A method for constructing a digital twin system of urban low-altitude wind field

Abstract

The application provides a kind of construction method of urban low-altitude wind field digital twin system, including constructing urban basic road network skeleton and regional division, generating building block model with real texture using oblique photography data, forming urban three-dimensional space geometric model library;Through radar real-time capture atmospheric information to generate high-resolution three-dimensional wind field scanning data covering the target area;Processing the detection data of laser wind radar, combining the wind field simulation calculation of urban three-dimensional space geometric model, generating urban low-altitude area sub-meter grid dynamic wind field data;Using visualization rendering technology to perform three-dimensional restoration and realistic performance on the obtained dynamic wind field data, forming an interactive urban low-altitude wind field digital twin system.The application integrates multi-scale modeling, laser wind radar, wind field simulation and visualization rendering technology, realizes real-time monitoring, dynamic simulation and intuitive display of urban low-altitude three-dimensional wind field, and improves low-altitude flight safety and operation efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing and simulation system technology, specifically to a method for constructing a digital twin system for urban low-altitude wind fields, used for real-time monitoring, analysis and visualization of urban low-altitude three-dimensional wind field information, thereby improving low-altitude flight safety and efficiency. Background Technology

[0002] The wind field over cities interacts in complex ways with urban buildings, easily generating small-scale, short-duration, and high-intensity wind field changes. For example, building obstruction can cause sudden changes in wind direction, the "canyon effect" between tall buildings can increase wind speed and generate turbulence, and the thermal effects of building surfaces can also cause local airflow changes. These phenomena can significantly impact the stability, maneuverability, energy consumption, and mission efficiency of low-altitude aircraft. However, current low-altitude ground meteorological monitoring facilities are inadequate and cannot monitor high-altitude wind field information, while aviation meteorological support primarily targets civil aviation, failing to meet the needs of low-altitude flights. Therefore, there is an urgent need to develop a three-dimensional digital twin system capable of real-time, high-precision reconstruction of urban low-altitude wind fields to support the large-scale development of the low-altitude economy. Summary of the Invention

[0003] To address the various shortcomings of existing technologies, this invention provides a method for constructing a digital twin system for urban low-altitude wind fields. By integrating multi-scale modeling, laser wind radar, wind field simulation, and visualization rendering technologies, it enables real-time monitoring, dynamic simulation, and intuitive display of urban low-altitude three-dimensional wind fields, thereby improving low-altitude flight safety and operational efficiency.

[0004] The present invention achieves the above objectives through the following technical solutions:

[0005] A method for constructing a digital twin system for urban low-altitude wind fields includes:

[0006] By integrating GIS vector data to construct the basic urban road network framework and regional division, using oblique photogrammetry data to generate building block models with realistic textures, and combining laser point cloud data to supplement terrain elevation information and complex building structural details, a multi-scale integrated geometric modeling from macro-urban layout to micro-building three-dimensional structure is completed, forming a three-dimensional spatial geometric model library for the city.

[0007] Deploy laser wind radar at key locations in low-altitude take-off and landing fields and pre-set flight routes to capture atmospheric information in real time and generate high-resolution three-dimensional wind field scanning data covering the target area.

[0008] Based on high-precision inversion algorithms and numerical simulation technology, the detection data of laser wind radar is processed, and the wind field is simulated and calculated in combination with the constructed three-dimensional spatial geometric model of the city. This generates sub-meter-level gridded dynamic wind field data in the urban low-altitude area, so as to realize the real-time updating and dynamic prediction of wind field parameters.

[0009] The system uses visualization rendering technology to recreate and realistically represent the obtained dynamic wind field data in three dimensions. Through a multi-scale visualization interface, it displays the macroscopic distribution characteristics and microscopic turbulence details of the urban low-altitude wind field, and overlays dangerous area markers and real-time early warning information to form an interactive digital twin system of urban low-altitude wind field.

[0010] According to a method for constructing a digital twin system for urban low-altitude wind fields provided by the present invention, the construction of the building block model includes:

[0011] Import city-level GIS vector data, extract road centerlines, intersection nodes and administrative boundary information, and construct the basic urban road network skeleton through topology analysis;

[0012] The road network skeleton is layered based on road grade attributes, and functional areas are divided in combination with administrative division data to generate a multi-level regional division model.

[0013] Spatial interpolation algorithms are used to correct geometric deviations in GIS vector data;

[0014] Using a tilting camera, multi-angle images of the target area are captured to obtain complete texture information of the building facade and roof.

[0015] Three-dimensional point cloud data is generated by image dense matching technology, and the geometric structure of the building surface is reconstructed by multi-view stereo vision algorithm.

[0016] Automatically identify building outlines based on semantic segmentation models and generate building block models with realistic material textures;

[0017] The generated road network skeleton is spatially registered with the building block model, and the coordinate deviation between the two is corrected by the ICP algorithm.

[0018] According to a method for constructing a digital twin system for urban low-altitude wind fields provided by the present invention, the formation of the urban three-dimensional spatial geometric model library includes:

[0019] An airborne LiDAR was used to perform a full-coverage scan of the target area, acquiring three-dimensional laser point cloud data including terrain surfaces, building facades, and roof structures; ground-based fixed LiDAR was used to supplement the measurements of obscured areas.

[0020] The ICP algorithm is used to spatially register the 3D laser point cloud data with the building block model and correct the coordinate deviation between the two.

[0021] The registered 3D laser point cloud data is divided into blocks, and the data blocks are divided according to individual buildings or terrain units.

[0022] The random sampling consistency algorithm is used to separate ground points from non-ground points in the registered 3D laser point cloud data, and the terrain point extraction effect is optimized by combining the cloth simulation filtering (CSF) algorithm.

[0023] The separated ground points are interpolated into a regular grid digital elevation model (DEM);

[0024] By combining contour lines and water system information from the generated GIS vector data, the DEM is locally corrected to eliminate elevation errors in LiDAR scanning blind spots and output a seamlessly stitched terrain DEM.

[0025] Based on Euclidean distance clustering and region growing algorithms, we can segment non-ground point clouds into individual building units and identify independent building bodies.

