Wind turbine shadow assessment method and system based on three-dimensional terrain occlusion determination

Abstract

This invention provides a method and system for assessing wind turbine shadows based on three-dimensional terrain occlusion judgment. This invention relates to the field of wind power engineering planning and environmental impact assessment technology. The assessment method includes: acquiring and locating one or more environmentally sensitive targets based on aerial real-scene perception, generating a set of measured sensitive targets; calculating multiple key acquisition points simulating the rotor space range based on the geometric parameters of the wind turbine; acquiring spatial image data including light source-side images and receiver-side images; analyzing the light source-side images in the spatial image data to determine the potential light and shadow time window for each key acquisition point; analyzing the receiver-side images and projecting the environmentally sensitive targets into the images to determine the visibility of each key acquisition point; if the proposed installation location of the wind turbine currently being assessed has light and shadow effects, then based on the differences in the visibility states of multiple key acquisition points, constructing an occlusion gradient vector to determine the optimal shadow avoidance displacement direction.

Description

Technical Field

[0001] This invention relates to the field of wind power engineering planning and environmental impact assessment technology, and in particular to a precise assessment method for the light and shadow flickering effect of wind turbine generators on surrounding sensitive points, taking into account the shading factors of complex terrain. Background Technology

[0002] With the advancement of the global clean energy transition, the scale of onshore wind farm construction is expanding rapidly, and wind turbine site selection is gradually extending from plains to complex terrain areas such as mountains and hills. Environmental impact assessment is a crucial step in determining the feasibility of wind farms during planning and construction. Among the assessments, the flickering effect of light and shadow generated by wind turbines during operation is an important indicator for evaluating the impact of wind farms on the quality of life of surrounding residents. Flickering refers to the phenomenon where, when sunlight shines from behind the rotating wind turbine blades, the moving blades periodically block sunlight, creating a flickering effect within the shaded area. If residential areas, schools, or other sensitive locations are within this affected area, prolonged flickering may cause visual fatigue, irritability, and even trigger health problems such as photosensitive epilepsy.

[0003] Currently, the industry's prevailing methods for assessing the impact of sunlight and shadow primarily rely on computer numerical simulations. Traditional assessment techniques are typically based on astronomical algorithms and geometric projection principles. They calculate the sun's trajectory throughout the year using the project's latitude and longitude, and combine this with geometric parameters such as the wind turbine's hub height and impeller diameter, as well as the relative positions of sensitive points, to calculate the theoretical range and duration of shadow projection. This process follows the worst-case scenario principle, assuming cloudless skies, continuous wind turbine operation, and blade planes always perpendicular to sunlight during all periods of sunshine.

[0004] However, existing assessment techniques exhibit significant limitations in practical applications, particularly in complex terrain conditions such as mountains and hills. The main technical challenges are as follows: traditional models are generally based on the assumption of flat terrain or use only low-precision digital elevation models. While the errors resulting from this simplification are acceptable in plains areas, in mountainous regions with significant topographic relief, actual physical shading is the key factor determining whether shadows can be cast on sensitive points. For example, between wind turbines located on ridgelines and villages in valleys, there are often secondary ridges, highlands, or dense vegetation. These natural obstacles often block sunlight or obstruct views, making the theoretical shadows shown in geometric calculations physically unable to reach sensitive points. Existing calculation software often struggles to handle this micro-level terrain shading with fine precision, leading to many areas that are not actually affected being misclassified as affected areas. Due to insufficient accuracy in the assessment results, wind power developers are often forced to take excessive avoidance measures. Based on assessment reports containing numerous false positives, designers may make extensive and unnecessary moves to wind turbine locations. This could not only cause turbines to deviate from their optimal wind resource positions, reducing overall power generation, but could even lead to abandoning some buildable areas to avoid non-existent light and shadow effects. Furthermore, during operation, to address potential complaints, wind farms often need to implement overly broad shutdown policies (sector management), resulting in significant power generation losses and reduced economic benefits.

[0005] In summary, existing wind power light and shadow assessment technologies, when faced with complex terrain, lack effective means to identify physical shading in the real geographical environment, leading to significant discrepancies between assessment conclusions and actual conditions. How to accurately determine the line-of-sight shading relationship between wind turbines and sensitive points during the wind farm planning stage, combining real three-dimensional terrain features, and eliminating invalid theoretical shadow times, thereby optimizing turbine placement while ensuring environmental friendliness, is a pressing technical challenge in the current wind power engineering field. Summary of the Invention

[0006] This invention provides a method for evaluating the shadow of a wind turbine based on three-dimensional terrain occlusion judgment, the method comprising the following steps:

[0007] S1: Based on aerial real-scene perception, acquire and locate one or more environmentally sensitive targets, and generate a set of measured sensitive targets;

[0008] S2: Based on the geometric parameters of the wind turbine, calculate multiple key acquisition points to simulate the spatial range of its rotor; control the UAV to fly to multiple key acquisition points in sequence, and collect spatial image data including light source side images and receiver side images respectively;

[0009] S3: Analyze the light source side image in the spatial image data, extract the physical skyline outline, and combine it with the solar trajectory calculation to determine the potential light and shadow time window for each key acquisition point; within the potential light and shadow time window, analyze the receiver side image, project environmentally sensitive targets into the image to determine the visibility of each key acquisition point;

[0010] S4: If the wind turbine installation location currently being evaluated is affected by light and shadow, then based on the differences in visibility status of multiple key acquisition points, construct the shading gradient vector to determine the optimal shadow avoidance displacement direction, generate the coordinates of the next candidate location, and repeat steps S2 to S4 for the candidate location.

[0011] In step S1, acquiring and locating one or more environmentally sensitive targets based on aerial real-scene perception includes: controlling a drone to collect high-altitude panoramic images, converting the panoramic images to the HSV color space, defining the characteristic color gamut set of artificial materials, and targeting each pixel in the image. Constructing a binarized mask :

[0012] ;

[0013] in, These are the hue, saturation, and luminance components of the pixel, respectively. This represents the numerical range of artificial materials in the hue channel. and To suppress shadows and low-saturation backgrounds, thresholds are applied; morphological operations and connected component labeling are performed on the binarized mask to extract candidate region blocks; the geographic coordinates of the effective candidate region blocks are calculated using a monocular vision geolocation algorithm to generate a set of measured sensitive targets.

