Offshore wind power blade angular velocity acquisition method and system based on key points

Abstract

The invention discloses a key point-based offshore wind power blade angular velocity acquisition method and system. The method comprises the following steps of: acquiring an offshore wind power blade image in real time and preprocessing the offshore wind power blade image; performing key point identification on the acquired offshore wind power blade images at the first moment and the second moment by adopting a key point detection algorithm; generating a first vector and a second vector by taking a cabin center point identified in the offshore wind power blade images of any selected blade at the first moment and the second moment as an original point and a blade tip as an end point; calculating an angle difference value between the first vector and the second vector, and dividing the sum of the angle difference value and an angle difference compensation value by a time interval between the first moment and the second moment to obtain an angular velocity sequence of the offshore wind power blade; and comparing the angular velocity of the offshore wind power blade obtained by the sensor with the angular velocity sequence of the offshore wind power blade, and determining the finally output angular velocity of the offshore wind power blade. According to the method, the angular velocity of the offshore wind power blade with relatively high precision and relatively high real-time performance can be obtained.

Description

Technical Field

[0002] The present application relates to the field of new energy and wind power technology, and specifically to a method and system for acquiring the angular velocity of offshore wind turbine blades based on key points. Background Art

[0003] With the increasing attention paid to renewable energy, offshore wind power generation has become an important part of clean energy. In the operation and maintenance of wind power equipment, it is crucial to accurately identify the angular velocity of wind turbine blades, which not only affects the operating efficiency of the wind turbine, but also affects the safety and service life of wind power equipment.

[0004] Traditional wind turbine blade angular velocity recognition methods are mainly based on physical sensors or image processing technology. Physical sensors have high installation and maintenance costs and are easily affected by the harsh marine environment, resulting in errors in measurement accuracy. Image processing-based technologies rely on high-quality image acquisition and complex image processing algorithms, and have high requirements for image acquisition equipment and processing platforms. Therefore, the existing technology has certain limitations and shortcomings.

[0005] At present, the methods for obtaining the angular velocity of offshore wind turbine blades have problems such as low accuracy and poor real-time performance, which makes it difficult to meet the requirements of efficient and safe operation of modern wind power equipment. Therefore, it is urgent to develop a method for obtaining the angular velocity of offshore wind turbine blades with high accuracy and strong real-time performance. Summary of the invention

[0006] In order to solve the above problems, the present application provides a method and system for acquiring the angular velocity of offshore wind turbine blades based on key points.

[0007] In a first aspect, the present application provides a method for obtaining the angular velocity of an offshore wind turbine blade based on key points, comprising: Real-time acquisition and preprocessing of offshore wind turbine blade images; A key point detection algorithm is used to identify key points of the offshore wind turbine blade images collected at the first moment and the offshore wind turbine blade images collected at the second moment; the key points include: the center point of the nacelle, the tip of the blade, and the connection point between the nacelle and the blade; the time interval between the first moment and the second moment is less than the preset time interval; the key point detection algorithm uses the yolov8-pose model as the basic model; A first vector is generated by taking the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin and the blade tip of any blade identified in the offshore wind turbine blade image at the first moment as the end point; a second vector is generated by taking the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin and the blade tip of the same blade identified in the offshore wind turbine blade image at the second moment as the end point; the same blade is determined by matching according to a key point matching algorithm; The angle difference between the first vector and the second vector is calculated, and the offshore wind turbine blade angular velocity sequence is calculated by dividing the sum of the angle difference and the angle difference compensation value by the time interval between the first moment and the second moment; the angle compensation value is n times 360 degrees, and n is an integer greater than 0; the offshore wind turbine blade angular velocity sequence is compared with the offshore wind turbine blade angular velocity sequence obtained by the sensor at the first moment or the second moment, and the offshore wind turbine blade angular velocity closest to the offshore wind turbine blade angular velocity collected by the sensor is selected from the offshore wind turbine blade angular velocity sequence as the offshore wind turbine blade angular velocity finally output.

[0008] By adopting the above scheme, the yolov8-pose model is selected to detect the key points of offshore wind turbine blades more accurately; the key points of each blade are matched based on the detected key points, and the angular velocity sequence is calculated according to the positional relationship of the identified key points. The angular velocity collected by the single sensor technology is combined to obtain a more accurate angular velocity from the angular velocity sequence.

[0009] Preferably, it also includes: Using sensors to collect environmental data in real time; the environmental data includes: wind direction, wind speed and wind force; The environmental data collected during the period from the first moment to the second moment are respectively input into the first neural network model, and the result of whether the offshore wind turbine blade is deformed during the period from the first moment to the second moment is output; the input of the first neural network model is the environmental data, and the output is whether the offshore wind turbine blade is deformed, which is generated by training through historical environmental data and the result of whether the offshore blade is deformed obtained at the corresponding moment; If the offshore wind turbine blade is deformed at any time between the first moment and the second moment, the similarity threshold of the key points in the key point matching algorithm is correspondingly lowered, and the same blade at the first moment and the second moment is matched based on the adjusted key point matching algorithm.