[0026] Each building is further segmented into roof, facade and ancillary structures to obtain non-ground point cloud data;

[0027] Perform geometric detail corrections, project non-ground point cloud data onto the surface of the generated building block model, and calculate the distance deviation between the non-ground point cloud data and the surface of the building block model;

[0028] The generated DEM is spatially overlaid with the corrected building block model, and Boolean operations are used to process the intersection area between the building base and the terrain to generate a seamless terrain-building fusion model.

[0029] The macro-scale model is optimized using quadtree segmentation, while the micro-scale model retains all element details, forming a library of urban three-dimensional spatial geometric models.

[0030] According to the present invention, a method for constructing a digital twin system for urban low-altitude wind fields, wherein generating high-resolution three-dimensional wind field scanning data covering a target area by real-time radar capture of atmospheric information includes:

[0031] Based on the urban three-dimensional spatial geometric model library, the geometric center of the low-altitude take-off and landing field and the preset flight path is identified;

[0032] Based on the urban canopy model, computational fluid dynamics (CFD) simulation was used to preliminarily determine the wind-sensitive areas;

[0033] In wind-sensitive areas, laser wind radars are deployed in a grid layout. By emitting laser pulses and receiving atmospheric backscattered signals, the Doppler frequency shift is calculated, and the radial wind speed is retrieved.

[0034] By combining radar azimuth and elevation information, the radial wind speed is converted into wind speed components (U, V, W) in a three-dimensional Cartesian coordinate system, and the composite wind speed is calculated. V = U 2+ V 2+ W 2 and wind direction θ =arctan( V / U );

[0035] The turbulence intensity TI is calculated using the structure function method, and the formula is as follows:

[0036]

[0037] in, The standard deviation of wind speed over 10 minutes. This represents the average wind speed.

[0038] According to the present invention, a method for constructing a digital twin system for urban low-altitude wind fields includes the reconstruction of high-resolution three-dimensional wind field scanning data, which includes:

[0039] Establish a multi-radar data alignment model based on time synchronization and spatial alignment;

[0040] A four-dimensional variational assimilation algorithm is used to fuse observation data from multiple lidar wind measuring radars with the background field of a numerical weather prediction model. The wind field reconstruction result is optimized by minimizing the cost function, as follows:

[0041]

[0042] in, J Let the cost function be minimized. J Find the wind field reconstruction result that best matches the background field of the numerical weather prediction model with radar observation data. N The total number of radar observations. These are radar observations. For the observation operator, For the background scene j The standard deviation of the background error for each grid point The first one to be optimized during the optimization process j Wind field state variables at each grid point For the background scene j Wind field state variables at each grid point The standard deviation of the observation error. M This represents the total number of grid points in the background field.

[0043] The fused wind field data is mapped to the generated urban 3D spatial geometric model library to generate a wind field vector field with terrain-building constraints;

[0044] A dynamic 3D wind field visualization model is generated using volume rendering technology. It supports layered display according to wind speed level and overlays a turbulence intensity heat map. The color gradient represents the TI value.

[0045] According to the method for constructing a digital twin system for urban low-altitude wind fields provided by the present invention, when generating sub-meter-level gridded dynamic wind field data for urban low-altitude areas, the following steps are performed: Turbulent atmospheric initial wind field dynamic modeling.

[0046] Based on the urban three-dimensional spatial geometric model library, and combined with the atmospheric boundary layer inflow conditions captured in real time by laser wind radar, an initial wind field model including buoyancy effect and Coriolis effect is constructed.

[0047] Gaussian filtering or box filtering is used to separate the initial wind field into scales, decomposing the turbulence into solvable scales and subgrid scales to obtain turbulence integral scale data. The filtering operation is defined as follows:

[0048]

[0049] In the formula, It is the filtered physical quantity field. It is the physical field before filtering, G It is a filter function. It is the filtering scale;

[0050] Implicit filtering is used for the near-wall region, which is automatically implemented through OpenFOAM mesh discretization to avoid explicit integration calculations.

[0051] According to the present invention, a method for constructing a digital twin system for urban low-altitude wind fields includes dynamically loading boundary conditions, comprising:

[0052] Inlet boundary: Load wind speed and wind direction profile data monitored in real time by laser wind radar, and apply random turbulent disturbance;

[0053] The exit boundary adopts a zero-gradient condition, allowing the wind field to diffuse naturally;

[0054] Wall boundary: The building surface is set as a no-slip wall, and a logarithmic law-based wall function is used to handle near-wall flow, with the following formula:

[0055]

[0056] in, It is a dimensionless velocity. It is the von Kármán constant. It is the wall roughness constant. It is a dimensionless wall distance. It is the friction speed. It is the vertical distance to the wall. It is kinematic viscosity. It is the wall normal velocity gradient.

[0057] According to the present invention, a method for constructing a digital twin system of urban low-altitude wind fields includes performing a sub-grid-scale turbulence model closure and optimization step when generating sub-grid-scale dynamic wind field data of urban low-altitude areas.

[0058] The dynamic Smagorinsky model is employed to dynamically adjust the subgrid eddy viscosity coefficient through local grid turbulent energy calculations. The eddy viscosity assumption equation for the subgrid stress model is:

[0059]

[0060] in, These are components of the sublattice stress tensor. It is a strain rate tensor with solvable scales. It is a trace of sublattice stress. It is the subgrid eddy viscosity coefficient. δ ij It is the Kronecker function;

[0061] The definition of subgrid eddy viscosity is:

[0062]

[0063] in, It is the Smagorinsky constant. For filtering scale, The modulus of the solvable scale strain rate tensor;

[0064] Dynamic correction is achieved by combining real-time turbulence integral scale data monitored by laser wind radar. C s value.

[0065] According to the present invention, a method for constructing a digital twin system for urban low-altitude wind fields includes performing a vegetation resistance source term parameterization step when generating sub-meter-level gridded dynamic wind field data for urban low-altitude areas.

[0066] Adding a momentum source term to the vegetated area to simulate the obstruction of airflow by the tree canopy, the formula is:

[0067]

[0068] in, It is a momentum source term caused by vegetation. It is the drag coefficient. It is leaf area density. It is a velocity vector. It refers to air density.

