A method for unmanned aerial vehicle autonomous tracking and inspection of wind turbine blades

Abstract

This invention relates to the field of wind turbine inspection technology, and more particularly to a method for autonomous tracking and inspection of wind turbine blades using a drone, comprising the following steps: S1, establishing a cylindrical enclosure model of the wind turbine, and calculating the inspection start and end points for each shooting surface; S2, performing collision detection on the inspection start and end points, as well as the lines connecting the start and end points; S3, calculating the stopping conditions for the drone's inspection route; S4, identifying the wind turbine blades at the initial position, adjusting the camera position, and initially aligning with the target; S5, the drone dynamically tracks the blades; S6, updating the blade stopping angle in real time based on changes in the flight vector, adjusting the drone's inspection end point, updating the collision enclosure, and performing collision detection. This invention, based on achieving coarse positioning of the wind turbine, relies on simple spatial relationships to dynamically track and photograph the blades in real time, performing real-time collision detection, thus achieving safe, reliable, and clear inspection.

Description

Technical Field

[0001] This invention relates to the field of wind turbine testing technology, and in particular to a method for autonomous tracking and inspection of wind turbine blades using unmanned aerial vehicles (UAVs). Background Technology

[0002] Existing wind turbine inspection methods require an RTK positioning system and pre-collected precise GPS geographic information of the wind turbine generator. This necessitates pre-entering manufacturing parameters of the turbine blades, such as impeller cone angle and impeller deflection angle. During operation, the impeller position must remain constant. Based on image recognition of the turbine azimuth and impeller stopping angle, fixed flight paths are designed for image capture. However, due to errors in GPS itself and in the image recognition azimuth and impeller stopping angle, a cumulative error deviation occurs between the planned blade imaging route and the actual blade position. This results in blades not appearing in the camera frame during imaging, preventing the drone from adjusting its attitude in time and forcing it to continue flying along the predetermined path. Therefore, we propose an autonomous drone tracking and inspection method for wind turbine blades. Summary of the Invention

[0003] Based on the technical problems existing in the background technology, this invention proposes a method for UAV autonomous tracking and inspection of wind turbine blades. On the basis of achieving coarse positioning of the wind turbine, it relies on simple spatial positional relationships to dynamically track and photograph the blades in real time and perform collision detection in real time, providing a safe, reliable and clear inspection method for UAV to conduct blade inspections without stopping the machine.

[0004] This invention provides the following technical solution: a method for autonomous tracking and inspection of wind turbine blades by unmanned aerial vehicles (UAVs), comprising the following steps:

[0005] S1. Establish a cylindrical enclosure model of the wind turbine and calculate the inspection start point and inspection end point of each shooting surface.

[0006] S2. Perform collision detection on the inspection start point and inspection end point, as well as the line connecting the start point and the end point;

[0007] S3. Calculate the stopping conditions for the drone inspection route;

[0008] S4. Recognize the wind turbine blades in the initial position, adjust the camera position, and initially align with the target;

[0009] S5, drones dynamically track blades;

[0010] S6. Based on the changes in flight vector, update the blade stopping angle in real time, adjust the drone inspection endpoint, update the collision bounding body, and perform collision detection.

[0011] Preferably, in step S1, a geographic coordinate system for the wind turbine hub center is established, and a cylindrical enclosing model is established with a radius of r, consisting of three blade vertices, a tower vertices, and a nacelle vertices.

[0012] A shooting trajectory model was established by translating the leaf tip apex and the shooting starting point based on the shooting angle and shooting distance.

[0013] Preferably, in step S2, the distance between the inspection start point, the inspection end point, the line connecting the inspection path and the central axis of the collision body is calculated, and it is determined whether the distance is greater than the radius of the collision body, so as to determine whether a collision has occurred and whether the inspection path needs to be adjusted.

[0014] Preferably, in step S3, the stopping condition for the UAV inspection route is based on the distance between the inspection start point and the inspection end point, with distance being the priority and time being the next priority.

[0015] Preferably, in step S4, at the initial shooting point, the position of the blade is identified using the drone's vision; the identified blade outline pixels and the image center pixels are used as the adjustment angle for the drone's gimbal angle and direction.