[0010] By adopting the above scheme, taking into account that the blades will undergo certain deformation under specific environmental conditions, the neural network model is used to determine whether deformation occurs in the current period. The similarity threshold of key point matching is adjusted accordingly for the deformation situation, so as to ensure that more matching key points are obtained as much as possible, and prevent the inability to obtain key point matching results due to deformation problems.

[0011] Preferably, it also includes: Taking the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin and the blade tip of each remaining blade not selected identified in the offshore wind turbine blade image at the first moment as the end point, a first vector corresponding to the blade is generated; taking the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin and the blade tip of the blade identical to each remaining blade not selected identified in the offshore wind turbine blade image at the second moment as the end point, a second vector corresponding to the blade is generated; The angle difference calculated by the first vector and the second vector of the corresponding leaf obtained for each leaf is compared. If any two calculated angle differences are greater than the preset angle difference, the key point types and quantities are increased to optimize the training of the key point detection algorithm. The key points are re-recognized according to the optimized key point detection algorithm, and the same leaf is re-determined based on the re-recognized key points.

[0012] By adopting the above scheme, considering that incorrect key point matching will lead to incorrect angle difference calculation, the angle difference corresponding to the angular velocity of the remaining blades is calculated to verify whether there is an error in the angle difference obtained corresponding to the selected blade calculation, and then the key point matching is re-performed according to the verification result to ensure that the accurate angular velocity is obtained.

[0013] Preferably, it also includes: When the angular velocity of the offshore wind turbine blade cannot be obtained in real time according to the sensor, the environmental data collected at the first moment or the second moment and the historical blade rotation speed at a certain interval from the first moment or the second moment are input into the second neural network model, and the predicted offshore wind turbine blade rotation speed at the first moment or the second moment is output; the input of the second neural network model is the environmental data at the current moment and the historical blade rotation speed at a certain interval from the current moment, and the output is the offshore wind turbine blade rotation speed at the current moment, which is generated by training the historical offshore wind turbine blade environmental data, the blade rotation speed at a certain interval from the time when the historical offshore wind turbine blade environmental data is collected, and the rotation speed at the time when the historical offshore wind turbine blade environmental data is collected; The offshore wind turbine blade angular velocity is calculated based on the predicted offshore wind turbine blade rotation speed, and the calculated offshore wind turbine blade angular velocity is compared with the offshore wind turbine blade angular velocity sequence.

[0014] By adopting the above scheme, considering that the angular velocity cannot be obtained according to the sensor, the rotation speed is generated according to the historical environmental data and the blade rotation speed prediction, and then the angular velocity of the offshore wind turbine blade is obtained to assist in determining the angular velocity of the offshore wind turbine blade in the offshore wind turbine blade angular velocity sequence.

[0015] Preferably, it also includes: Determine whether the offshore wind turbine blade angular velocity finally outputted is greater than a preset offshore wind turbine blade angular velocity threshold; if greater, use the offshore wind turbine blade angular velocity collected by the backup sensor at the first moment or the second moment to replace the offshore wind turbine blade angular velocity collected by the original sensor to compare the offshore wind turbine blade angular velocity sequence.

[0016] By adopting the above scheme, considering that when the offshore wind force and wind speed reach certain values, the offshore wind turbine blade control system will control the offshore wind turbine blades to operate at a certain angular velocity to avoid blade damage, there is a corresponding angular velocity threshold. Once the angular velocity obtained by the sensor is greater than the angular velocity threshold, it indicates that the current sensor is damaged, and the corresponding value collected by the backup sensor is used for auxiliary judgment.

[0017] Preferably, it also includes: Obtain the final output offshore wind turbine blade angular velocity sequence, monitor the moment when the absolute value of the angular velocity change value in the offshore wind turbine blade angular velocity sequence is greater than a preset angular velocity change threshold, record it as a distortion moment, obtain the corresponding environmental data at the distortion moment and analyze it; If the number of types of environmental data that meet the first condition is greater than the preset number, the angular velocity of the offshore wind turbine blades collected by the backup sensor at the distortion moment is used to replace the angular velocity of the offshore wind turbine blades collected by the original sensor to compare the angular velocity sequence of the offshore wind turbine blades; the first condition is: the absolute value of the change value of a type of environmental data at the distortion moment compared with the corresponding type of environmental data at the moment before the distortion moment is greater than the corresponding preset environmental data change threshold.

[0018] By adopting the above scheme, considering that the angular velocity change of offshore wind turbine blades mainly depends on the change of wind speed and wind force in the environmental data, once the moment when the angular velocity changes greatly corresponds to and a huge change in the environmental data occurs, it is verified that there is a calculation error in the current angular velocity, indicating that the current sensor is damaged, and the corresponding value collected by the backup sensor is used for auxiliary judgment.

[0019] Preferably, Pose as CompositionalTokens is added to the selected yolov8-pose model network architecture, and DCNv3 is used to replace the convolutional layer.

[0020] By adopting the above solution, preferably using Pose as Compositional Tokens and DCNv3 to improve the accuracy of the model, due to the introduction of these technologies, the model has higher performance and stability in key point detection, further improving the accuracy of speed recognition.