[0069] According to a method for constructing a digital twin system of urban low-altitude wind fields provided by the present invention, a volumetric force term caused by terrain is added to simulate the vertical airflow acceleration caused by hillsides / buildings, expressed as:

[0070]

[0071] in, It is a volume force caused by the terrain. It is air density. It is gravitational acceleration. It is the topographic elevation gradient. It is the inertial acceleration term.

[0072] Therefore, compared with existing technologies, the method for constructing a digital twin system for urban low-altitude wind fields proposed in this invention organically combines multi-scale urban modeling, laser wind radar, wind field simulation calculation, and visualization rendering technology to form a complete digital twin system for urban low-altitude wind fields, which has the following beneficial effects:

[0073] 1. This invention employs multi-scale urban modeling technology to achieve refined modeling from macro-level urban layout to micro-level building three-dimensional structures. Combined with the high-precision real-time detection capabilities of laser wind radar, it can accurately acquire wind field data within a set horizontal and vertical range. Through wind field simulation calculation technology, it generates sub-meter-level gridded high-precision wind field data. Compared with traditional monitoring methods, the spatial resolution is significantly improved, enabling precise capture of subtle changes in urban low-altitude wind fields and providing a reliable data foundation for low-altitude economic activities.

[0074] 2. This invention utilizes a laser wind-measuring radar to capture atmospheric information in real time, providing real-time, high-resolution three-dimensional wind field scanning data. Combined with real-time wind field parameter inversion and processing technology, it can rapidly process radar detection data, perform simulation calculations using urban spatial geometric models, and promptly generate the latest wind field data. This ensures that low-altitude economic activities can make decisions based on the latest wind field information, effectively responding to rapid changes in the wind field.

[0075] 3. This invention uses visualization rendering technology to present the three-dimensional urban spatial structure and wind field information in an intuitive way. Relevant departments and managers do not need professional wind field analysis knowledge; they can clearly see the macroscopic distribution and microscopic details of the urban low-altitude wind field through the front-end visualization platform. For example, they can intuitively observe areas of sudden wind direction changes caused by building obstruction, areas of increased wind speed and turbulence caused by the "canyon effect" between high-rise buildings, and local airflow changes caused by the thermal effects of building surfaces, greatly improving the ability to perceive the urban low-altitude wind field. With the help of the visualization platform, relevant departments and managers can monitor the low-altitude wind field in real time. The system can quickly display dangerous areas, such as strong wind areas and highly turbulent areas, providing timely guidance and support for low-altitude flight path planning, decision-making, and scheduling. For example, at low-altitude take-off and landing sites, managers can rationally arrange the take-off and landing times and sequences of aircraft based on real-time wind field information, avoiding operations under dangerous wind field conditions and improving operational safety and efficiency.

[0076] 4. This invention can be widely used in multiple scenarios, including low-altitude take-off and landing fields, low-altitude test fields, low-altitude flight paths, and large-scale low-altitude airspace. At low-altitude take-off and landing fields, it can provide accurate wind field information for aircraft take-off and landing, ensuring safety. At low-altitude test fields, it can provide realistic wind field environment simulations for aircraft performance testing, improving the accuracy and reliability of the tests. At low-altitude flight paths, it can monitor wind field changes along the flight path in real time, providing a basis for aircraft flight path planning and adjustments. In large-scale low-altitude airspace, it can comprehensively grasp the wind field conditions within the airspace, providing support for airspace management and scheduling.

[0077] 5. By providing accurate and real-time wind field information and intuitive visual monitoring, this invention effectively improves the safety level and operational efficiency of low-altitude flight. Aircraft can adjust their flight attitude and speed based on real-time wind field information, avoiding flight accidents caused by sudden wind field changes and reducing flight risks. Simultaneously, reasonable flight path planning and decision-making scheduling can reduce flight time, improve flight efficiency, and lower operating costs, thereby promoting the accelerated development of the low-altitude economy.

[0078] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0079] Figure 1 This is a flowchart of an embodiment of a method for constructing a digital twin system for urban low-altitude wind fields according to the present invention.

[0080] Figure 2 This is a flowchart illustrating the construction process of a building block model in an embodiment of a construction method for a digital twin system of urban low-altitude wind fields according to the present invention.

[0081] Figure 3This is a flowchart illustrating the formation of a three-dimensional spatial geometric model library for a city, as described in an embodiment of the construction method for a digital twin system of urban low-altitude wind fields according to the present invention.

[0082] Figure 4 This is a flowchart illustrating the generation of high-resolution three-dimensional wind field scanning data in an embodiment of the construction method of a digital twin system for urban low-altitude wind fields according to the present invention.

[0083] Figure 5 This is a flowchart illustrating the reconstruction of high-resolution three-dimensional wind field scanning data in an embodiment of the construction method of a digital twin system for urban low-altitude wind fields according to the present invention.

[0084] Figure 6 This is a flowchart illustrating the dynamic loading of boundary conditions in an embodiment of the construction method of a digital twin system for urban low-altitude wind fields according to the present invention. Detailed Implementation

[0085] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0086] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0087] See Figures 1 to 6 This embodiment provides a method for constructing a digital twin system for urban low-altitude wind fields, including:

[0088] Step S1: By integrating GIS vector data, the basic urban road network framework and regional division are constructed. Oblique photogrammetry data is used to generate building block models with realistic textures. Laser point cloud data is combined to supplement terrain elevation information and complex building structural details, thus completing multi-scale integrated geometric modeling from macro-urban layout to micro-building three-dimensional structure, forming a three-dimensional spatial geometric model library for the city.

[0089] Step S2: Deploy laser wind-measuring radar at key locations in the low-altitude take-off and landing field and the preset flight path, and generate high-resolution three-dimensional wind field scanning data covering the target area by capturing atmospheric information (atmospheric turbulence, wind speed and wind direction information) in real time through radar.

[0090] Step S3: Based on high-precision inversion algorithm and numerical simulation technology, the detection data of laser wind radar is processed, and the wind field is simulated and calculated in combination with the constructed three-dimensional spatial geometric model of the city to generate sub-meter level gridded dynamic wind field data in the urban low-altitude area, so as to realize real-time updating and dynamic prediction of wind field parameters.