[0016] Preferably, in step S5, at each shooting surface, the UAV generates a direction vector for flying towards the blade tip based on the starting point and intermediate key point information in a speed control manner. Image recognition obtains the center point of the wind turbine blade in real time, and adjusts the direction of the UAV's flight speed vector to achieve dynamic tracking of the blade by the UAV.

[0017] Calculate the blade length and flight path length and distance, and with flight time and flight distance as constraints, end the inspection path and move to the next inspection and shooting area;

[0018] The speed is adjusted within each image recognition cycle, and if the adjustment cycle is exceeded, it returns to the initial speed to continue flying, ensuring that the flight speed vector maintains the initial calculation direction. Local adjustments are made in the direction of image recognition to achieve the purpose of tracking the blades in the drone camera image.

[0019] Preferably, in step S6, the blade stopping angle is updated in real time according to the flight vector change, the UAV inspection endpoint is adjusted, and after the UAV stably tracks the blade, the current blade angle is calculated based on the current stable flight position and the starting inspection point position, and the updated inspection endpoint is calculated from the current blade angle.

[0020] Preferably, the update method for step S6 is as follows:

[0021] leaf_angle_temp=atan((Ptemp.x-Pstart.x) / (Ptemp.y-Pstart.y));

[0022] Where leaf_angle_temp is the updated blade stopping angle, atan is the arctangent function, Ptemp is the current position, Pstart is the starting inspection point position, and x and y represent the x coordinate value and y coordinate value, respectively;

[0023] Based on the coordinate system, the calculated leaf_angle_temp is converted into the leaf angle leaf_angle in the coordinate system. The updated leaf_angle is used to calculate the drone inspection endpoint, update the leaf bounding model, and perform collision detection.

[0024] Update the angles of leaf a, b, and c: leaf_angle_a, leaf_angle_b, leaf_angle_c

[0025] leaf_angle_a=leaf_angle_temp+leaf_angle_a;

[0026] leaf_angle_b=leaf_angle_temp+leaf_angle_b;

[0027] leaf_angle_c=leaf_angle_temp+leaf_angle_c;

[0028] After detecting the blade update angle, update the coordinates of the start / end point of the inspection, update the wind turbine collision enclosure model in step S1, and then use the updated angle to update steps S2-S5 in sequence to complete the entire update process.

[0029] This invention provides a method for autonomous tracking and inspection of wind turbine blades using a drone. With only GPS positioning, the method relies solely on the drone's own camera for visual recognition and motion control to photograph and inspect wind turbine blades. It can also update the blade angle in real time, enabling wind turbine blade inspection without locking the drone. Because the method identifies the blade position in real time during the inspection process, it ensures that the blade is always centered in the captured image, thus ensuring shooting accuracy and overlap rate. Attached Figure Description

[0030] Figure 1 This is a system block diagram of the present invention;

[0031] Figure 2 This is a schematic diagram of the cylindrical enclosure model of the present invention;

[0032] Figure 3 This is a schematic diagram of the blade tracking of the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] This invention provides a technical solution: a method for autonomous tracking and inspection of wind turbine blades by unmanned aerial vehicles (UAVs), comprising the following steps:

[0035] S1. Establish a cylindrical bounding model of the wind turbine and calculate the inspection start and end points for each shooting surface; S2. Perform collision detection on the inspection start and end points, as well as the lines connecting them; S3. Calculate the stopping conditions for the UAV inspection route; S4. Recognize the wind turbine blades at the initial position, adjust the camera position, and initially align with the target; S5. The UAV dynamically tracks the blades; S6. Update the blade stopping angle in real time based on flight vector changes, adjust the UAV inspection end point, update the collision bounding body, and perform collision detection; The system block diagram is as follows: Figure 1 As shown in the figure. This method, based on achieving coarse positioning of the wind turbine, relies on simple spatial relationships to dynamically track and photograph the blades in real time, and performs collision detection in real time, providing a safe, reliable, and clear inspection method for unmanned aerial vehicles (UAVs) to conduct blade inspections without shutting down the turbine.