[0021] In a second aspect, the present application provides an offshore wind turbine blade angular velocity system based on key points, comprising: Image acquisition module, used for real-time acquisition and preprocessing of offshore wind turbine blade images; A key point detection module is used to identify key points of the first moment offshore wind turbine blade image and the second moment offshore wind turbine blade image respectively using a key point detection algorithm; the key points include: the center point of the cabin, the blade tip, and the connection point between the cabin and the blade; the key point detection algorithm uses the yolov8-pose model; A vector acquisition module is used to generate a first vector with the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin and the blade tip of any blade identified in the offshore wind turbine blade image at the first moment as the end point; and to generate a second vector with the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin and the blade tip of the same blade identified in the offshore wind turbine blade image at the second moment as the end point; the same blade is determined by matching according to a key point matching algorithm; The angular velocity calculation module is used to calculate the angle difference between the first vector and the second vector, and obtain the offshore wind turbine blade angular velocity sequence by dividing the sum of the angle difference and the angle difference compensation value by the time interval between the first moment and the second moment; the angle compensation value is n times 360 degrees, and n is an integer greater than 0; the offshore wind turbine blade angular velocity sequence is compared with the offshore wind turbine blade angular velocity obtained by the sensor, and the offshore wind turbine blade angular velocity that is closest to the offshore wind turbine blade angular velocity collected by the sensor is selected from the offshore wind turbine blade angular velocity sequence.

[0022] By adopting the above scheme, the key points of the offshore wind turbine blades are detected through the key point detection algorithm, the angle between the blades within a certain time interval is calculated based on these key points, and then the angular velocity of the blades is calculated based on the change in the angle, thereby improving the accuracy and efficiency of angular velocity recognition.

[0023] In a third aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method as described above.

[0024] In a fourth aspect, the present application provides a computer device, comprising a memory, a processor, and a program stored and executable on the memory, wherein the program implements the steps of the above method when executed by the processor.

[0025] In summary, this application has the following beneficial effects: 1. Combining image detection technology and sensor detection technology, the yolov8-pose model is used to detect the key points of offshore wind turbine blades more accurately; the angular velocity sequence is calculated based on the positional relationship of the identified and matched key points, and the angular velocity collected by the single sensor technology is combined to obtain a more accurate angular velocity from the angular velocity sequence to improve the accuracy of angular velocity recognition; 2. Combined with the collected environmental data, the key point matching is judged and optimized to further improve the accuracy of the final calculated angular velocity; 3. Use the calculated angle differences of each blade to verify each other, and further optimize the key points according to the verification results to improve the accuracy of the final calculated angular velocity. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of the method for acquiring the angular velocity of offshore wind turbine blades based on key points described in a specific embodiment; Figure 2 This is a diagram of identifying key points of offshore wind turbine blades using the key point-based offshore wind turbine blade angular velocity acquisition method in a specific embodiment; Figure 3 A diagram of the position relationship of key points of offshore wind turbine blades using the key point-based offshore wind turbine blade angular velocity acquisition method in a specific embodiment; Figure 4 It is a structural schematic diagram of the key point-based offshore wind turbine blade angular velocity acquisition system described in a specific embodiment. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0028] like Figure 1 As shown, the embodiment of the present application discloses a method for obtaining the angular velocity of an offshore wind turbine blade based on key points, which specifically includes: S1. Real-time acquisition and preprocessing of offshore wind turbine blade images.

[0029] Specifically, a drone is used to take real-time photos of offshore wind turbine blades within a preset shooting distance. In this embodiment, the preset shooting distance is set to within 80 meters of the center of the diameter of the circle.

[0030] The real-time offshore wind turbine blade images captured and captured are preprocessed, including adjusting the brightness and contrast of the images, and performing image enhancement and other processing.

[0031] S2. Use key point detection algorithm to identify key points of preprocessed image data.

[0032] Specifically, the key point detection algorithm uses the yolov8-pose model as a basic model or selects other neural network models for identifying key points; the key points include: the center point of the cabin, the tip of the blade, the maximum bending point of the blade, and the connection point between the cabin and the blade, etc., wherein the connection points between the cabin and the blade are symmetrically arranged, such as the connection points between the three blades and the cabin are symmetrically arranged at 120° clockwise intervals, such as Figure 2 As shown, the blade tip and the connection point between the nacelle and the blade are set in association with each other with the blade length as the distance.

[0033] The yolov8-pose model is constructed by training historical offshore wind turbine blade images from different angles that are manually annotated with key point information (position, type, correlation, etc.). The loss function is set, the parameters of the yolov8-pose model are adjusted, and the training is iterative to generate a yolov8-pose model with a loss function value greater than the preset loss threshold. In addition, Pose as Compositional Tokens is added to the selected yolov8-pose model network architecture, and DCNv3 is used to replace the convolutional layer. Among them, the key points in the historical offshore wind turbine images of different angles and forms are encoded to further improve the accuracy of key detection.

[0034] Input each frame of image data after preprocessing into the trained yolov8-pose model, select the result of key recognition of offshore wind turbine blade image at the first moment and the result of key offshore wind turbine blade image recognition at the second moment, the second moment is taken as the next moment of the first moment, and the time interval between the first moment and the second moment is less than the preset time interval, which is relatively small, and is 1s in this embodiment. Figure 3 As shown, the key point information of a frame of image is obtained by using the key point detection algorithm.