[0091] Step S4: The obtained dynamic wind field data is rendered in three dimensions and realistically represented using visualization rendering technology. The macroscopic distribution characteristics and microscopic turbulence details of the urban low-altitude wind field are displayed through a multi-scale visualization interface. Danger zone markers and real-time early warning information are superimposed to form an interactive digital twin system of urban low-altitude wind field.

[0092] In this embodiment, as Figure 2 As shown, the construction of the building block model includes:

[0093] Import city-level GIS vector data, extract road centerlines, intersection nodes and administrative boundary information, and construct the basic urban road network skeleton through topology analysis;

[0094] The road network skeleton is layered based on road grade attributes (such as arterial roads, secondary arterial roads, and local roads), and functional areas (such as commercial areas, residential areas, and industrial areas) are divided in combination with administrative division data to generate a multi-level regional division model.

[0095] Spatial interpolation algorithms are used to correct geometric deviations in GIS vector data, ensuring that the elevation matching accuracy between the road network framework and the actual terrain is ≤0.5 meters.

[0096] In this embodiment, a five-lens tilting camera is used to acquire multi-angle (forward, rear, left, right, and downward) images of the target area within a flight altitude range of 200-500 meters to obtain complete texture information of the building facade and roof.

[0097] Three-dimensional point cloud data with a point cloud density of ≥50 points / square meter is generated by image dense matching technology, and the geometric structure of the building surface is reconstructed by combining the multi-view stereo vision (MVS) algorithm.

[0098] The semantic segmentation model automatically identifies building outlines, extracts detailed features such as windows and balconies, and generates building block models with realistic material textures (such as glass, brick walls, and concrete).

[0099] The generated road network skeleton is spatially registered with the building block model, and the coordinate deviation between the two is corrected by the ICP (Iterative Closest Point) algorithm to ensure that the overlap between the building base and the road boundary is ≥95%.

[0100] Local refinement processing is performed on overlapping areas (such as where buildings are adjacent to roads), and NURBS surface fitting technology is used to optimize the transition details between building edges and terrain;

[0101] The output includes a unified spatial database containing road networks, regional divisions, and building models, and supports export in multiple formats such as SHP, OBJ, and FBX.

[0102] In this embodiment, as Figure 3 As shown, the formation of the urban three-dimensional spatial geometric model library includes:

[0103] Airborne LiDAR (Light Detection and Ranging) was used to perform a full-coverage scan of the target area, acquiring high-density three-dimensional laser point cloud data including terrain surfaces, building facades, and roof structures. Ground-based fixed LiDAR was used to supplement the measurements in obscured areas (such as building shadows or dense vegetation) to ensure data integrity.

[0104] Based on the generated photographic architectural model, features such as building corners and roof ridge lines are extracted as control points;

[0105] The laser point cloud and the building block model are spatially registered using the ICP (Iterative Closest Point) algorithm to correct the coordinate deviation between the two and ensure that the overlap between the building base and the building block model is ≥98%.

[0106] The registered point cloud is divided into blocks, with data blocks divided according to individual buildings or terrain units, to facilitate subsequent fine-grained processing.

[0107] In this embodiment, terrain elevation information extraction and DEM generation include:

[0108] The Random Sample Consensus (RANSAC) algorithm is used to separate ground points from non-ground points in the point cloud, and the terrain point extraction effect is optimized by combining the cloth simulation filtering (CSF) algorithm.

[0109] For complex terrain (such as mountains and hills), slope analysis is performed, and when the slope is ≥20°, the ground point sampling density is automatically increased to 300 points / square meter.

[0110] The separated ground points are interpolated into a regular grid digital elevation model (DEM), and a high-precision DEM with a resolution of ≤0.5 meters is generated using the natural neighborhood interpolation method.

[0111] By combining contour lines and water system information from the generated GIS vector data, the DEM is locally corrected to eliminate elevation errors in LiDAR scanning blind spots (such as water areas and under bridges), and a seamlessly stitched terrain DEM is output, supporting seamless integration with building models.

[0112] Based on Euclidean distance clustering and region growing algorithms, we can segment non-ground point clouds into individual building units and identify independent building bodies.

[0113] For each building, the roof, facade, and ancillary structures (such as balconies and canopies) are further segmented to obtain non-ground point cloud data. Principal component analysis (PCA) is then used to determine the building's main orientation and assist in structural classification.

[0114] Perform geometric detail corrections, project non-ground point cloud data onto the surface of the generated building block model, and calculate the distance deviation between the non-ground point cloud data and the surface of the building block model;

[0115] For areas with a deviation of ≥0.3 meters (such as building deformation or missing parts in the building block model), an implicit surface reconstruction algorithm is used to generate a NURBS surface model to supplement details such as curtain wall segmentation and decorative lines; feature lines are extracted for roof structures (such as pitched roofs and domes), and Hough transform is used to detect roof edges and intersections to optimize the roof geometry.

[0116] The generated DEM is spatially overlaid with the corrected building block model, and Boolean operations are used to process the intersection area between the building base and the terrain to generate a seamless terrain-building fusion model.

[0117] Quadtree segmentation optimization is used for macro-scale (city-level) models, while full element details are preserved for micro-scale (building-level) models, forming a city 3D spatial geometric model library, such as a multi-level model library from LOD1 (regional level) to LOD4 (component level).

[0118] The model library is stored hierarchically by region, scale, and feature type, supporting dynamic retrieval and real-time updates. For example:

[0119] Macro level: roads, water systems, administrative divisions;

[0120] Mesoscopic level: building complexes, green spaces, public facilities;

[0121] Microscopic layer: building facade, roof structure, and auxiliary components.

[0122] In this embodiment, as Figure 4 As shown, high-resolution three-dimensional wind field scanning data covering the target area is generated by real-time radar capture of atmospheric information, including:

[0123] Based on the urban 3D spatial geometric model library, the geometric center of low-altitude take-off and landing sites (such as helicopter landing pads and drone take-off and landing points) and preset routes (such as logistics delivery routes and emergency rescue channels) is identified.