[0036] Example:

[0037] S1. Establish a cylindrical enclosure model of the wind turbine and calculate the inspection start and end points for each camera; details are as follows:

[0038] like Figure 2 As shown, first, establish the geographic coordinate system Og of the wind turbine hub center, with the hub center as the origin, east as the Xg axis, north as the Yg axis, and vertically upward as the Zg axis;

[0039] Establish a ground coordinate system O at the center of the wind turbine hub, with the center of the hub as the origin, the direction of the wind turbine nacelle as the x-axis, perpendicular to the x-axis, the left side of the front of the wind turbine as the y-axis, and the vertical upward direction of the xy plane as the z-axis direction;

[0040] By identifying key parameters of the wind turbine blades using the drone itself, including the hub position P0[lat, lon, height], azimuth, leaf angle, and leaf length, and using the optimal safe shooting distance of 15 meters (photo_distance), the starting point P1 and ending point P2 of each shooting surface are calculated. The calculation method is as follows:

[0041] The transformation matrix from the hub geographic coordinate system to the hub center ground coordinate system is L(azimuth), and the transformation variable is the azimuth angle;

[0042] Based on the three blade angles of 120 degrees, and knowing the angle of one blade in the hub center coordinate system O, the angles of the other two blades can also be calculated.

[0043] The position of the leaf tip can be calculated based on the leaf length and leaf angle.

[0044]

[0045] Where P0 is the hub center position, leaf_angle_a is the stopping angle of blade a obtained from image recognition, leaf_angle_b is the stopping angle of blade b obtained from image recognition, leaf_lengh is the blade length, and L[leaf_angle_a], L[leaf_angle_b], and L[leaf_angle_c] are the transformation matrices from the hub geographic coordinate system to the hub center coordinate system.

[0046] Based on the tower height (tower_height), the tower position is Pt[tower_distance, 0, 0]. T `tower_distance` represents the distance from the center of the tower to the center of the hub. The tower endpoint Pj[xyz] is calculated. T

[0047] Pj = P0 + Pt + [0 0 tower_height] T (Formula 2)

[0048] The endpoint Pe[xyz] of the computer cabin is calculated based on the cabin length, engine_length. T :

[0049] Pe = P0 + [engine_length 0 0]T (Formula 3)

[0050] Where engine_length is the length of the engine compartment.

[0051] A cylindrical enclosure model is established with a radius of r, consisting of three blade vertices, a tower vertices, and a nacelle vertices.

[0052] A shooting flight path model was established by translating the leaf tip apex and the shooting starting point based on the shooting angle and shooting distance;

[0053] S2. Perform collision detection on the inspection start point and inspection end point, as well as the line connecting the start point and the end point;

[0054] Collision detection: Calculate the distance from the inspection start point Pstart and the inspection end point Pend to the central axis of the cylinder surrounding the blade, and calculate the distance from the line connecting the inspection start point and the inspection end point to the central axis of the cylinder surrounding the blade.

[0055] The mathematical representation of collision detection between the inspection path and the cylindrical enclosure model of the wind turbine can be simplified as follows:

[0056] Calculate the distance between the inspection start point, inspection end point, and the line connecting the inspection path to the centerline of the collision body, determine whether the distance is greater than the radius of the collision body, and thus determine whether a collision has occurred and whether the inspection path needs to be adjusted.

[0057] The calculation method is as follows:

[0058] 1. Calculate the distance between the point and the central axis of the colliding body;

[0059] (1) Calculate the distance between the point and the vertex of the central axis;

[0060] (2) Calculate the minimum distance between the point and the central axis;

[0061] (3) Calculate the minimum distance between the inspection path and the centerline, that is, the minimum distance between the two line segments;

[0062] (4) Select the minimum distance and compare the radii of the colliding bodies;

[0063] (5) If the calculated minimum distance is less than the radius of the colliding body, then the reverse extension line between the calculated point and the central axis is extended along the reverse extension line to a distance greater than the radius of the colliding body.

[0064] 2. Determine the minimum collision distance between the current point and the three blades, nacelle, and tower. When the distance between the point and one or more of the colliding bodies is less than the radius of the colliding body, continue to extend the line from the intersection of the extension lines of multiple colliding bodies and the point to avoid the colliding body.