[0035] S3. Obtaining angle differences of blades in different frames of image data according to the identified key points.

[0036] Specifically, a first vector is generated with the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin, and the blade tip of any blade identified in the offshore wind turbine blade image at the first moment as the end point; a second vector is generated with the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin, and the blade tip of the same blade identified in the offshore wind turbine blade image at the second moment as the end point; and the angle difference between the first vector and the second vector is calculated.

[0037] In this embodiment, the offshore wind turbine blade has three blades; a first vector corresponding to a blade (such as the first blade) is randomly selected according to the key point recognition result in the offshore wind turbine blade image at the first moment, and a second vector corresponding to the first blade is determined in the key point recognition result in the offshore wind turbine blade at the second moment, and the angle between the same blades at the two moments is calculated. The specific calculation formula is: Where θ is the angle between the same blades at two moments, that is, the angle difference between the same blades at two moments. is the first vector, is the second vector.

[0038] Among them, the same leaves can be determined by matching according to a key point matching algorithm; the key point matching algorithm matches the feature information of key points identified in different frame image data, and the feature information of the key points includes: position, texture, color, mark and shape, etc.; in addition to comparing the similarity of the feature information between the key points during the matching process, key point matching is also performed based on the type of each key point and the distance relationship between the key points.

[0039] In addition, considering that the characteristic information between the three blades in the offshore wind turbine blades is relatively similar, the type of training data, that is, the number and type of key points, can be increased when training the key point detection algorithm. For example, a key point is determined when each blade is separated by a certain distance; or key point matching and identification are performed in combination with vibration characteristic signals. Vibration sensors are installed at each key point to obtain the characteristic information of vibration signals corresponding to different key points, and the vibration characteristic information is combined with other characteristic information of the key points for comprehensive matching.

[0040] In this embodiment, the scenario where the offshore wind turbine blades rotate at a low speed is selected, that is, the scenario where the offshore wind turbine blade rotation speed is lower than the preset rotation speed is obtained according to the sensor installed on the offshore wind turbine blade. Wherein, at the preset rotation speed, the blade rotation angle in the interval between the first moment and the second moment is less than the two blade angles, that is, the blade rotation angle within 1s is less than 120 degrees. In this scenario, the blade whose blade tip position is closest to the blade tip position selected in the offshore wind turbine blade image at the first moment among the blade tip positions identified in the offshore wind turbine blade image at the second moment can be directly used as the same blade as the offshore wind turbine blade image at the first moment.

[0041] S4. Calculate the angular velocity of the offshore wind turbine blade according to the calculated angle difference between the first vector and the second vector.

[0042] Specifically, the offshore wind turbine blade angular velocity sequence is calculated by dividing the sum of the angle difference value and the angle difference compensation value by the time interval between the first moment and the second moment; the specific formula is: {θ}=θ+θ0 θ0=n*360° Where {θ} is the angular velocity sequence of offshore wind turbine blades, θ0 is the angle difference compensation value, and n is an integer between 0 and greater than 0.

[0043] Since the time interval from the first moment to the second moment is short, it is assumed that the difference in the angular velocity values ​​at the two moments is not large. Therefore, the angular velocity of the offshore wind turbine blades at the first moment or the second moment obtained by the sensor is compared with the angular velocity sequence of the offshore wind turbine blades to assist in determining the final angular velocity of the offshore wind turbine blades; that is, the angular velocity of the offshore wind turbine blades closest to the angular velocity of the offshore wind turbine blades collected by the sensor is selected from the angular velocity sequence of the offshore wind turbine blades as the final output angular velocity of the offshore wind turbine blades; wherein, the angular velocity of the offshore wind turbine blades at a certain moment obtained by the sensor can be the average value of the angular velocities obtained by the sensors installed on multiple blades.

[0044] Alternatively, the offshore wind turbine blade angular velocity sequence is directly compared with the average offshore wind turbine blade angular velocity at the first moment and the second moment to assist in determining the final offshore wind turbine blade angular velocity; that is, the offshore wind turbine blade angular velocity that is closest to the average offshore wind turbine blade angular velocity collected by the sensor is selected from the offshore wind turbine blade angular velocity sequence as the final output offshore wind turbine blade angular velocity.

[0045] In this embodiment, in the scenario where the offshore wind turbine blades rotate at a low speed, since it is impossible for the same blade to rotate one or more circles, n is taken as 0, and the angular velocity of the offshore wind turbine blades is directly calculated by dividing the angle difference between the first vector and the second vector by the time interval between the first moment and the second moment.

[0046] By adopting the solution described in this embodiment, the angular velocity of the target offshore wind turbine blade at the corresponding moment can be acquired with a high accuracy and the angular velocity can be acquired in real time.