[0124] By combining the urban canopy model (including building height and density distribution), computational fluid dynamics (CFD) simulation is used to preliminarily determine wind-sensitive areas (such as the high-rise canyon effect area and the heat island circulation area).

[0125] In wind-sensitive areas, laser wind-measuring radars are deployed in a grid-like layout, with an overlap rate of ≥30% between adjacent radar detection ranges to ensure no monitoring blind spots. This embodiment uses a Doppler laser wind-measuring radar with a working wavelength of 1.5μm (eye-safe band), a pulse repetition frequency ≥10kHz, and a wind measurement range covering low altitudes from 0 to 3000 meters. The radar scanning mode is set to a three-dimensional cone scan + vertical plane scan composite mode, with a cone scan angle range of ±15°, a vertical resolution ≤10 meters, and a horizontal resolution ≤50 meters. The radar data output frequency is configured to ≥1Hz, supporting real-time transmission of wind speed, wind direction, and turbulence intensity (TI) data.

[0126] Laser wind radar calculates the Doppler frequency shift and inverts the radial wind speed by emitting laser pulses and receiving atmospheric backscattered signals.

[0127] By combining radar azimuth and elevation information, the radial wind speed is converted into wind speed components (U, V, W) in a three-dimensional Cartesian coordinate system, and the composite wind speed is calculated. V = U 2+ V 2+ W 2 and wind direction θ =arctan( V / U );

[0128] The turbulence intensity TI is calculated using the structure function method, and the formula is as follows:

[0129]

[0130] in, The standard deviation of wind speed over 10 minutes. The average wind speed; a single radar scan cycle of ≤5 seconds, covering a low-altitude area with a radius of ≥5 kilometers.

[0131] Furthermore, the Kalman filter algorithm is used to remove high-frequency noise from the original data while preserving the trend changes in the wind field;

[0132] Furthermore, outliers are eliminated through cross-validation: when the wind speed at a certain point deviates from the average wind speed of the eight surrounding points by more than 30%, it is marked as an anomaly and corrected by inverse distance weighted interpolation.

[0133] Furthermore, a Long Short-Term Memory (LSTM) network model was used to perform spatiotemporal prediction and completion for missing data areas (such as radar maintenance periods). The model training data consisted of historical wind field data from the past 72 hours.

[0134] In this embodiment, as Figure 5 As shown, the reconstruction of high-resolution three-dimensional wind field scanning data includes:

[0135] Establish a multi-radar data alignment model based on time synchronization and spatial alignment to unify the acquisition time of each radar to millisecond-level accuracy;

[0136] A four-dimensional variational assimilation (4D-Var) algorithm is used to fuse multiple radar observation data with the background field of a numerical weather prediction (WRF) model. The wind field reconstruction result is optimized by minimizing the cost function, as follows:

[0137]

[0138] in, J Let the cost function be minimized. J It is possible to find the wind field reconstruction result that best matches the background field of the numerical weather prediction model with the radar observation data. N The total number of radar observations. These are radar observations. For the observation operator, For the background scene j The standard deviation of the background error for each grid point The first one to be optimized during the optimization process j Wind field state variables at each grid point For the background scene j Wind field state variables at each grid point The standard deviation of the observation error. M The total number of background field grid points; after fusion, the wind field data resolution is improved to 15 meters horizontally and 5 meters vertically, covering a target area of ​​≥100 square kilometers, and the wind speed error in the building canyon area is ≤0.5m / s.

[0139] The fused wind field data is mapped to the generated urban 3D spatial geometric model library to generate a wind field vector field with terrain-building constraints;

[0140] The system uses volume rendering technology to generate a dynamic 3D wind field visualization model, which supports layered display according to wind speed level (such as 0-3m / s, 3-6m / s) and overlays turbulence intensity heat map. The color gradient represents the TI value.

[0141] In this embodiment, a dynamic update and real-time verification mechanism for wind field data is implemented:

[0142] Establish a two-way transmission channel between radar data and the digital twin system, synchronizing the latest wind field data to the system model every 10 seconds;

[0143] A three-level early warning mechanism is triggered for sudden wind field events (such as gusts and shear lines):

[0144] Level 1 warning (wind speed change rate ≥ 1.5 m / s / s): High-risk areas are marked and pushed to low-altitude aircraft terminals;

[0145] Level 2 Warning (TI≥20%): Restrict takeoff and landing operations and initiate alternative route planning;

[0146] Level 3 warning (wind speed ≥ 15m / s): Forcefully terminate low-altitude flight mission and trigger emergency landing procedure.

[0147] Deploy micro-drones (such as quadcopter drones) as mobile calibration platforms, carrying lightweight anemometers (such as pitot tubes) to fly along preset routes and collect measured wind speed data;

[0148] The measured data were compared with the radar wind field model. When the deviation exceeded 15%, the model adaptive correction was initiated, and the background field weight parameters in the variational assimilation algorithm were adjusted.

[0149] A wind field data quality report is generated monthly, including error distribution statistics, radar health status assessment, and optimization suggestions.

[0150] In this embodiment, when generating sub-meter-level gridded dynamic wind field data for urban low-altitude areas, the following dynamic modeling steps for the initial wind field of turbulent atmosphere are performed:

[0151] Based on the urban three-dimensional spatial geometric model library, and combined with the atmospheric boundary layer inflow conditions captured in real time by laser wind radar, including wind speed, wind direction, and turbulence intensity, an initial wind field model including buoyancy effect (thermal convection) and Coriolis effect (Earth rotation) is constructed.

[0152] Gaussian filtering or box filtering is used to separate the initial wind field into scales, decomposing the turbulence into a solvable scale (analyzable mesh) and a subgrid scale (SGS, requiring a closed model), obtaining turbulence integral scale data. The filtering operation is defined as follows:

[0153]

[0154] In the formula, It is the filtered physical quantity field. It is the filtered physical quantity field. It is a filtering function (such as Gaussian filtering, box filtering). It is the filtering scale (usually the grid size).

[0155] Implicit filtering is used for the near-wall area (building surface, terrain slope), which is automatically implemented by discretization through OpenFOAM mesh, avoiding explicit integration calculation and improving computational efficiency.