[0065] S3. Calculate the stopping conditions for the drone inspection route;

[0066] Based on the distance between the inspection start point and the inspection end point, prioritize distance, then time.

[0067] Calculate the straight-line distance L from the inspection start point to the inspection end point, and obtain the theoretical flight time T using the flight speed V of the drone when taking pictures during inspection.

[0068] When calculating, the distance Lt between the current position and the initial position of the UAV is taken first. When Lt-L>0.01, the UAV is considered to have reached the destination and the inspection path is stopped. If the condition is not met, but the flight time has exceeded T, the leaf tip is identified by image recognition. If the leaf tip is not detected, the inspection path is stopped.

[0069] S4. Recognize the wind turbine blades in the initial position, adjust the camera position, and initially align with the target;

[0070] At the initial shooting point, the drone's visual recognition is used to identify the blade position; based on the identified blade outline pixel point P[x1,y1] and the image center pixel point P0[x0,y0], the adjustment angles for the drone's gimbal angles pitch and yaw are determined, and the calculation method is as follows:

[0071]

[0072]

[0073] Kpitch_y, Kpitch_z, Kyaw_y, and Kyaw_z are proportional parameters set based on the current angle of the blades and the shooting angle. After adjusting the gimbal angle to align the blades with the target center, you can proceed to the next step.

[0074] S5, drone dynamic tracking of blades

[0075] At each shooting location, the drone generates a direction vector for flight towards the blade tip based on the starting point and intermediate key points, using speed control. Image recognition acquires the center point of the wind turbine blade in real time, adjusting the drone's flight speed vector direction to achieve dynamic tracking of the blade. The drone calculates the blade length and flight path distance, and, constrained by flight time and distance, terminates the inspection route and moves to the next shooting location. Figure 3 As shown;

[0076] The initial flight speed for each shooting surface is calculated as follows;

[0077]

[0078] Where p1[x,y,z] is the starting point of the control flight path, p2[x,y,z] is the ending point of the flight path; Lxyz is the distance between the two points; sqrt represents the square root formula; V is the vector velocity of the aircraft, and Vx, Vy, Vz are the velocity components of the aircraft in the x, y, and z directions in the collective coordinate system.

[0079] During flight, the image captures the real-time pixel coordinates of the blade center point, Pleaf[x,y]. Using the PD control method, the difference between the identified pixel coordinates and the image center P0[x,y] is used as the input to adjust the UAV's flight parameters. Based on the initial speed, the speed is increased or decreased to adjust the direction of the flight speed vector.

[0080]

[0081] Where deltaVy and deltaVz are the calculated velocity increments, Kvy_x and Kvy_y are the target displacement gain values ​​in the image coordinate system corresponding to the flight velocity y direction; Kvz_x and Kvz_y are the target displacement gain values ​​in the image coordinate system corresponding to the flight velocity z direction; Pimage_center is the coordinate of the image center point in the image coordinate system; and leaf_length is the blade length.

[0082] The speed is adjusted within each image recognition cycle, and if the adjustment cycle is exceeded, it returns to the initial speed to continue flying, ensuring that the flight speed vector maintains the initial calculation direction. Local adjustments are made in the direction of image recognition to achieve the purpose of tracking the blades in the drone camera image.

[0083] S6. Based on changes in flight vectors, update the blade stopping angle in real time, adjust the drone inspection endpoint, and perform collision detection.

[0084] After the UAV stably tracks the blade, it calculates the current blade angle based on the current stable flight position Ptemp and the starting inspection point position Pstart, and then calculates the updated inspection endpoint based on the current blade angle. The update method is as follows:

[0085] leaf_angle_temp=atan((Ptemp.x-Pstart.x) / (Ptemp.y-Pstart.y))(Formula 8)

[0086] Where leaf_angle_temp is the updated blade stopping angle, and atan is the arctangent function.

[0087] Based on the coordinate system, the calculated leaf_angle_temp is converted into the leaf angle leaf_angle in the coordinate system. The updated leaf_angle is then used to calculate the drone inspection endpoint, update the leaf enclosure model, and perform collision detection.