[0047] In a specific embodiment, considering that certain environmental conditions, such as strong winds, may cause the shape of offshore wind turbine blades to deform, in order to avoid affecting key point matching due to deformation and thus reducing the accuracy of angular velocity acquisition, the method further includes: using sensors to collect environmental data in real time; the environmental data includes: wind direction, wind speed, wind force, etc.; The environmental data collected during the period from the first moment to the second moment are respectively input into the first neural network model, and the result of whether the offshore wind turbine blades are deformed during the period from the first moment to the second moment is output; the input of the first neural network model is the environmental data, and the output is whether the offshore wind turbine blades are deformed; the first neural network is generated by training the historical environmental data and the results of whether the offshore blades are deformed obtained at the corresponding moments.

[0048] If the offshore wind turbine blade is deformed at any time between the first moment and the second moment, the similarity threshold of the key points in the key point matching algorithm is correspondingly lowered, and the same blade at the first moment and the second moment is matched based on the adjusted key point matching algorithm; in this embodiment, the similarity threshold of the key points is reduced by 5%, for example: from 95% to 90%.

[0049] In addition, in the key point matching process, if the number of blades in the offshore wind turbine blade image at the second moment that matches the first blade identified in the offshore wind turbine blade image at the first moment is more than one, such as 3, then multiple (3 types) of blade matching results are determined based on the association between each matching result and the key point, that is, the first blade, the second blade, and the third blade identified in the offshore wind turbine blade image at the first moment, and multiple matching results are retained, and the angle differences corresponding to different matching results are calculated accordingly. The angular velocities of multiple offshore wind turbine blades are subsequently calculated, and the angular velocities of the offshore wind turbine blades collected by the sensor are compared to determine the final angular velocities of the offshore wind turbine blades.

[0050] In a specific embodiment, the accuracy of the angular velocity acquisition result is verified by calculating multiple angle differences, thereby optimizing the training key point detection algorithm and improving the accuracy of key point detection; the method includes: Taking the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin and the blade tip of each remaining blade not selected identified in the offshore wind turbine blade image at the first moment as the end point, a first vector corresponding to the blade is generated; taking the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin and the blade tip of the blade identical to each remaining blade not selected identified in the offshore wind turbine blade image at the second moment as the end point, a second vector corresponding to the blade is generated; For example: based on the key points of the second blade identified in the offshore wind turbine blade image at the first moment and the second moment, the first vector and the second vector of the second blade are generated; based on the key points of the third blade identified in the offshore wind turbine blade image at the first moment and the second moment, the first vector and the second vector of the third blade are generated.

[0051] Compare the angle difference calculated by the first vector and the second vector of the corresponding blade obtained according to each blade; for example, the angle difference calculated for the first blade is 30 degrees, the angle difference calculated for the second blade is 28 degrees, and the angle difference calculated for the third blade is 30 degrees; Ignoring the accuracy of the matching algorithm, if there is any two calculated angle differences that are greater than the preset angle difference, the key point types and quantities are increased, that is, the training key point detection algorithm is optimized by increasing the training data; in this embodiment, the preset angle difference is 1 degree, then the angle difference corresponding to the first blade and the angle difference corresponding to the second blade are 2 degrees greater than 1 degree, then the training key point detection algorithm needs to be further optimized.

[0052] The key points are re-identified according to the optimized key point detection algorithm, and the key points are re-matched according to the re-identified key points to determine the same leaves.

[0053] In a specific embodiment, considering the scenario where the sensor is damaged and the angular velocity of the offshore wind turbine blade cannot be obtained according to the sensor to assist in determining the final angular velocity, the auxiliary angular velocity is generated by predicting the rotation speed and calculating the angular velocity; the method includes: Taking into account that the offshore wind turbine blade control system will adjust the offshore wind turbine blade speed according to the environmental data, when the sensor cannot obtain the offshore wind turbine blade angular velocity in real time, the environmental data collected at the first moment or the second moment and the historical blade speed a period of time apart from the first moment or the second moment are input into the second neural network model, and the predicted offshore wind turbine blade speed at the first moment or the second moment is output.

[0054] The input of the second neural network model is the environmental data at any moment and the historical blade speed at a certain interval from any moment, and the output is the offshore wind turbine blade speed at the current moment; the second neural network is generated by training the historical offshore wind turbine blade environmental data, the blade speed at a certain interval from the historical offshore wind turbine blade environmental data collection time, and the speed at the historical offshore wind turbine blade environmental data collection time; wherein the preset period of time is set to 2 hours.

[0055] The offshore wind turbine blade angular velocity is calculated based on the predicted offshore wind turbine blade rotation speed, and the offshore wind turbine blade angular velocity sequence is compared with the offshore wind turbine blade angular velocity sequence, wherein the calculation formula between the offshore wind turbine blade rotation speed and the offshore wind turbine blade angular velocity is w=2πn / 60.

[0056] In a specific embodiment, considering the scenario where the sensor is damaged, the sensor detection result is reversely verified based on the acquired angular velocity to avoid reducing the accuracy of the finally acquired angular velocity; the method includes: Considering that the offshore wind turbine blade control system will set the maximum offshore wind turbine blade speed during the adjustment of the offshore wind turbine blade speed according to the environmental data, a corresponding offshore wind turbine blade angular velocity threshold is set. Therefore, it is judged whether the offshore wind turbine blade angular velocity outputted finally is greater than the preset offshore wind turbine blade angular velocity threshold; if greater, the offshore wind turbine blade angular velocity collected by the backup sensor at the first moment or the second moment is used to replace the offshore wind turbine blade angular velocity collected by the original sensor to compare the offshore wind turbine blade angular velocity sequence.