[0014] In step S2, the calculation of multiple key acquisition points for simulating the impeller space range includes: taking the hub center coordinates of the proposed installation location as the origin, and calculating the three-dimensional coordinates of the center point, upper limit point, lower limit point, left limit point and right limit point based on the impeller diameter parameters;

[0015] Among them, the upper limit point The coordinates are calculated as follows:

[0016] ;

[0017] in, Here are the planar coordinates of the proposed installation location. This refers to the ground elevation. For wheel hub height, The impeller diameter is [value missing].

[0018] In step S3, the light source-side image in the spatial image data is analyzed to extract the physical skyline contour, including: extracting the set of pixel coordinates of the skyline contour in the light source-side image. ,in The number of pixels in the skyline outline; for each pixel... Using the camera intrinsic parameter matrix and the camera rotation matrix during shooting Calculate its normalized direction vector in the camera coordinate system. and the world direction vector in the global geographic coordinate system :

[0019] ;

[0020] ;

[0021] World Direction Vector Convert to azimuth in astronomical coordinates and elevation angle A terrain occlusion map is generated. Combined with the calculation of the sun's trajectory, if the sun's altitude angle at a certain moment is lower than the terrain altitude angle corresponding to the sun's azimuth angle at that moment in the terrain occlusion map, it is determined to be an invalid sunshine period. After removing the invalid sunshine period, the potential light and shadow time window is obtained.

[0022] In step S3, the environmentally sensitive targets are projected onto the image, including: acquiring targets from the measured set of sensitive targets. 3D geographic coordinates And the location of the drone when acquiring recipient-side images. Camera intrinsic parameter matrix and camera rotation matrix ; Calculate the target Projected pixel position on the image :

[0023] ;

[0024] ;

[0025] ;

[0026] Based on the projected pixel position Construct a trapezoidal visual corridor on the image that extends from the bottom center of the image to the projection position.

[0027] In step S3, the visual corridor from the key acquisition point to the projection position is analyzed to determine if there is any occlusion, including:

[0028] An edge detection operator is applied within a trapezoidal visual corridor to calculate the intensity of the lateral edges. ;

[0029] The trapezoidal visual corridor is divided into multiple horizontal strips, and the texture abrupt change gradient between adjacent strips is calculated. :

[0030] ;

[0031] in, For the first Local binary pattern histogram of horizontal stripes, Represents the chi-square distance; if Below the preset edge strength threshold and Below the preset texture gradient threshold If the key acquisition point is visible, it is considered visible; otherwise, it is considered obscured.

[0032] In step S4, constructing the occlusion gradient vector includes:

[0033] Define a local coordinate base that contains a unit line-of-sight vector pointing towards the sensitive target. Global vertical upward unit vector and the unit vector perpendicular to the line of sight in the horizontal plane. Calculate the occlusion gradient vector based on the visibility Boolean values ​​of the upper, lower, left, and right extreme points. vertical component and horizontal components :

[0034] ;

[0035] ;

[0036] ;

[0037] in, This is an indicator function that takes a value of 1 when visibility is visible, and a value of 0 otherwise. and These are the weighting coefficients; This represents the visibility state of the k-th limit point with respect to the target q.

[0038] In step S4, the coordinates of the next candidate aircraft position are generated using an adaptive step-size strategy, including: calculating the number of exposure points. ;

[0039] The displacement step size for the current iteration is determined based on the number of exposure points. :

[0040] ;

[0041] in, The preset maximum search step size, This is the decay index.

[0042] The coordinates of the next candidate camera position are generated using the following formula:

[0043] ;

[0044] in, The coordinates of the current assessment position. The coordinates of the next candidate aircraft position are generated. After generating the candidate aircraft position coordinates, the constraint verification is also included: verifying whether the coordinates are located within the preset effective working ring and whether the surface slope at the coordinates is less than the preset maximum slope threshold.

[0045] The present invention also provides a wind turbine shadow assessment system based on three-dimensional terrain occlusion judgment, the system comprising:

[0046] Environmental perception module: Based on aerial real-scene perception, acquire and locate one or more environmentally sensitive targets, and generate a set of measured sensitive targets;

[0047] Image acquisition module: Based on the geometric parameters of the wind turbine, calculate multiple key acquisition points to simulate the spatial range of its rotor; control the UAV to fly to multiple key acquisition points in sequence, and acquire spatial image data including light source side image and receiver side image respectively;

[0048] The visibility analysis module analyzes the light source side image in the spatial image data, extracts the physical skyline outline, and combines it with the solar trajectory calculation to determine the potential light and shadow time window for each key acquisition point; within the potential light and shadow time window, it analyzes the receiver side image and projects environmentally sensitive targets into the image to determine the visibility of each key acquisition point.

[0049] Iterative module: If the wind turbine installation location currently being evaluated is affected by light and shadow, then based on the differences in visibility status of multiple key acquisition points, an occlusion gradient vector is constructed to determine the optimal shadow avoidance displacement direction, the coordinates of the next candidate location are generated, and image acquisition and visibility analysis are repeated.

[0050] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for evaluating the shadow of a wind turbine based on three-dimensional terrain occlusion.

[0051] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned wind turbine shadow assessment method based on three-dimensional terrain occlusion judgment.

[0052] The method and system for accurate assessment of wind turbine shadows based on three-dimensional terrain occlusion judgment provided by this invention have significant technical advantages and application value in the planning and environmental impact assessment of complex mountain wind farms.

[0053] This invention fundamentally overcomes the limitations of existing technologies that heavily rely on static digital elevation models and outdated geographic information data. In complex mountainous scenarios, conventional satellite remote sensing data often fails to accurately reflect the micro-undulations of the terrain, the seasonal growth of vegetation, and the changes in temporary structures. This invention utilizes UAV aerial real-scene perception technology to directly acquire high-resolution panoramic images of the area to be evaluated, and to lock the true coordinates of environmentally sensitive targets in real time. This method, based on on-site physical measurements, ensures the location accuracy and timeliness of the evaluated object, effectively avoiding the shift in evaluation benchmarks caused by map data lag or elevation errors, and laying a solid physical foundation for subsequent accurate analysis.

[0054] This invention overcomes the shortcomings of traditional evaluation methods that simplify large wind turbines into a single point light source. With the accelerating trend towards larger wind turbines, the hundreds-meter diameter of the rotor means that the blade tip and the hub center are in drastically different spatial positions. This invention constructs a virtual rotor spatial model, driving a drone to simulate the critical extreme positions swept by the rotor's rotation in physical space. This achieves a leap from single-point estimation to array-based full-domain perception, enabling the sensitive detection of edge-related light and shadow risks such as "hub obscured but blade tip exposed," completely eliminating blind spots caused by model simplification and ensuring comprehensive coverage of the light and shadow effects of ultra-large rotors.