[0088] Update the angles of leaf a, b, and c: leaf_angle_a, leaf_angle_b, leaf_angle_c

[0089]

[0090] After detecting the updated blade angle, leaf_angle_a, leaf_angle_b, and leaf_angle_c are substituted into Formula 1 to update the coordinates of the inspection start / end point and update the wind turbine collision bounding model. Then, using the updated angle, the process iterates through Formulas 2 to 7 to complete the entire update process. This update enables wind turbine blade inspection and photography, and allows for real-time updates of the blade angle, enabling wind turbine blade inspection without locking the machine.

[0091] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for autonomous tracking and inspection of wind turbine blades using unmanned aerial vehicles (UAVs), characterized in that: Includes the following steps: S1. Establish a cylindrical enclosure model of the wind turbine and calculate the inspection start point and inspection end point of each shooting surface. S2. Perform collision detection on the inspection start point and inspection end point, as well as the line connecting the start point and the end point; S3. Calculate the stopping conditions for the drone inspection route; S4. Recognize the wind turbine blades in the initial position, adjust the camera position, and initially align with the target; S5, drones dynamically track blades; S6. Based on the changes in flight vector, update the blade stopping angle in real time, adjust the UAV inspection endpoint, update the collision bounding body, and perform collision detection. In step S1, a geographic coordinate system for the wind turbine hub center is established, and a cylindrical enclosed model is established with the three blade vertices, the tower vertices, and the nacelle vertices enclosing a radius of r. A shooting flight path model was established by translating the leaf tip apex and the shooting starting point based on the shooting angle and shooting distance; In step S2, the inspection start point, inspection end point, and the distance between the inspection path line and the central axis of the collision body are calculated. It is then determined whether the distance is greater than the radius of the collision body to determine whether a collision has occurred and whether the inspection path needs to be adjusted. In step S3, the stopping condition for the UAV inspection route is based on the distance between the inspection start point and the inspection end point, with distance being the priority and time being the next priority. In step S4, at the initial shooting point, the position of the blade is identified using the drone's vision; the identified blade outline pixels and the center pixel of the image are used as the adjustment angle for the drone's gimbal angle and direction. In step S5, at each shooting surface, the UAV generates a direction vector for flying towards the blade tip based on the starting point and intermediate key point information in a speed control manner. Image recognition obtains the center point of the wind turbine blade in real time and adjusts the direction of the UAV's flight speed vector to achieve dynamic tracking of the blade by the UAV. Calculate the blade length and flight path length and distance, and with flight time and flight distance as constraints, end the inspection path and move to the next inspection and shooting area; The speed is adjusted within each image recognition cycle, and if the adjustment cycle is exceeded, it returns to the initial speed to continue flying, ensuring that the flight speed vector maintains the initial calculation direction. Local adjustments are made in the direction of image recognition to achieve the purpose of tracking the blades in the drone camera image.

2. The method for autonomous tracking and inspection of wind turbine blades by an unmanned aerial vehicle (UAV) according to claim 1, characterized in that: In step S6, the blade stopping angle is updated in real time according to the flight vector change, the inspection endpoint of the UAV is adjusted, and collision detection is performed. After the UAV stably tracks the blade, the current blade angle is calculated based on the current stable flight position and the starting inspection point position, and the updated inspection endpoint is calculated from the current blade angle.

3. The method for autonomous tracking and inspection of wind turbine blades by an unmanned aerial vehicle (UAV) according to claim 2, characterized in that: The update method for step S6 is as follows: ； Where leaf_angle_temp is the updated blade stopping angle, atan is the arctangent function, Ptemp is the current position, Pstart is the starting inspection point position, and x and y represent the x coordinate value and y coordinate value, respectively; Based on the coordinate system, the calculated leaf_angle_temp is converted into the leaf angle leaf_angle in the coordinate system. The updated leaf_angle is used to calculate the drone inspection endpoint, update the leaf bounding model, and perform collision detection. Update the angles of leaf a, b, and c: leaf_angle_a, leaf_angle_b, leaf_angle_c ； After detecting the blade update angle, update the coordinates of the start / end point of the inspection, update the wind turbine collision enclosure model in step S1, and then use the updated angle to update steps S2-S5 in sequence to complete the entire update process.

Images