[0057] Considering that the offshore wind turbine blade control system will adjust the speed of the offshore wind turbine blade according to the environmental data, in order to avoid damage to the offshore wind turbine blade caused by rapid adjustment, the adjustment will be completed according to a certain speed change rate. Thus, the final output offshore wind turbine blade angular velocity sequence is obtained, and the moment when the absolute value of the angular velocity change value in the offshore wind turbine blade angular velocity sequence is greater than the preset angular velocity change threshold is monitored and recorded as the distortion moment; such as: the angular velocity change value at the first moment is the angular velocity at the first moment minus the initial angular velocity; the angular velocity change value at the second moment is the angular velocity at the second moment minus the angular velocity at the first moment; if the absolute value of the angular velocity change value at the second moment is greater than the preset angle change threshold, the second moment is recorded as the distortion moment.

[0058] The environmental data at the distortion moment is collected and analyzed accordingly; if the number of types of environmental data that meet the first condition is greater than the preset number, the angular velocity of the offshore wind turbine blades at the distortion moment collected by the backup sensor is used to replace the angular velocity of the offshore wind turbine blades collected by the original sensor to compare the angular velocity sequence of the offshore wind turbine blades; the first condition is: the absolute value of the change value of a type of environmental data at the distortion moment compared with the corresponding type of environmental data at the moment before the distortion moment is greater than the corresponding preset environmental data change threshold.

[0059] For example, the wind data corresponding to the second moment is X1, and the wind data corresponding to the first moment is X2. The change value of the wind data at the first moment compared to the wind data at the second moment is X2-X1, and the absolute value of the change value is greater than the corresponding wind data change threshold.

[0060] like Figure 4 As shown, the embodiment of the present application discloses a system for acquiring angular velocity of offshore wind turbine blades based on key points, comprising: An image acquisition module 101 is used to acquire images of offshore wind turbine blades in real time and perform preprocessing; The key point detection module 102 is used to use a key point detection algorithm to identify key points of the collected offshore wind turbine blade image at the first moment and the collected offshore wind turbine blade image at the second moment; the key points include: the center point of the cabin, the tip of the blade, and the connection point between the cabin and the blade; the key point detection algorithm uses the yolov8-pose model; The vector acquisition module 103 is used to generate a first vector with the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin and the blade tip of any blade identified in the offshore wind turbine blade image at the first moment as the end point; generate a second vector with the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin and the blade tip of the same blade identified in the offshore wind turbine blade image at the second moment as the end point; the same blade is determined by matching according to the key point matching algorithm; the angular velocity calculation module 104 is used to calculate the angle difference between the first vector and the second vector, and obtain the offshore wind turbine blade angular velocity sequence by dividing the sum of the angle difference and the angle difference compensation value by the time interval between the first moment and the second moment; the angle compensation value is n times 360 degrees, and n is 0 and an integer greater than 0; the offshore wind turbine blade angular velocity sequence is compared with the offshore wind turbine blade angular velocity sequence according to the offshore wind turbine blade angular velocity acquired by the sensor, and the offshore wind turbine blade angular velocity closest to the offshore wind turbine blade angular velocity acquired by the sensor is selected from the offshore wind turbine blade angular velocity sequence.

[0061] The system further comprises: The key point detection optimization module 105 is used to collect environmental data in real time using sensors; the environmental data includes: wind direction, wind speed and wind force; the environmental data collected during the first moment to the second moment are respectively input into the first neural network model, and the result of whether the offshore wind turbine blades are deformed during the first moment to the second moment is output; the input of the first neural network model is environmental data, and the output is whether the offshore wind turbine blades are deformed, and the result of whether the offshore blades are deformed is obtained through historical environmental data and corresponding moments; if the offshore wind turbine blades are deformed at any moment between the first moment and the second moment, the similarity threshold of the key points in the key point matching algorithm is correspondingly reduced, and the same blades at the first moment and the second moment are matched based on the adjusted key point matching algorithm.

[0062] The key point detection optimization module 105 is also used to generate a first vector corresponding to the blade with the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin and the blade tip of each remaining unselected blade identified in the offshore wind turbine blade image at the first moment as the end point; generate a second vector corresponding to the blade with the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin and the blade tip of the blade identical to each remaining unselected blade identified in the offshore wind turbine blade image at the second moment as the end point; compare the angle difference calculated between the first vector of the corresponding blade obtained according to each blade and the second vector, and if any two calculated angle differences are greater than the preset angle difference, increase the key point type and quantity to optimize the training of the key point detection algorithm, re-recognize the key points according to the optimized key point detection algorithm, and re-determine the same blade according to the re-recognized key points.