[0055] This invention establishes a bidirectional line-of-sight logic verification mechanism based on real image content, achieving precise removal in both temporal and spatial dimensions. In the temporal dimension, by extracting the physical skyline contour from the light source-side image and combining it with a solar trajectory algorithm, invalid periods of sunlight obscured by distant mountains can be accurately removed, solving the problem of discrepancies between theoretical sunlight duration and actual terrain occlusion. In the spatial dimension, by constructing visual corridors in the recipient-side image and utilizing edge intensity and texture gradient analysis techniques, the blocking effect of near-field obstacles on the line of sight can be directly identified from the image. This WYSIWYG analysis method avoids the drawback of ignoring subtle occlusions such as trees and sharp edges due to terrain smoothing in 3D model collision detection, greatly reducing the false positive rate of shadow impact assessment and making the assessment conclusions highly approximate the real physical environment.

[0056] This invention represents a technological shift from passive evaluation to proactive optimization. Traditional evaluation methods can only provide static conclusions about whether a specific wind turbine location meets the standards. However, the intelligent displacement guidance strategy based on shading gradients proposed in this invention can automatically analyze the optimal shadow avoidance direction and step length according to the visibility differences at various extreme points. Through iterative optimization, the system can autonomously find the best installation point in complex terrain by utilizing natural terrain features such as ridges and passes for shading. This not only significantly improves the efficiency of micro-site selection for wind farms and reduces the cost of repeated manual trial and error, but also maximizes the utilization of wind-rich areas while ensuring environmental friendliness. It effectively balances the economics of engineering construction with the quality of life of surrounding residents, providing strong technical support for the large-scale development of mountain wind power. Attached Figure Description

[0057] Figure 1 This is a flowchart of the wind turbine shadow assessment based on three-dimensional terrain occlusion judgment according to the present invention. Detailed Implementation

[0058] This embodiment provides a basic framework and application scenario configuration for a wind turbine shadow assessment method based on three-dimensional terrain occlusion judgment. The system and method are mainly applied to wind farm planning areas with complex terrain features, specifically referring to mountainous or hilly wind farms where the surface has significant undulations such as ridges, ravines, and steep cliffs, and where light-sensitive targets such as villages, scattered farmhouses, schools, or roads are located nearby.

[0059] In this embodiment, the wind turbine shadow accuracy assessment system mainly consists of two parts: a data acquisition subsystem and a data processing subsystem.

[0060] The data acquisition subsystem is used to simulate the coverage area of ​​wind turbine rotors in physical space and acquire multi-dimensional line-of-sight data. The core equipment of this subsystem is a multi-rotor unmanned aerial vehicle (UAV) with high-precision positioning and hovering capabilities.

[0061] To meet the requirements of accurate assessment, the UAV needs to be equipped with an RTK differential positioning module to provide centimeter-level spatial positioning accuracy. In complex mountainous environments, conventional GPS positioning errors may cause elevation data to deviate from the true value, thus affecting the accuracy of line-of-sight occlusion judgment. Therefore, this embodiment preferably adopts a positioning method that connects to a network RTK base station or sets up a mobile D-RTK base station to ensure strict alignment between the UAV's flight coordinates and the wind turbine's design coordinate system. The UAV is also equipped with a three-axis stabilized gimbal camera with a high-resolution optical imaging lens and a wide-angle field of view. The dynamic range of the camera's image sensor meets the requirement of clear imaging under both strong light (facing the sun) and backlight (facing the ground shadow) conditions to ensure that it can simultaneously identify the sunlight background in the sky and the terrain texture on the ground. The UAV also includes a flight control unit with a built-in waypoint planning algorithm, which supports automatic cruising and hovering in three-dimensional space based on the input coordinate point array.

[0062] The data processing subsystem is a high-performance graphics workstation deployed at a ground station or back office. This subsystem is equipped with Geographic Information System (GIS) software, photogrammetry processing software, and the shadow analysis algorithm module of this invention. Its main functions include: importing macroscopic site selection data for wind farms, generating UAV flight waypoint instructions, receiving and processing acquired spatial image data, building a 3D terrain model, and performing line-of-sight occlusion logic operations.

[0063] Before performing the evaluation task, the basic data must be initialized in the data processing subsystem.

[0064] The system first imports the digital elevation model (DEM) data and orthophoto map of the area to be evaluated. This DEM data can be preliminary data acquired based on satellite remote sensing, mainly used for planning the UAV's flight path and safe obstacle avoidance altitude, rather than being directly used for the final shadow determination. At the same time, the system marks the center coordinates (longitude, latitude, and altitude) of the proposed wind turbine installation site and the outline boundaries of surrounding sensitive targets.

[0065] Input the geometric parameters of the wind turbine generator set to be selected. Key parameters include: the vertical distance from the top of the wind turbine tower to the center of rotation of the rotor, i.e., the hub height; the diameter of the disk plane formed by the rotation of the wind turbine blades, i.e., the rotor diameter; and the maximum blade tip height, which represents the highest point of the shadow projection source, calculated by adding half of the rotor diameter to the hub height.

[0066] To overcome the shortcomings of existing technologies that treat the wind turbine as a single point light source, this embodiment constructs a virtual impeller spatial model in the data processing subsystem. This model uses the hub center coordinates of the proposed installation location as the origin and calculates the three-dimensional coordinates of five key feature points based on the impeller diameter parameter. These five points constitute the target hovering position for subsequent array-based data acquisition by the UAV. The five key feature points include: Center point: corresponding to the physical spatial position of the wind turbine hub center; Upper limit point: corresponding to the blade tip position when the blade rotates to its highest point, with its elevation being the hub center elevation plus the impeller radius; Lower limit point: corresponding to the blade tip position when the blade rotates to its lowest point, with its elevation being the hub center elevation minus the impeller radius; Left and right limit points: corresponding to the widest positions of the impeller rotation plane in the horizontal direction.

[0067] Next, this embodiment elaborates on the initial planning and global coarse screening process in the method of the present invention. As the preliminary stage of the entire evaluation process, this process does not rely solely on static geographic information system data, but rather leverages the advantage of the UAV's high-altitude perspective to actively discover and locate potential light- and shadow-sensitive targets through real-time perception algorithms, thereby defining the evaluation boundary and initializing flight parameters.

