data; The step 1 of generating a three-dimensional refined wind field also includes the following steps: Step 1-1: Construct a refined wind field numerical simulation model for the airport area; Step 1-2: Construct a simulated wind farm database based on the CFD model; Step 1-3: Calculate the matching degree between the current measured multi-source wind field data and each group of CFD data, and find the CFD simulation data with the highest matching degree with the measured multi-source wind field data; Step 1-4: Numerical correction of CFD simulation data based on current measured multi-source wind field data; Step 1-5: Obtain the current three-dimensional refined wind field after data reconstruction; The step 2 wind field data fusion also includes the following steps: Step 2-1: Export WRF data and process it; Step 2-2: Generate CFD refined network; Step 2-3: Exchange data between WRF and CFD; Step 2-4: Finally output WRF data and process it.

2. The method for intelligently identifying wind shear in an airport area based on multi-source wind field data according to claim 1, characterized in that: The numerical correction also includes the following steps: Step 3-1: For areas with reliable measured wind field data, the measured data are mainly used; Step 3-2: For areas without measured wind field data, supplement them with CFD simulation data under the current meteorological conditions; Step 3-3: When there are two or more wind measurement results at the same location, a weighted calculation is performed. The weight of each wind measurement device is determined by the weather conditions at the time, the wind measurement principle of each device, and the observation location; Step 3-4: Areas with measured wind field data have different weights of influence on wind field data within a certain range around them. The specific influence range and influence weight need to be determined in combination with sensitivity experiments.

3. The method for intelligently identifying wind shear in an airport area based on multi-source wind field data according to claim 1, characterized in that: The intelligent identification of wind shear based on convolutional neural network also includes the following steps: Step 4-1: Construct a wind shear recognition convolutional neural network structure and establish a wind field plane image set with wind shear characteristics; Step 4-2: Extract image features and train wind shear recognition convolutional neural network node parameters; Step 4-3: Construct a plane scanning scheme for regional refined three-dimensional wind field data; Step 4-4: Output the location and shear intensity of wind shear; Step 4-5: Mark the shear position and shear intensity in the 3D wind field data map.

Images