A method for the clearance of a wind turbine based on machine vision ground wind measurement

Abstract

The present invention provides a method for measuring the clearance of a wind turbine based on machine vision on the ground, which includes formulating a calibration table for the pixel scale of a camera, finding the tip projection points, starting to shoot, and calculating the clearance result. The method for measuring the clearance of a wind turbine based on machine vision on the ground provided by the present invention adopts a ground measurement clearance scheme, which is different from the existing long-distance vertical measurement. It obtains the images of the tower barrel and blades directly below the nacelle through machine vision means, and through image processing means, obtains the equivalent tip points of the blades in each frame of image, as well as the tip projection points found by a specific method. Referring to the calibration table of the pixel scale, the blade clearance result is calculated.

Description

Technical Field

[0001] The present invention relates to the field of wind turbine monitoring, and particularly to a method for measuring the clearance of a wind turbine based on machine vision on the ground. Background Art

[0002] The blades of wind turbines in the wind power industry are developing towards longer, softer and lighter trends. With the development of onshore wind turbines, the proportion of wind farm construction under conditions such as high temperature and low temperature, high altitude, high humidity and complex terrain that exceed the standard design environment of the original wind turbines is increasing day by day, and the blade clearance problem has become increasingly serious. The blade clearance actually refers to the minimum clearance between the blade tip and the tower surface. The GL certification specification requires that during the operation of the unit, this minimum clearance shall not be less than 30% of the blade in the undeformed state. Blade clearance is very important in the development and certification stage of wind turbines, and is related to the stable and safe operation of the unit.

[0003] Therefore, the detection and monitoring of blade clearance have become increasingly important. At present, there are nacelle clearance measurement schemes and ground clearance measurement schemes. The nacelle clearance measurement scheme installs clearance equipment in the nacelle and calculates the clearance result through image processing technology or laser technology. The ground clearance measurement scheme generally takes a video of the impeller rotation perpendicular to the nacelle axis at a distance of several hundred meters from the tower and calculates the clearance.

[0004] Due to the difficulty in installing and implementing the hanging equipment in the nacelle clearance measurement scheme, which involves a huge amount of work, such as the nacelle transformation scheme, process and the development and testing of the supporting main control program, and being affected by environmental factors such as the movement, vibration and electromagnetic interference of the unit, the data measured by the current clearance test means is unstable and the accuracy of the clearance data needs to be verified by more accurate means. The ground clearance measurement scheme will be affected by the camera elevation angle and the deviation of the perpendicularity to the nacelle axis, resulting in errors, and it is not easy to solve. Summary of the Invention

[0005] In order to solve the deficiencies of the prior art, the present invention provides a method for measuring the clearance of a wind turbine based on machine vision on the ground. Adopting the ground measurement clearance scheme, different from the existing long-distance vertical measurement, it obtains the tower and blade images directly below the nacelle through machine vision means, and through image processing means, obtains the equivalent points of the blade tips in each frame of image, as well as the blade tip projection points found by a specific method, and calculates the blade clearance result with reference to the pixel scale calibration table.

[0006] The technical solution adopted by the present invention to solve its technical problems is: providing a method for measuring the clearance of a wind turbine based on machine vision on the ground, including the following steps:

[0007] S1. Develop a calibration table for the camera pixel scale: Use the camera to photograph a marker with a test length of L, and obtain a relationship table of the calculated pixel scale S when different imaging scaling ratios R are used at different distances D. The pixel scale S is the ratio of the test length to the number of imaging pixels.

[0008] S2. Obtain the camera installation position: Place the camera at a certain distance from the tower barrel at the bottom of the wind turbine tower, and make the camera shoot vertically upward. Image the impeller to ensure that the reference line of the impeller end face is parallel to the reference line of the horizontal axis of the ground camera image. If not parallel, rotate the camera direction for adjustment, and then translate the camera left and right to ensure that the impeller is in the center position of the left and right of the image.

[0009] S3. Find the tip projection point: When the blade rotates to the position where the tip is at the lowest point, the projection point of the tip on the tower barrel is point A. Place a laser rangefinder at any point D on the straight line formed by the ground camera and the tower bottom. Calculate the tip height h based on the distance d of the laser rangefinder from the tower barrel OD and the known tower height and blade length. Calculate the distance d between point A and point D through the Pythagorean theorem OA , and make the laser rangefinder project a light spot onto point A. AD S4. Start shooting: The camera starts shooting the rotating impeller image, record the position of point A in the starting image of the video, and then remove the laser rangefinder.

[0010] S5. Calculate the clearance result: For each frame of the video processed by image processing means, obtain the equivalent point pixels of the blade tip. Calculate the measured distance CL through the following formula:

[0011] CL = N

[0012] *S pix *S OA

[0013] where N pix is the number of pixels between the equivalent point of the blade tip and point A, and S OA is the pixel scale of the height of point A from the ground queried in the camera pixel scale calibration table.