[0063] The angular velocity calculation module 104 is also used to input the environmental data collected at the first moment or the second moment and the historical blade speed at a certain interval from the first moment or the second moment into the second neural network model when the sensor cannot obtain the angular velocity of the offshore wind turbine blade in real time, and output the predicted offshore wind turbine blade speed at the first moment or the second moment; the input of the second neural network model is the environmental data at the current moment and the historical blade speed at a certain interval from the current moment, and the output is the offshore wind turbine blade speed at the current moment, which is generated by training the historical offshore wind turbine blade environmental data, the blade speed at a certain interval from the historical offshore wind turbine blade environmental data collection time, and the speed at the historical offshore wind turbine blade environmental data collection time; the offshore wind turbine blade angular velocity is calculated according to the predicted offshore wind turbine blade speed, and the offshore wind turbine blade angular velocity is compared with the offshore wind turbine blade angular velocity sequence by the calculated offshore wind turbine blade angular velocity.

[0064] The angular velocity calculation module 104 is also used to determine whether the offshore wind turbine blade angular velocity finally outputted is greater than a preset offshore wind turbine blade angular velocity threshold; if greater, the offshore wind turbine blade angular velocity collected by the backup sensor at the first moment or the second moment is used to replace the offshore wind turbine blade angular velocity collected by the original sensor to compare the offshore wind turbine blade angular velocity sequence.

[0065] The angular velocity calculation module 104 obtains the final output offshore wind turbine blade angular velocity sequence, monitors the moment when the absolute value of the angular velocity change value in the offshore wind turbine blade angular velocity sequence is greater than a preset angular velocity change threshold, records it as a distortion moment, and obtains and analyzes the environmental data at the distortion moment; if the number of types of environmental data that meet the first condition analyzed is greater than a preset number, the offshore wind turbine blade angular velocity at the distortion moment collected by the backup sensor is used to replace the offshore wind turbine blade angular velocity collected by the original sensor to compare the offshore wind turbine blade angular velocity sequence; the first condition is: the absolute value of the change value of a type of environmental data at the distortion moment compared with the corresponding type of environmental data at the moment before the distortion moment is greater than the corresponding preset environmental data change threshold.

[0066] The embodiment of the present application also discloses a computer-readable storage medium.

[0067] Specifically, the computer-readable storage medium stores a computer program that can be loaded by a processor and executed such as the above-mentioned method for obtaining the angular velocity of offshore wind turbine blades based on key points. The computer-readable storage medium includes, for example: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0068] The embodiment of the present application also discloses a computer device.

[0069] Specifically, the computer device includes a memory and a processor, and the memory stores a computer program that can be loaded by the processor and execute the above-mentioned method for obtaining the angular velocity of offshore wind turbine blades based on key points.

[0070] The above are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Any feature disclosed in this specification (including the abstract and drawings), unless otherwise stated, can be replaced by other equivalent or alternative features with similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

Claims

1. A method for obtaining the angular velocity of offshore wind turbine blades based on key points. It is characterized in that include: Real-time acquisition and preprocessing of offshore wind turbine blade images; A key point detection algorithm is used to identify key points of the offshore wind turbine blade images collected at the first moment and the offshore wind turbine blade images collected at the second moment; the key points include: the center point of the nacelle, the tip of the blade, and the connection point between the nacelle and the blade; the time interval between the first moment and the second moment is less than the preset time interval; the key point detection algorithm uses the yolov8-pose model; A first vector is generated by taking the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin and the blade tip of any blade identified in the offshore wind turbine blade image at the first moment as the end point; a second vector is generated by taking the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin and the blade tip of the same blade identified in the offshore wind turbine blade image at the second moment as the end point; the same blade is determined by matching according to a key point matching algorithm; Calculate the angle difference between the first vector and the second vector, and take the sum of the angle difference and the angle difference compensation value The offshore wind turbine blade angular velocity sequence is obtained by dividing by the time interval between the first moment and the second moment; the angle compensation value is n times 360 degrees, and n is an integer between 0 and 0; the offshore wind turbine blade angular velocity sequence is compared with the offshore wind turbine blade angular velocity sequence obtained by the sensor at the first moment or the second moment, and the offshore wind turbine blade angular velocity sequence that is closest to the offshore wind turbine blade angular velocity collected by the sensor is selected as the offshore wind turbine blade angular velocity that is finally output.

2. The method for acquiring the angular velocity of offshore wind turbine blades based on key points according to claim 1, It is characterized in that Also includes: Use sensors to collect environmental data in real time; The environmental data include: wind direction, wind speed and wind force; The environmental data collected during the period from the first moment to the second moment are respectively input into the first neural network model, and the result of whether the offshore wind turbine blade is deformed during the period from the first moment to the second moment is output; the input of the first neural network model is the environmental data, and the output is whether the offshore wind turbine blade is deformed, which is generated by training through historical environmental data and the result of whether the offshore blade is deformed obtained at the corresponding moment; If the offshore wind turbine blade is deformed at any time between the first moment and the second moment, the similarity threshold of the key points in the key point matching algorithm is correspondingly lowered, and the same blade at the first moment and the second moment is matched based on the adjusted key point matching algorithm.