[0068] S1.1 Sensitive Target Extraction Based on Aerial Reality Perception

[0069] In complex mountainous wind farm scenarios, ground vegetation typically exhibits irregular textures and specific green / brown color bands, while human settlements such as villages and detached houses usually have roofs made of corrugated steel sheets (blue, red) or concrete (grayish-white) with regular geometric edges. Based on these physical differences, this step obtains the coordinates of sensitive points through the following logic.

[0070] S1.1.1: Scan to obtain a high-altitude panoramic view, and control the drone to fly to the center coordinates of the wind turbine to be installed. A preset reconnaissance altitude above Preferably, the distance is 50 meters above the wheel hub height. The drone performs 360-degree panoramic photography to acquire high-resolution panoramic images of the environment. .

[0071] S1.1.2: Morphological segmentation based on HSV color space. To overcome the influence of varying lighting conditions in mountainous areas on the RGB color space, the system first segments the panoramic image... Convert to HSV color space.

[0072] Define the characteristic color gamut set of artificial materials This characteristic color gamut set contains the highlight portions of blue corrugated steel, red roof tiles, and white cement. For each pixel in the image... Let u and v be pixel coordinates, and construct a binary mask. :

[0073] ;

[0074] in, These are the hue, saturation, and luminance components of the pixel, respectively. and A threshold for suppressing shadows and low-saturation backgrounds; This represents the range of numerical values ​​for artificial materials in the hue channel.

[0075] right Morphological closing operations are performed to fill texture holes within the roof area, and several candidate region blocks are extracted using a connected component labeling algorithm. .

[0076] S1.1.3: False target removal based on geometric regularity. In mountainous areas, there may be misclassifications due to exposed rocks or ponds. This method utilizes the rectangular or polygonal outlines of man-made structures to filter candidate regions. Perform geometric constraint screening.

[0077] Calculate each region block The area of ​​the minimum bounding rectangle and its actual pixel area Define the rectangularity operator :

[0078] ;

[0079] Simultaneously, the edge smoothness operator of the calculated region block is used. Candidate region blocks are extracted using the Canny edge detection operator. boundary pixel set The Hough transform is then applied to the boundary to detect straight lines. The proportion of all detected straight line segments in the total length of the boundary is calculated; this proportion is defined as... .

[0080] The outline of man-made structures is usually composed of long straight line segments, therefore its The value will be significantly higher than the irregular boundaries of natural objects (such as rocks).

[0081] Only when and At that time, the area was determined. For effective environmentally sensitive target areas.

[0082] in For shape threshold, This represents the minimum house pixel area calculated based on flight altitude.

[0083] S1.1.4: Monocular visual geolocation, which uses the UAV's attitude information to map image pixel coordinates to geographic coordinates. The camera intrinsic parameter matrix is ​​known. The rotation matrix of the drone at the moment of shooting. and position coordinates For the center pixel of the identified sensitive target region. Construct the line-of-sight vector :

[0084] ;

[0085] Calculate the ray using a ray casting algorithm. With rough digital elevation model The intersection point is the geographic coordinate of the j-th identified sensitive target. This generates a set of measured sensitive targets. .

[0086] S1.2 Physical Shadow Limits and Global Coarse Screening

[0087] Based on the measured set of sensitive targets, and combined with the physical parameters of the wind turbines and the laws of solar motion, targets that are theoretically impossible to be affected are eliminated.

[0088] Obtain wind turbine parameters: hub height and impeller diameter Calculate the elevation of the highest point at the leaf tip. :

[0089] ;

[0090] Set the critical solar altitude angle one hour after sunrise / sunset on the winter solstice (when the solar altitude angle is lowest and the shadow is longest). Based on the three-dimensional geometric projection relationship, calculate the maximum physical shadow projection radius generated by the wind turbine. Considering that sensitive targets in mountainous areas inherently possess elevation... The radius is a function of the target elevation:

[0091] ;

[0092] For sets Each target in Calculate the target Horizontal Euclidean distance from the wind turbine plane :

[0093] ;

[0094] like If the target is outside the physical shadow limit, it is determined that the target is directly eliminated; if If the target is identified as a potential risk target, it will be added to the list. This will be the focus of subsequent refined inspections.

[0095] S1.3 Flight Path Initialization and Safety Envelope Setting

[0096] Based on certainty Determine the distribution orientation and initialize the drone's data acquisition array parameters.

[0097] Construct a virtual impeller coordinate system with the wind turbine installation point Establish a local coordinate system with the origin as the origin. Calculate the positions of five key acquisition points (center point, upper / lower / left / right extreme points) in the global coordinate system.

[0098] The above extreme points For example, its coordinates are calculated as follows:

[0099] ;

[0100] To prevent drones from subsequently targeting If the aircraft crashes into a mountain while performing low-altitude side-flying maneuvers or adjusting its perspective, the system will respond accordingly. The closest target distance for medium-range wind turbines and the farthest distance Define the effective working loop .

[0101] exist Retrieve digital elevation models within the specified range Maximum terrain height Set an absolute safe flight limit for drone operations. :

[0102] ;

[0103] in For safety margin, The elevation of the lowest sensitive point.

[0104] If the calculated lower limit point of the wind turbine, i.e., the lowest point elevation of the blade tip, is obtained... If the terrain interference warning is triggered, the system will automatically generate a terrain interference warning, indicating that there may be a flight risk at the location at a certain low angle, and the area should be marked as a restricted detection zone in the subsequent path planning.

[0105] To make the embodiments of this application more complete and clear, the present invention provides an exemplary implementation to illustrate the initialization and security envelope settings.

[0106] Assuming the wind farm is located on a ridge, the initial parameters obtained are as follows: wind turbine installation point ( Ground elevation =900.0 meters, with the plane coordinate system set to origin (0,0). Hub height =100.0 meters, impeller diameter =160.0 meters (radius R=80.0 meters). Set of measured sensitive targets. Two main risk objectives have been identified: Objective A azimuth (Due East), Horizontal Distance =600.0 meters. Target B azimuth (Southeast), horizontal distance =1200.0 meters. Nearest target distance. =600.0 meters. Distance to the furthest target. =1200.0 meters. Topographic data. Recorded on the east side of the wind turbine (azimuth angle) At a horizontal distance of 180.0 meters from the center of the wind turbine, there is a prominent small hill (a terrain obstacle point). Its summit has an elevation of 920.0 meters.