[0014] Take the minimum value before and after each blade passes through the tower barrel as the result of this clearance recognition, or calculate the distance from point A to the fitted straight line by linearly fitting the trajectory of the tip equivalent points as the result of this clearance recognition.

[0015] Before step S1, first estimate whether enough clearance can be in the shooting frame according to the maximum opening angle of the camera.

[0016] In step S4, for the images of the captured video, the auxiliary line of the impeller end face is compared with the image horizontal line in real time to ensure the parallel relationship between the heading of the unit and the axis of the camera. If the hub end face tilts during the shooting, there is a deviation from the image reference line and exceeds a certain range, it indicates that the unit has a yaw action. At this time, the measurement is stopped and the new yaw position is tracked.

[0017] The method for obtaining the equivalent point pixels of the blade tip described in step S5 adopts image processing means or machine learning target recognition methods. The image processing means specifically adopts the frame difference method in OpenCV, and the machine learning target recognition method specifically adopts the edge detection contour recognition method in YOLO.

[0018] The beneficial effects of the present invention based on its technical solution are as follows:

[0019] A method for measuring the clearance of a wind turbine based on machine vision on the ground provided by the present invention utilizes the natural convenience of the blade clearance detection means implemented on the ground. The blade clearance result can be conveniently and effectively measured in a fixed coordinate system, ensuring high resolution and accuracy, and being convenient and easy to operate, and can be reused in different units and environments. At the same time, the image recognition technology can obtain accurate ranging results, and the recognition accuracy is at the centimeter level, so both the accuracy and resolution are very high. The method for measuring the clearance of a wind turbine based on machine vision on the ground provided by the present invention can solve technical problems such as the selection of the impeller projection point, the acquisition of the pixel scale, and the acquisition of the tip projection point, so as to conveniently, effectively and accurately obtain the clearance result of the unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a schematic diagram of the principle of ground measurement of clearance.

[0021] Figure 2 is a schematic diagram of the principle of formulating the camera pixel scale calibration table.

[0022] Figure 3 is a schematic diagram of obtaining the camera installation position.

[0023] Figure 4 is a schematic diagram of finding the tip projection point.

[0024] Figure 5 is a schematic diagram of a frame of image.

[0025] In the figure: 1 - hub end face, 2 - camera horizontal reference line. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The present invention will be further described below with reference to the drawings and embodiments.

[0027] Principle description:

[0028] Refer toFigure 1 , by placing the camera directly below the bottom of the tower and shooting the video of the blade operation upward, and by calculating the number of pixels from the marked tip projection point A to the tip point P, referring to the pixel scale calibration table, the real-time clearance result can be obtained. Point O represents the intersection of the tower barrel and the ground, point A represents the projection point of the tip point on the tower barrel, point B represents the point where the tower barrel enters the picture, point C represents the equivalent point of the camera position, point E represents the projection point of point C at the tip height, point F represents the point where the sky enters the picture, and point P represents the tip point.

[0029] First, estimate whether enough clearance can be made to be within the shooting frame according to the maximum opening angle of the camera. Taking the maximum opening angle α of the ground camera as 60° as an example, the distance d AP is estimated according to 20m, and the height h OA is estimated according to 20m. If point C can be placed 10m away from the tower barrel, then:

[0030] d OC = 10m,

[0031] h OB = d OC * tan60° = 17.32m,

[0032] d AE = d OC = 10m,

[0033] d EF = h OA = 20 * tan30° = 11.55m,

[0034] d AF = d AE + d EF = 21.55m.

[0035] In this way, enough clearance for measurement can be within the picture.

[0036] Based on the above principle, the present invention provides a method for measuring the clearance of a wind turbine based on machine vision on the ground, including the following steps:

[0037] S1. Formulate a pixel scale calibration table for the camera: Refer to Figure 2 , use the camera to shoot a marker with a test length of L, and obtain a relationship table of the pixel scale S calculated when different imaging scaling ratios R are used at different distances D, where the pixel scale S is the ratio of the test length to the number of imaging pixels.

[0038] The pixel scale calibration table used in this embodiment is shown in the following table:

[0039]

[0040] Table 1 Camera Pixel Scale Calibration Table

[0041] S2. Obtain the camera installation position: Place the ground camera used for shooting videos at an appropriate position at the bottom of the wind turbine tower, ensuring that the device shoots vertically upward. Through the imaging of the impeller, ensure that the reference line 1 of the impeller end face is parallel to the reference line 2 of the horizontal axis of the ground camera image. If not parallel, rotate the device to adjust. Then move the camera left and right to ensure that the impeller is centered horizontally in the image.