3. The method for acquiring the angular velocity of offshore wind turbine blades based on key points according to claim 1, It is characterized in that Also includes: Taking the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin and taking the blade tip of each remaining unselected blade identified in the offshore wind turbine blade image at the first moment as the end point, generating a first vector corresponding to the blade; Taking the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin and taking the blade tip of the blade identical to each of the remaining unselected blades identified in the offshore wind turbine blade image at the second moment as the end point, generating a second vector corresponding to the blade; The angle difference calculated by the first vector and the second vector of the corresponding leaf obtained for each leaf is compared. If any two calculated angle differences are greater than the preset angle difference, the key point types and quantities are increased to optimize the training of the key point detection algorithm. The key points are re-recognized according to the optimized key point detection algorithm, and the same leaf is re-determined based on the re-recognized key points.

4. The method for acquiring angular velocity of offshore wind turbine blades based on key points according to claim 1, It is characterized in that Also includes: When the angular velocity of the offshore wind turbine blade cannot be obtained in real time according to the sensor, the environmental data collected at the first moment or the second moment and the historical blade rotation speed at a certain interval from the first moment or the second moment are input into the second neural network model, and the predicted offshore wind turbine blade rotation speed at the first moment or the second moment is output; the input of the second neural network model is the environmental data at the current moment and the historical blade rotation speed at a certain interval from the current moment, and the output is the offshore wind turbine blade rotation speed at the current moment, which is generated by training the historical offshore wind turbine blade environmental data, the blade rotation speed at a certain interval from the time when the historical offshore wind turbine blade environmental data is collected, and the rotation speed at the time when the historical offshore wind turbine blade environmental data is collected; The offshore wind turbine blade angular velocity is calculated based on the predicted offshore wind turbine blade rotation speed, and the calculated offshore wind turbine blade angular velocity is compared with the offshore wind turbine blade angular velocity sequence.

5. The method for acquiring angular velocity of offshore wind turbine blades based on key points according to claim 1, It is characterized in that Also includes: Determine whether the offshore wind turbine blade angular velocity finally outputted is greater than a preset offshore wind turbine blade angular velocity threshold; If it is greater, the offshore wind turbine blade angular velocity collected by the backup sensor at the first moment or the second moment is used to replace the offshore wind turbine blade angular velocity collected by the original sensor to compare the offshore wind turbine blade angular velocity sequence.

6. The method for acquiring angular velocity of offshore wind turbine blades based on key points according to claim 2, It is characterized in that Also includes: Obtain the final output offshore wind turbine blade angular velocity sequence, monitor the moment when the absolute value of the angular velocity change value in the offshore wind turbine blade angular velocity sequence is greater than a preset angular velocity change threshold, record it as a distortion moment, obtain the corresponding environmental data at the distortion moment and analyze it; If the number of types of environmental data that meet the first condition is greater than the preset number, the angular velocity of the offshore wind turbine blades collected by the backup sensor at the distortion moment is used to replace the angular velocity of the offshore wind turbine blades collected by the original sensor to compare the angular velocity sequence of the offshore wind turbine blades; the first condition is: the absolute value of the change value of a type of environmental data at the distortion moment compared with the corresponding type of environmental data at the moment before the distortion moment is greater than the corresponding preset environmental data change threshold.

7. The method for acquiring angular velocity of offshore wind turbine blades based on key points according to claim 1, It is characterized in that Add Pose as Compositional Tokens to the selected yolov8-pose model network architecture, and use DCNv3 to replace the convolutional layer.

8. An offshore wind turbine blade angular velocity system based on key points, It is characterized in that include: Image acquisition module, used for real-time acquisition and preprocessing of offshore wind turbine blade images; A key point detection module is used to identify key points of the first moment offshore wind turbine blade image and the second moment offshore wind turbine blade image respectively using a key point detection algorithm; the key points include: the center point of the cabin, the blade tip, and the connection point between the cabin and the blade; the key point detection algorithm uses the yolov8-pose model; A vector acquisition module is used to generate a first vector with the center point of the nacelle identified in the offshore wind turbine blade image at the first moment as the origin and the blade tip of any blade identified in the offshore wind turbine blade image at the first moment as the end point; and to generate a second vector with the center point of the nacelle identified in the offshore wind turbine blade image at the second moment as the origin and the blade tip of the same blade identified in the offshore wind turbine blade image at the second moment as the end point; the same blade is determined by matching according to a key point matching algorithm; The angular velocity calculation module is used to calculate the angle difference between the first vector and the second vector, and obtain the offshore wind turbine blade angular velocity sequence by dividing the sum of the angle difference and the angle difference compensation value by the time interval between the first moment and the second moment; the angle compensation value is n times 360 degrees, and n is an integer greater than 0; the offshore wind turbine blade angular velocity sequence is compared with the offshore wind turbine blade angular velocity obtained by the sensor, and the offshore wind turbine blade angular velocity that is closest to the offshore wind turbine blade angular velocity collected by the sensor is selected from the offshore wind turbine blade angular velocity sequence.

9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

10. A computer device, It is characterized in that The computer device comprises a memory, a processor and a program stored and executable on the memory, and the program implements the steps of the method according to any one of claims 1 to 7 when executed by the processor.

Images