[0156] In this embodiment, as Figure 6As shown, the boundary conditions are dynamically loaded, including:

[0157] Inlet boundary: Load wind speed and wind direction profile data monitored in real time by laser wind radar, update it every 10 seconds, and apply random turbulence disturbance (turbulence intensity error ≤8%).

[0158] The exit boundary adopts a zero-gradient condition, allowing the wind field to diffuse naturally;

[0159] Wall boundary: The building surface is set as a no-slip wall, and a logarithmic law-based wall function is used to handle near-wall flow, with the following formula:

[0160]

[0161] in, It is a dimensionless velocity. It is the von Kármán constant ( ), It is the wall roughness constant. It is a dimensionless wall distance. It is the friction speed. It is the vertical distance to the wall. It is kinematic viscosity. It is the wall normal velocity gradient.

[0162] The continuity equation (mass conservation) is expressed as:

[0163]

[0164] in, It is a divergence operator. It is a velocity vector field.

[0165] The momentum equation (the filtered Navier-Stokes equation) is expressed as:

[0166]

[0167] in, It is a local acceleration term. It is a momentum transport term. It is the pressure gradient term. It is the viscous diffusion term. It is the sublattice stress divergence term. These are external volume forces (such as vegetation resistance, topographic effects, etc.).

[0168] In this embodiment, when generating sub-meter-scale gridded dynamic wind field data for urban low-altitude areas, a sub-grid-scale turbulence model closure and optimization step is performed:

[0169] The dynamic Smagorinsky model is employed to dynamically adjust the subgrid eddy viscosity coefficient through local grid turbulent energy calculations. The eddy viscosity assumption equation for the subgrid stress model is as follows:

[0170]

[0171] in, These are components of the sublattice stress tensor. It is a strain rate tensor with solvable scales. It is a trace of sublattice stress. It is the subgrid eddy viscosity coefficient. δ ij It is the Kronecker function.

[0172] The definition of subgrid eddy viscosity is:

[0173]

[0174] in, This is the Smagorinsky constant, typically taken as 0.1-0.2. For filtering scale, The modulus of the solvable scale strain rate tensor.

[0175] Dynamic correction is achieved by combining real-time turbulence integral scale data monitored by laser wind radar. C s The value is set such that the deviation between the model-predicted turbulent energy dissipation rate and the measured value is ≤12%.

[0176] In this embodiment, when generating sub-meter-level gridded dynamic wind field data for urban low-altitude areas, a vegetation resistance source term parameterization step is performed:

[0177] Adding a momentum source term to the vegetated area to simulate the obstruction of airflow by the tree canopy, the formula is:

[0178]

[0179] in, It is a momentum source term caused by vegetation. It is the drag coefficient. It is leaf area density. It is a velocity vector. It refers to air density.

[0180] Dynamically adjust based on vegetation type (trees / shrubs) and season (leaf area index LAI). C d and A p The parameter library covers common urban vegetation (such as camphor trees, plane trees, and hedges).

[0181] In this embodiment, a body-fitted mesh is generated using a digital elevation model (DEM) or laser point cloud data. The near-wall mesh requires a dimensionless wall distance. Local densification is carried out in complex areas (such as ridges and building wakes). , (Height of building features).

[0182] To simulate the acceleration of vertical airflow caused by hillsides / buildings, a volumetric force term induced by the terrain is added, expressed as:

[0183]

[0184] In the formula, It is a volume force caused by the terrain. It is air density. It is gravitational acceleration. It is the topographic elevation gradient. It is the inertial acceleration term; the volume force caused by the terrain is activated when the slope is greater than 15° to avoid digital noise in plain areas.

[0185] Specifically, this embodiment uses the SnappyHexMesh tool to generate a hexahedral-dominated body-fitting mesh based on digital elevation model (DEM) or laser point cloud data, with a total number of mesh cells ≥ 5 million, ensuring that the mesh orthogonality is ≥ 0.3 in complex areas (building wake zones, ridges);

[0186] Dimensionless distance of near-wall mesh y + Controlled between 30-300, building feature height H The number of grid layers within the range is ≥10, which satisfies the applicable conditions for wall functions.

[0187] Adaptive mesh refinement is implemented in wind-sensitive areas (such as within a 200-meter radius of helicopter landing points and in the canyon effect area of ​​high-rise buildings), with a refinement factor of 2-4 times, thereby improving the local mesh resolution to 0.5-1 meters.

[0188] The encryption determination is based on the wind speed gradient ∂ V / ∂ x and turbulence intensity TI When ∂ V / ∂ x ≥0.5s⁻¹ or TI Encryption is triggered when the concentration is ≥18%.

[0189] The governing equations (continuity equation and momentum equation) are discretized using the finite volume method, and all variables (velocity, pressure, and subgrid turbulence variables) are stored at the center of the grid cells.

[0190] The interpolation from the cell center to the surface uses a linear format (second-order precision), and the time stepping uses Issa's PISO algorithm (implicit prediction-correction scheme), with ≤3 iterations per time step.

[0191] The convection term is stabilized using a second-order upwind scheme, the diffusion term uses a central difference scheme, and the time integral uses a second-order backward difference scheme.

[0192] The computational domain is partitioned using Scotch decomposition, and multi-process parallel computing is achieved by combining the MPI communication protocol, with a process count ≤64 and a parallel efficiency ≥85%.

[0193] Vegetation drag and terrain lift terms are explicitly processed to reduce iterative computation.

[0194] Output sub-meter level gridded wind field data (horizontal resolution ≤ 1 meter, vertical resolution ≤ 2 meters), covering a target area of ​​≥ 50 square kilometers, and with wind speed error ≤ 0.3 m / s in the building canyon area.

[0195] Therefore, the governing equations solved in this embodiment employ the finite volume method, with all variables (velocity, pressure, subgrid turbulence variables, etc.) stored at the center of the grid cells. Interpolation from the cell center to the surface uses a linear scheme (equivalent to second-order accuracy). Time propagation uses Issa's PISO algorithm, an implicit prediction-correction scheme. In the momentum equation, vegetation drag and topographic lift terms are explicitly handled to reduce iterative computation, convection terms use a second-order upwind scheme to enhance stability, and diffusion terms use a central difference scheme. Time integration uses a second-order backward difference scheme. The computational domain is partitioned using Scotch decomposition, and efficient parallel computation is achieved using the MPI communication protocol.