[0042] S3. Find the tip projection point: When the blade rotates to the position where the tip is at the lowest point, the projection point of the tip on the tower barrel is point A. Place a laser rangefinder at any point D on the straight line formed by the ground camera and the tower bottom. Measure the distance d from the laser rangefinder to the tower barrel OD and calculate the tip height h based on the known tower height and blade length OA , calculate the distance d between point A and point D through the Pythagorean theorem AD , and make the laser rangefinder project a light spot onto point A;

[0043] S4. Start shooting: Refer to Figure 4 , place the camera at the tower bottom and shoot vertically upward facing the root of the blade. Record the position of point A in the starting frame of the video, and then remove the laser rangefinder. Refer to Figure 3 , (refer to the above text), ensure that the impeller end face is parallel to the horizontal axis of the ground camera image, which means the camera is shooting vertically upward facing the root of the blade; if not parallel, move the camera to adjust; if the reference line 1 of the hub end face is tilted and deviated from the image reference line 2 during shooting, it indicates that the unit has a yaw action. At this time, stop the measurement and track the new yaw position.

[0044] S5. Calculate the clearance result: Refer to Figure 5 , for each frame of the video processed by image processing means, obtain the equivalent point pixels of the blade tip in it; calculate the measured distance CL through the following formula:

[0045] CL = N pix *S OA

[0046] where N pix is the number of pixels between the equivalent point (point P) of the blade tip and point A, and S OA is the pixel scale of the height of point A from the ground queried in the camera pixel scale calibration table;

[0047] Take the minimum value of each blade before and after passing the tower barrel as the clearance recognition result of this time.

[0048] A method for measuring the clearance of a wind turbine based on machine vision on the ground provided by the present invention adopts a ground measurement clearance scheme, which is different from the existing long-distance vertical measurement. It obtains the tower barrel and blade images directly below the nacelle through machine vision means, obtains the equivalent tip points of the blades in each frame of image through image processing means, and finds the tip projection points through a specific method. Referring to the pixel scale calibration table, the blade clearance result is calculated.

Claims

1. A method for the clearance of a wind turbine based on machine vision for ground wind measurement, characterized in that It includes the following steps: S1. Develop a calibration table for the camera pixel scale: Use the camera to photograph a marker with a test length of L, and obtain a relationship table of the pixel scale S calculated when different imaging scaling ratios R are used at different distances D, where the pixel scale S is the ratio of the test length to the number of imaging pixels; S2. Obtain the installation position of the camera: Place the camera at a certain distance from the tower barrel at the bottom of the wind turbine tower, and make the camera shoot vertically upward. Through the imaging of the impeller, ensure that the reference line of the impeller end face is parallel to the reference line of the horizontal axis of the ground camera image. If it is not parallel, rotate the camera direction for adjustment, and then translate the camera left and right to ensure that the impeller is in the center position of the left and right of the image; S3. Find the tip projection point: When the blade rotates to the position where the tip is at the lowest point, the projection point of the tip on the tower barrel is point A. Place a laser rangefinder at any point D on the straight line formed by the ground camera and the tower base. Measure the distance d from the laser rangefinder to the tower barrel OD and calculate the tip height h based on the known tower height and blade length OA , and calculate the distance d between point A and point D through the Pythagorean theorem AD , and make the laser rangefinder project a light spot onto point A; S4. Start shooting: The camera starts shooting the rotating picture of the impeller, record the position of point A in the starting picture of the video, and then remove the laser rangefinder; S5. Calculate the clearance result: For each frame of the video processed by image processing means, obtain the equivalent point pixels of the blade tip; calculate the measured distance CL through the following formula: CL = N pix *S OA where N pix is the number of pixels between the equivalent point of the blade tip and point A, and S OA is the pixel scale of the height of point A from the ground queried in the camera pixel scale calibration table; Take the minimum value before and after each blade passes through the tower barrel as the clearance recognition result of this time, or fit the trajectory of the equivalent points of the blade tip by a straight line, and then calculate the distance from point A to the fitted straight line as the clearance recognition result of this time.

2. The method for the clearance of a wind turbine based on machine vision ground wind measurement according to claim 1, wherein: Before step S1, first estimate whether enough clearance can be made to be in the shooting picture according to the maximum opening angle of the camera.

3. The method for the clearance of a wind turbine based on machine vision ground wind measurement according to claim 1, wherein: In step S4, for the images of the captured video, compare the auxiliary line of the impeller end face with the horizontal line of the image in real time to ensure the parallel relationship between the heading of the unit and the axis of the camera; if the hub end face tilts during shooting, there is a deviation from the image reference line and exceeds a certain range, it indicates that the unit has a yaw action. At this time, stop the measurement and track the new yaw position.

4. The method for clearing the airspace of a wind turbine based on machine vision according to any one of claims 1 or 2, characterized in that: The method for obtaining the equivalent point pixels of the blade tip described in step S5 adopts image processing means or machine learning target recognition method. The image processing means specifically adopts the frame difference method in OpenCV, and the machine learning target recognition method specifically adopts the edge detection contour recognition method in yolo.

Images