[0107] To ensure that the drone does not fly out of the safe range during simulated blade rotation and subsequent possible intelligent side-flight, the system relies on... Construction of distribution characteristics .

[0108] Extract the azimuth angles of all sensitive targets and calculate the coverage sector.

[0109] ;

[0110] ;

[0111] in As the field of view redundancy angle, this invention sets it to be The flight radius of the drone Two physical constraints must be met: the physical range of the virtual impeller must be included. When conducting side-flight maneuvers to locate line-of-sight gaps, the flight distance should not be too far to avoid excessive parallax leading to simulation distortion.

[0112] The system defines the maximum permissible maneuver radius. for Functions:

[0113] ;

[0114] Set maximum physical limit Meters, parallax control coefficient Substitute the values:

[0115] .

[0116] Therefore, effective working loop Defined as: With the center as the azimuth angle at the center, Between, radial distance A fan-shaped area between meters.

[0117] In the generated Within the region, for digital elevation models Perform a detailed scan to search for the highest terrain point.

[0118] exist (Specific coordinates: location) (180 meters away, within a 240-meter radius) The system detected the obstacle point. Elevation =920.0 meters. Calculate the absolute safe flight lower limit. Set up a safe airspace around the ground. =30.0 meters.

[0119]

[0120] Calculate the theoretical altitude of the lower limit point that the drone needs to simulate based on the wind turbine parameters:

[0121]

[0122] Comparing theoretical waypoints with the lower safety limit:

[0123]

[0124] A terrain collision high-risk alert has been triggered. Although the elevation directly below the wind turbine is only 900 meters, this is within the operational ring where drone activity is permitted. There is a 920-meter-high hill inside (specifically 180 meters to the east). If the drone descends to 920 meters to simulate the position of a leaf tip and then attempts to fly eastward, it is highly likely to collide with the hill (because 920 meters is exactly the height of the mountaintop, with no clearance).

[0125] The system automatically clamps the flight altitude at the lower limit point to the safe lower limit:

[0126]

[0127] Simultaneously, a no-fly zone command is generated: Drones are prohibited from flying towards the azimuth angle when their altitude is below 950 meters. The sector is laterally displaced.

[0128] Traditional assessment methods treat the wind turbine as a single point (usually the hub center). If the hub center is blocked by a small hill in front, the traditional method considers it to have no shading effect. However, modern wind turbines are enormous, with rotor diameters reaching 180 meters or even larger. This means that the turbine blades not only move vertically (up and down), but also extend horizontally (left and right) by nearly 90 meters (radius). This presents a very subtle risk: while the center is blocked, light leaks in from both sides. For example, there might be a small hill directly in front of the turbine tower, blocking the hub center. However, when the blades rotate to the 3 o'clock or 9 o'clock position, the blade tips might extend beyond the hill's obstruction, exposed into a valley or pass. At this point, sunlight will penetrate this lateral gap, casting the blade's shadow into the village.

[0129] Side-flying involves not just keeping the drone stationary in the center, but simulating a blade rotating to the far left and far right positions to check for light leakage at these edges. This is one of the key methods used in this invention to achieve "precise evaluation."

[0130] Next, this embodiment details the multi-point arrayed spatial image acquisition method of the present invention. In previous embodiments, the present invention calculated five key acquisition points capable of simulating the spatial range of a wind turbine impeller. The input of this embodiment is the generated set of virtual impeller spatial coordinates. and potential risk target set .in, These represent the coordinates of the center, highest, lowest, leftmost, and optimal positions of the virtual impeller, respectively.

[0131] S2.1 Waypoint Cruise and High-Precision Hovering Stabilization

[0132] To ensure that the acquired images accurately reflect the theoretical viewpoint, the drone flies to each key acquisition point. After that, a strict hovering maneuver must be executed. .

[0133] In complex mountainous environments, the terrain often creates irregular updrafts or turbulence, posing a challenge to the hovering stability of drones. To address this, the system employs a three-dimensional spatial position tolerance sphere. Only when the drone's position remains stable for a preset duration... Within a given timeframe, a hovering action is considered stable and triggers the image acquisition process only when the following conditions are met consecutively:

[0134] ;

[0135] in, It is the real-time three-dimensional coordinate of the UAV RTK module at time t. These are the coordinates of the target's hovering point. It is the location tolerance threshold. This represents the Euclidean distance.

[0136] S2.2 Dual-hemispherical panoramic spatial image array acquisition

[0137] After achieving stable hovering, the drone completed its mission at each key data collection point. At all locations, a standardized dual-hemispherical panoramic image acquisition strategy is implemented. The dual-hemispherical panoramic image acquisition strategy is not a simple 360-degree planar scan, but rather a differentiated and focused data acquisition strategy for the sky (light source side) and the ground (receiver side) based on the light source-path-receiver three-element model of light and shadow analysis in this invention.

[0138] S2.2.1: Dynamic calculation of acquisition parameters. Before shooting begins, the sequence parameters required for panoramic shooting are dynamically calculated based on the camera lens parameters carried by the drone.

[0139] Given the horizontal field of view of the camera lens and vertical field of view And the overlap rate set to ensure image stitching quality. Number of shots taken in the horizontal direction (yaw angle) The calculation is as follows:

[0140] ;

[0141] in This is the floor function.

[0142] S2.2.2: Light source side (upper hemisphere) scanning

[0143] The drone gimbal tilts the camera lens upwards at a preset pitch angle. The drone uses incremental yaw angles. Rotate sequentially, at each yaw angle position A single shot was taken. This process captured a complete upper hemisphere view, including the sky, clouds, and distant ridgelines.

[0144] S2.2.3: Receptor-side (lower hemisphere) scan

[0145] The drone gimbal presses the camera lens down at a preset pitch angle. Similarly, the drone at each yaw angle position Perform a single shot. This process ensures coverage from the current viewpoint to all potentially hazardous targets. Along the way, the terrain and landforms (such as nearby hills, tree lines, and other buildings) were all recorded without omission.