[0196] In summary, this invention employs multi-scale urban modeling technology to achieve refined modeling from macro-level urban layout to micro-level building three-dimensional structures. Combined with the high-precision real-time detection capabilities of laser wind radar, it can accurately acquire wind field data within a set horizontal and vertical range. Through wind field simulation calculation technology, it generates sub-meter-level gridded high-precision wind field data. Compared to traditional monitoring methods, the spatial resolution is significantly improved, accurately capturing subtle changes in urban low-altitude wind fields and providing a reliable data foundation for low-altitude economic activities.

[0197] Furthermore, this invention utilizes a laser wind-measuring radar to capture atmospheric information in real time, providing real-time, high-resolution three-dimensional wind field scanning data. Combined with real-time wind field parameter inversion and processing technology, it can rapidly process radar detection data, perform simulation calculations using urban spatial geometric models, and promptly generate the latest wind field data. This ensures that low-altitude economic activities can make decisions based on the latest wind field information, effectively responding to rapid changes in the wind field.

[0198] Furthermore, this invention uses visualization rendering technology to present the three-dimensional urban spatial structure and wind field information in an intuitive way. Relevant departments and managers do not need professional wind field analysis knowledge; they can clearly see the macroscopic distribution and microscopic details of the urban low-altitude wind field through the front-end visualization platform. For example, they can intuitively observe areas of sudden wind direction changes caused by building obstruction, areas of increased wind speed and turbulence caused by the "canyon effect" between high-rise buildings, and local airflow changes caused by the thermal effects of building surfaces, greatly improving the ability to perceive the urban low-altitude wind field. With the help of the visualization platform, relevant departments and managers can monitor the low-altitude wind field in real time. The system can quickly display dangerous areas, such as strong wind areas and highly turbulent areas, providing timely guidance and support for low-altitude flight path planning, decision-making, and scheduling. For example, at low-altitude take-off and landing sites, managers can rationally arrange the take-off and landing times and sequences of aircraft based on real-time wind field information, avoiding operations under dangerous wind field conditions and improving operational safety and efficiency.

[0199] Furthermore, this invention can be widely used in various scenarios, including low-altitude take-off and landing fields, low-altitude test fields, low-altitude flight paths, and large-scale low-altitude airspace. At low-altitude take-off and landing fields, it can provide accurate wind field information for aircraft take-off and landing, ensuring safety. At low-altitude test fields, it can provide realistic wind field environment simulations for aircraft performance testing, improving the accuracy and reliability of the tests. At low-altitude flight paths, it can monitor wind field changes along the flight path in real time, providing a basis for aircraft flight path planning and adjustments. In large-scale low-altitude airspace, it can comprehensively grasp the wind field conditions within the airspace, providing support for airspace management and scheduling.

[0200] Furthermore, by providing accurate and real-time wind field information and intuitive visual monitoring, this invention effectively improves the safety level and operational efficiency of low-altitude flight. Aircraft can adjust their flight attitude and speed based on real-time wind field information, avoiding flight accidents caused by sudden wind field changes and reducing flight risks. Simultaneously, reasonable flight path planning and decision-making scheduling can reduce flight time, improve flight efficiency, and lower operating costs, thereby promoting the accelerated development of the low-altitude economy.

[0201] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0202] The above embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-substantial changes and substitutions made by those skilled in the art based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. A method for constructing a digital twin system of urban low-altitude wind field, characterized in that, The method comprises the following steps: Constructing a city basic road network skeleton and regional division by fusing GIS vector data, generating a building block model with real texture using oblique photography data, and combining laser point cloud data to supplement terrain elevation information and complex building structure details, completing multi-scale integrated geometric modeling from macro city layout to micro building three-dimensional structure, and forming a city three-dimensional space geometric model library; Deploying laser wind measurement radars at key positions on the low-altitude take-off and landing field and the preset flight route, generating high-resolution three-dimensional wind field scanning data covering the target area by real-time capture of atmospheric information by the radars; Processing the detection data of the laser wind measurement radars based on high-precision inversion algorithms and numerical simulation techniques, and simulating and calculating the wind field based on the constructed city three-dimensional space geometric model to generate dynamic wind field data of the city low-altitude area in a sub-meter grid, so as to realize real-time updating and dynamic prediction of the wind field parameters; Using visual rendering technology to perform three-dimensional restoration and realistic representation on the obtained dynamic wind field data, displaying the macro distribution characteristics and micro turbulence details of the city low-altitude wind field through a multi-scale visual interface, and superimposing dangerous area identification and real-time early warning information to form an interactive city low-altitude wind field digital twin system; In the step of generating dynamic wind field data of the city low-altitude area in a sub-meter grid, the following steps are performed: Based on the city three-dimensional space geometric model library and the inflow conditions of the atmospheric boundary layer captured by the laser wind measurement radars in real time, an initial wind field model containing buoyancy effect and Coriolis force effect is constructed; Scale separation is performed on the initial wind field by filtering operation to decompose the turbulence into a resolvable scale and a sub-grid scale to obtain turbulence integral scale data, wherein the filtering operation is defined as: wherein is the filtered physical quantity field, is the unfiltered physical quantity field, G is the filter function, is the filter scale; For the near-wall region, implicit filtering is used, which is automatically realized by OpenFOAM grid discretization to avoid explicit integral calculation.

2. The method of claim 1, wherein, The construction of the building block model comprises: Importing city-level GIS vector data, extracting road centerline, intersection node and administrative boundary information, and constructing a city basic road network skeleton through topological analysis; Based on the road grade attribute, the road network skeleton is processed in layers, and a multi-level regional division model is generated by combining administrative division data to divide functional areas; Geometric deviations in the GIS vector data are corrected using a spatial interpolation algorithm; Multi-angle image acquisition of the target area is performed using an oblique photography camera to obtain complete texture information of building facades and roofs; Three-dimensional point cloud data is generated through image dense matching technology, and the building surface geometry is reconstructed using a multi-view stereo vision algorithm; A building block model with real material texture is generated by automatically identifying the building contour based on a semantic segmentation model; The generated road network skeleton and building block model are spatially registered, and the coordinate deviation between them is corrected by ICP algorithm.