[0146] S2.2.4: High Dynamic Range Image Capture

[0147] In mountainous environments, the difference in light intensity between the sunlit and shaded sides is significant. To avoid overexposure or underexposure in the images, which could lead to loss of skyline contours or details in the darker parts of the ground, specific camera positions (i.e., specific yaw angles) are crucial. and pitch angle or In all cases (combined), automatic exposure bracketing (AEB) mode is used for shooting. Preferably, a three-shot bracketing sequence is executed: These correspond to exposure compensation of -1EV, 0EV, and +1EV, respectively.

[0148] S2.3 Data Structured Storage

[0149] Complete a key data collection point After all the images are captured, the acquired image data and metadata are structured, bound together, and stored. For points... Its output is a data packet. Its structure is represented as:

[0150] ;

[0151] in, It is a specific image file. It contains metadata including the precise 3D coordinates of the drone, gimbal attitude, camera intrinsic parameters, and other information at the time the image was captured.

[0152] By sequentially The above S2.1 to S2.3 processes are executed at all five points. In this embodiment, five structured high dynamic range panoramic image data packets are finally generated. .

[0153] Next, this embodiment details the bidirectional line-of-sight logic verification process in the method of the present invention. By performing depth analysis on the light source and receiver images, the invalid light and shadow effects caused by real terrain occlusion are eliminated from both temporal and spatial dimensions.

[0154] S3.1 Validity verification of the light source side, i.e., elimination in the time dimension.

[0155] Based on the real sky background captured by drones at various key collection points, a physical skyline is constructed, and based on this, invalid sunshine periods throughout the year are eliminated because the sunlight cannot reach the wind turbines due to being blocked by distant mountains.

[0156] S3.1.1 Batch Extraction of Skyline Contours

[0157] To address the interference of complex lighting conditions in mountainous terrain on the identification of the sky and mountain boundaries, the system uses five structured high dynamic range panoramic image data packets. Each viewpoint k in the (where Perform contour extraction algorithm on the upper hemisphere image:

[0158] Convert the image from the RGB color space to the HSV color space. Select the V (luminance) channel of the image. This process maximizes the contrast between the bright areas of the sky and the low-brightness areas of the mountains.

[0159] Applying a Gaussian difference filter to The image is processed to enhance appropriately scaled edges, i.e., the skyline, while suppressing high-frequency noise such as clouds and slow changes in illumination.

[0160] The filtered result is binarized to generate a binary image. The skyline is represented by continuous or semi-continuous boundaries.

[0161] Using contour tracking algorithm in The longest transverse image is traced and identified as the skyline.

[0162] Generate a set of pixel coordinates for the skyline outline for each viewpoint k. . The number of pixels in the skyline outline.

[0163] S3.1.2 Pixel Outline to Astronomical Coordinate Mapping

[0164] Five sets of skyline outline points generated and the camera intrinsic parameter matrix associated with each image. and the camera rotation matrix during shooting .

[0165] Convert the skyline on a 2D image into physical directions in 3D space. Each pixel in Perform the following back projection transformation:

[0166] Calculate the normalized direction vector of the pixel in the camera coordinate system. :

[0167] ;

[0168] Transform the direction vector from the camera coordinate system to the global geographic coordinate system to obtain the world direction vector. : ;

[0169] World Direction Vector Convert to azimuth in astronomical coordinates and elevation angle :

[0170] ;

[0171] ;

[0172] Based on skyline contour point set For each pixel in the map, the corresponding azimuth and elevation angles are used to generate a terrain occlusion map in functional form for each viewpoint k. This function takes an azimuth angle as input and returns the highest elevation angle of the physical skyline in that direction through interpolation.

[0173] S3.1.3 Comparison of solar trajectory and terrain occlusion

[0174] The system employs a standard Solar Position Algorithm (SPA), combined with the wind turbine's geographical location, to determine the validity of sunshine periods throughout the year. It iterates through every day 'd' of the year, and every time point from sunrise to sunset on that day. At a certain point in time Calculate the precise position of the sun to obtain its azimuth angle. and elevation angle .

[0175] For each viewpoint k, query the terrain shading height of that viewpoint at that solar azimuth angle: .like Then it is determined at this moment The sunlight was blocked by the distant mountains and could not reach the viewpoint k.

[0176] Generate a set of potential lighting and shadow time windows for each viewpoint k. This set consists of a series of time intervals, representing all periods of the year during which sunlight can physically reach that particular viewpoint.

[0177] S3.2 Receptor-side visibility analysis, i.e., spatial dimension elimination

[0178] Within the effective time window determined in S3.1, this step further analyzes whether the line of sight from the wind turbine viewpoint to the sensitive target is blocked by near- or mid-range obstacles.

[0179] S3.2.1 3D Coordinate Projection of the Target Area

[0180] This invention has obtained a set of potential risk targets. Each target 3D geographic coordinates The data also includes panoramic images of the lower hemisphere (receptor side) and their metadata (camera intrinsic and extrinsic parameters) from five data packets. and drone location ).

[0181] For each viewpoint-target pair Perform standard camera forward projection and calculate the target. The position of the projected pixel on the image at viewpoint k :

[0182] ;

[0183] ;

[0184] ;

[0185] On each relevant lower hemisphere image, the theoretical location where the sensitive target should appear was marked as the focus of subsequent analysis.

[0186] S3.2.2 Visual Occlusion Analysis

[0187] Construct a projection point on the image that extends from the bottom center of the image to the target projection point. Trapezoidal visual corridor And perform double testing within this corridor:

[0188] Lateral edge strength detection: The Canny edge detection operator is applied internally. For each detected edge pixel, its gradient direction is calculated. Lateral edge strength is also considered. This is the sum of the gradient magnitudes of all edge pixels within the corridor whose dot product with the corridor's principal axis (from bottom to top) is close to zero (i.e., the edges are essentially horizontal). A larger value indicates the presence of a strong ridge or canopy line that laterally obstructs the view.

[0189] Texture discontinuity analysis: Divide into M horizontal strips of equal height from bottom to top. .

[0190] For each strip , Extract its Local Binary Pattern (LBP) histogram , which serves as the texture descriptor for that region.

[0191] Calculate the differences in texture descriptors between adjacent stripes. Texture abrupt change gradient. The maximum chi-square distance between all adjacent histograms:

[0192] .

[0193] Larger values ​​indicate abrupt changes in land cover types along the visual path, such as a sudden shift from nearby grassland to distant rocky outcrops.

[0194] Set a preset edge strength threshold and texture gradient threshold For viewpoint-target pairs Its visibility Defined as:

[0195] .

[0196] That is, the line of sight is considered visible (True) only when both the horizontal edge intensity and the texture abrupt gradient are below their respective thresholds; otherwise, it is considered occluded.

[0197] For each wind turbine viewpoint - sensitive target This generates a final Boolean result. .

[0198] Next, this embodiment elaborates on the intelligent displacement guidance based on occlusion gradient in the method of the present invention. This is done for the current proposed installation location. (where n is the number of iterations, and the initial position is...) This process begins after the assessment of the light and shadow effects is completed. If the assessment results indicate the presence of light and shadow effects, this step analyzes the differences in visibility at four extreme points (up / down / left / right) to construct an occlusion gradient vector indicating the optimal shadow avoidance direction. Combined with an adaptive step-size strategy, this autonomously guides the drone and its simulated virtual wind turbine to a new candidate location with better light and shadow effects. The evaluation is repeated for new locations until the best location that meets the environmental requirements is found or all preset feasible areas are traversed.

[0199] For the current camera position and a specific sensitive target First, based on the generated vector consisting of the Boolean values ​​of the visibility of the four limit points... .

[0200] Construct an occlusion gradient vector The construction of this vector transforms discrete visibility states into a continuous directional command that guides movement in physical space. To this end, a local coordinate base is defined, consisting of a unit line-of-sight vector pointing towards the sensitive target. Global vertical upward unit vector and the unit vector perpendicular to the line of sight in the horizontal plane. Composition. The vertical component of the occlusion gradient vector. and horizontal components The calculation is as follows:

[0201] ;

[0202] ;

[0203] in, This is an indicator function; it takes the value 1 when its internal logic is true and 0 when it is false.

[0204] In a typical mountain wind farm scenario, if the upper limit point is visible while the lower limit point is obscured... The downward pointing indicates that lowering the camera's elevation is an effective strategy to avoid obstruction by the ridge in front; similarly, if the right extreme point is visible while the left extreme point is obstructed,... This will point to the left, guiding the camera position to move towards an area that can provide more occlusion. The final occlusion gradient vector is a weighted synthesis of these two components: The weighting coefficient and Configuration can be made based on engineering experience.

[0205] After determining the optimal direction of movement, an adaptive displacement step size strategy is employed to determine the movement distance, balancing search efficiency in large-scale regions with convergence accuracy in near-optimal regions. The core logic of this strategy is that the displacement step size is negatively correlated with the risk level at the current location (i.e., the degree of impeller exposure). First, a simplified number of exposure points is calculated. :

[0206] ;

[0207] Based on this value, the displacement step size for the current iteration Determined by the following piecewise function:

[0208] ;

[0209] in, It is the preset maximum search step size, preferably the impeller radius. ; It is the minimum fine-tuning step size; It is an exponent that controls the nonlinear decay rate of the step size, preferably 1.5. This strategy ensures that when the exposed surface of the wind turbine is large ( The system uses its maximum step size to quickly escape high-risk areas; when only one or two critical points are exposed, indicating that it is approaching the critical edge of the obstruction, the system will automatically reduce its step size for a finer search to accurately pinpoint the best hiding place; when all points are obstructed ( =0), the step size is zero, indicating that the ideal position has been found, and the iterative process terminates naturally for this sensitive target.

[0210] By combining the direction indicated by the occlusion gradient vector and the distance determined by the adaptive step size, the system calculates the coordinates of the next candidate camera position. .

[0211] To ensure that the new location is feasible in engineering, it will be subject to strict constraint checks.

[0212] First, verify whether the point exceeds the defined effective working loop centered on the initial machine position. The geographical scope is defined to prevent the search path from spreading indefinitely. Secondly, the digital elevation model is queried. ,calculate The slope of the ground at the location And verify whether it is less than the maximum slope threshold allowed for wind turbine foundation construction. Only when all constraints are met is the candidate site confirmed as a valid next evaluation site. This triggers a new round of evaluation (returning to execute S2 to S4); if any constraint is not met, the system will determine that the current movement path is blocked and attempt to use suboptimal strategies such as halving the step size or fine-tuning the direction to avoid it.

[0213] If multiple attempts fail, the search direction will be terminated to ensure that all evaluation points are engineering feasible. Through this iterative optimization process, the present invention can automatically plan an efficient evaluation path in a terrain-complex area, ultimately converging to one or more optimal wind turbine installation points that meet the light and shadow environmental protection requirements.

[0214] The present invention also provides a wind turbine shadow assessment system based on three-dimensional terrain occlusion judgment, the system comprising:

[0215] Environmental perception module: Based on aerial real-scene perception, acquire and locate one or more environmentally sensitive targets, and generate a set of measured sensitive targets;

[0216] Image acquisition module: Based on the geometric parameters of the wind turbine, calculate multiple key acquisition points to simulate the spatial range of its rotor; control the UAV to fly to multiple key acquisition points in sequence, and acquire spatial image data including light source side image and receiver side image respectively;

[0217] The visibility analysis module analyzes the light source side image in the spatial image data, extracts the physical skyline outline, and combines it with the solar trajectory calculation to determine the potential light and shadow time window for each key acquisition point; within the potential light and shadow time window, it analyzes the receiver side image and projects environmentally sensitive targets into the image to determine the visibility of each key acquisition point.

[0218] Iterative module: If the wind turbine installation location currently being evaluated is affected by light and shadow, then based on the differences in visibility status of multiple key acquisition points, an occlusion gradient vector is constructed to determine the optimal shadow avoidance displacement direction, the coordinates of the next candidate location are generated, and image acquisition and visibility analysis are repeated.

[0219] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for evaluating the shadow of a wind turbine based on three-dimensional terrain occlusion.

[0220] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned wind turbine shadow assessment method based on three-dimensional terrain occlusion judgment.

[0221] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0222] In this specification, the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the descriptions of the embodiments described later are relatively simple, and relevant parts can be referred to the descriptions of the foregoing embodiments.