3. The method of claim 1, wherein, The formation of the city three-dimensional space geometric model library comprises: Full coverage scanning of the target area is performed using an airborne LiDAR to obtain three-dimensional laser point cloud data containing terrain surface, building facade and roof structure; wherein, ground fixed LiDAR is used for supplementary measurement of the sheltered area; The three-dimensional laser point cloud data is spatially registered with the building block model by the ICP algorithm to correct the coordinate deviation between the two; The registered three-dimensional laser point cloud data is processed in blocks, and the data blocks are divided according to building units or ground element units; The ground points and non-ground points are separated from the registered three-dimensional laser point cloud data by using the random sample consensus algorithm, and the terrain point extraction effect is optimized by combining the cloth simulation filtering CSF algorithm; The separated ground points are interpolated into a regular grid digital elevation model DEM; The DEM is locally corrected in combination with the contour line and water system information in the generated GIS vector data to eliminate the elevation error of the LiDAR scanning blind area, and a seamless spliced terrain DEM is output; Based on the Euclidean distance clustering and region growing algorithm, the non-ground point cloud is segmented into building units to identify independent building bodies; Each building body is further segmented into a roof, a facade and an accessory structure to obtain non-ground point cloud data; Geometric details are corrected, the non-ground point cloud data is projected onto the surface of the generated building block model, and the distance deviation between the non-ground point cloud data and the surface of the building block model is calculated; The generated DEM and the corrected building block model are spatially superimposed, the intersection area of the building base and the terrain is processed by Boolean operation, and a seamless terrain-building fusion model is generated; The macro-scale model is divided and optimized by using a quadtree, and the micro-scale model retains full-factor details, forming a city three-dimensional space geometric model library.

4. The method of claim 1, wherein, The high-resolution three-dimensional wind field scanning data covering the target area is generated by real-time capturing of atmospheric information by radar, including: Based on the city three-dimensional space geometric model library, the geometric center of the low-altitude take-off and landing field and the preset flight route is identified; Combined with the city canopy model, the computational fluid dynamics CFD simulation is used to preliminarily determine the wind field sensitive area; In the wind field sensitive area, the laser wind radar is deployed according to the grid layout, the Doppler shift amount is calculated by emitting a laser pulse and receiving the atmospheric backscatter signal, and the radial wind speed is inverted; Combining radar azimuth and elevation information, the radial wind speed is converted into wind speed components in a three-dimensional rectangular coordinate system. U , V , W ), and calculate the composite wind speed. V ′ and horizontal wind direction The formula for calculating the turbulence intensity TI is: =arctan( V ′ / U ); The reconstruction of high-resolution three-dimensional wind field scanning data includes: wherein is the standard deviation of wind speed over 10 minutes, is the average wind speed.

5. The method of claim 4, wherein, A multi-radar data alignment model based on time synchronization-space alignment is established; The observation data of multiple laser wind radars is fused with the background field of the numerical weather prediction model by using a four-dimensional variational assimilation algorithm, and the wind field reconstruction result is optimized by minimizing the cost function, which is represented as: The fused wind field data is mapped to the generated city three-dimensional space geometric model library to generate a wind field vector field with terrain-building constraints; wherein J is a cost function, by minimizing J finds the wind field reconstruction that best matches the radar observation data to the numerical weather prediction model background field, N is the total number of radar observations, is a radar observation, is the observation operator, is the background error standard deviation for the j grid point in the background field, is the wind field state variable for the j grid point to be solved in the optimization process, is the wind field state variable for the j grid point in the background field, is the observation error standard deviation, M is the total number of background field grid points; A dynamic three-dimensional wind field visualization model is generated by using volume rendering technology, which supports hierarchical display according to wind speed levels and superimposes a turbulence intensity heat map with color gradient representing TI values. The boundary conditions are dynamically loaded, including:

6. The method of claim 1, wherein, Inlet boundary: load the wind speed and wind direction profile data monitored by the laser wind radar in real time, and apply random turbulence disturbance; Outlet boundary: use zero gradient condition to allow natural diffusion of the wind field; Wall boundary: the building surface is set as a no-slip wall, and the near-wall flow is processed by using a wall function based on the logarithmic law, with the formula being: ​ wherein, is the dimensionless velocity, is the von Karman constant, is the wall roughness constant, is the dimensionless wall distance, is the friction velocity, is the perpendicular distance to the wall, is the kinematic viscosity, is the wall normal velocity gradient.

7. The method of claim 1, wherein, In generating the urban low-altitude area sub-millimeter grid dynamic wind field data, the sub-grid scale turbulence model closure and optimization steps are performed: The dynamic Smagorinsky model is adopted to dynamically adjust the sub-grid eddy viscosity coefficient by calculating the local grid turbulent energy The eddy viscosity hypothesis equation of the sub-grid stress model is as follows: wherein is a component of the sub-lattice stress tensor, is a strain rate tensor of the solvable scale, is a trace of the sub-lattice stress, is a sub-lattice eddy viscosity coefficient, The definition of the sub-grid eddy viscosity coefficient is: ij is a Kronecker function; In generating the urban low-altitude area sub-millimeter grid dynamic wind field data, the vegetation resistance source term parameterization step is performed: wherein, is a Smagorinsky constant, is a filter scale, is a modulus of the resolvable scale strain rate tensor; The dynamic correction is combined with the real-time monitored turbulence integral scale data of the laser wind radar C s value.

8. The method of claim 1, wherein, A momentum source term is added in the vegetation area to simulate the resistance effect of the tree crown on the airflow, and the formula is:

9. The method according to any one of claims 1 to 8, characterized in that: wherein, is the vegetation-induced momentum source term, is the drag coefficient, is the leaf area density, is the velocity vector, is the air density. In order to simulate the vertical airflow acceleration caused by the mountain slope / building, a terrain-induced body force term is added, which is expressed as: ​ where is the terrain-induced volume force, is the air density, is the gravitational acceleration, is the terrain elevation gradient, is the inertial acceleration term.

Images