[0223] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for evaluating the shadow of a wind turbine based on three-dimensional terrain occlusion judgment, characterized in that, The method includes: S1: Based on aerial real-scene perception, acquire and locate one or more environmentally sensitive targets, and generate a set of measured sensitive targets; S2: Based on the geometric parameters of the wind turbine, calculate multiple key acquisition points to simulate the spatial range of its rotor; the multiple key acquisition points specifically include the three-dimensional coordinates of the center point, upper limit point, lower limit point, left limit point and right limit point calculated based on the rotor diameter parameters, with the hub center coordinates of the proposed installation location as the origin; The drone was controlled to fly sequentially to multiple key acquisition points and collect spatial image data, including images from the light source side and the receptor side, respectively. S3: Analyze the light source side image in the spatial image data to extract the physical skyline outline, specifically: Extract the set of pixel coordinates of the skyline outline from the light source side image. ,in The number of pixels in the skyline outline; for each pixel... Using the camera intrinsic parameter matrix and the camera rotation matrix during shooting Calculate its normalized direction vector in the camera coordinate system. and the world direction vector in the global geographic coordinate system : ； ； World Direction Vector Convert to azimuth in astronomical coordinates and elevation angle A terrain occlusion map is generated. Combined with solar trajectory calculations, if the solar elevation angle at a certain moment is lower than the corresponding terrain elevation angle in the terrain occlusion map, it is determined to be an invalid sunshine period. After removing invalid sunshine periods, a potential light and shadow time window is obtained. Within the potential light and shadow time window, the receiver-side image is analyzed, and environmentally sensitive targets are projected onto the image. Based on the target... Projected pixel position A trapezoidal visual corridor extending from the bottom center of the image to the projected location is constructed on the image to determine the visibility of each key acquisition point; specifically, an edge detection operator is applied within the trapezoidal visual corridor to calculate the intensity of the lateral edges. ; The trapezoidal visual corridor is divided into multiple horizontal strips, and the texture abrupt change gradient between adjacent strips is calculated. : ； in, For the first A local binary pattern histogram of horizontal stripes, where M is the number of horizontal stripes of equal height. Represents the chi-square distance; if Below the preset edge strength threshold and Below the preset texture gradient threshold If the key acquisition point is visible, it is determined to be visible; otherwise, it is determined to be occluded. S4: If the proposed installation location of the wind turbine is affected by light and shadow, then based on the visibility differences of multiple key acquisition points, an shading gradient vector is constructed to determine the optimal shadow avoidance displacement direction, and the coordinates of the next candidate turbine location are generated. The formula for generating the coordinates of the next candidate turbine location is as follows: ; in, To occlude the gradient vector, It is calculated based on the Boolean values ​​of the visibility of the upper limit point, lower limit point, left limit point, and right limit point; Calculate the number of exposure points The displacement step size for the current iteration is determined based on the number of exposure points. : in, The coordinates of the current assessment position. The coordinates of the next candidate camera position; This is an indicator function that takes a value of 1 when visibility is visible, and a value of 0 otherwise. and These are the weighting coefficients; This represents the visibility state of the k-th limit point with respect to the target q; Repeat steps S2 to S4 for the candidate location.

2. The method for evaluating wind turbine shadows based on three-dimensional terrain occlusion judgment according to claim 1, characterized in that, In S1, acquiring and locating one or more environmentally sensitive targets based on aerial real-world perception includes: controlling a drone to collect high-altitude panoramic images, converting the panoramic images to the HSV color space, defining the characteristic color gamut set of artificial materials, and targeting each pixel in the image. Constructing a binarized mask : ； in, These are the hue, saturation, and luminance components of the pixel, respectively. This represents the numerical range of artificial materials in the hue channel. and To suppress shadows and low-saturation backgrounds, a threshold is set; morphological operations and connected component labeling are performed on the binarized mask to extract candidate region blocks; the geographic coordinates of the effective candidate region blocks are calculated using a monocular vision geolocation algorithm to generate a set of measured sensitive targets.

3. The method for evaluating wind turbine shadows based on three-dimensional terrain occlusion judgment according to claim 1, characterized in that, Upper limit point The coordinates are calculated as follows: ； in, Here are the planar coordinates of the proposed installation location. This refers to the ground elevation. For wheel hub height, The impeller diameter is [value missing].

4. The method for evaluating wind turbine shadows based on three-dimensional terrain occlusion judgment according to claim 1, characterized in that, In S3, projecting environmentally sensitive targets onto the image includes: acquiring targets from the measured set of sensitive targets. 3D geographic coordinates And the location of the drone when acquiring recipient-side images. Camera intrinsic parameter matrix and camera rotation matrix ; Calculate the target Projected pixel position on the image : ； ； ； Based on the projected pixel position Construct a trapezoidal visual corridor on the image that extends from the bottom center of the image to the projection position.

5. The method for evaluating wind turbine shadows based on three-dimensional terrain occlusion judgment according to claim 1, characterized in that, In S4, constructing the occlusion gradient vector includes: Define a local coordinate base that contains a unit line-of-sight vector pointing towards the sensitive target. Global vertical upward unit vector and the unit vector perpendicular to the line of sight in the horizontal plane. ; ； ； 。 6. The wind turbine shadow assessment method based on three-dimensional terrain occlusion judgment according to claim 1 or 5, wherein the displacement step size of the iteration... Determined by the following piecewise function: ； in, The preset maximum search step size, This is the decay index.

7. The method for evaluating wind turbine shadows based on three-dimensional terrain occlusion judgment according to claim 6, characterized in that, After generating the candidate machine location coordinates, the process also includes constraint verification: verifying whether the coordinates are located within the preset effective working ring and whether the surface slope at the coordinates is less than the preset maximum slope threshold.

8. A wind turbine shadow assessment system based on three-dimensional terrain occlusion judgment, characterized in that, The system is used to perform the wind turbine shadow assessment method based on three-dimensional terrain occlusion determination as described in claim 1, the system comprising: Environmental perception module: Based on aerial real-scene perception, acquire and locate one or more environmentally sensitive targets, and generate a set of measured sensitive targets; Image acquisition module: Based on the geometric parameters of the wind turbine, calculate multiple key acquisition points to simulate the spatial range of its rotor; control the UAV to fly to multiple key acquisition points in sequence, and acquire spatial image data including light source side image and receiver side image respectively; The visibility analysis module analyzes the light source side image in the spatial image data, extracts the physical skyline outline, and combines it with the solar trajectory calculation to determine the potential light and shadow time window for each key acquisition point; within the potential light and shadow time window, it analyzes the receiver side image and projects environmentally sensitive targets into the image to determine the visibility of each key acquisition point. Iterative module: If the wind turbine installation location currently being evaluated is affected by light and shadow, then based on the differences in visibility status of multiple key acquisition points, an occlusion gradient vector is constructed to determine the optimal shadow avoidance displacement direction, the coordinates of the next candidate location are generated, and image acquisition and visibility analysis are repeated.

Images