A method for monitoring impeller unbalance based on machine vision

Abstract

This invention provides a method for monitoring impeller imbalance based on machine vision. First, a camera is installed at the lower front of the nacelle of a wind turbine. The camera acquires video data and transmits it to a processor in real time. Finally, the processor processes each frame of the video data to determine impeller imbalance. This invention provides a method or device for monitoring impeller imbalance based on machine vision or other measurement methods that can identify the trajectory of the blade tip or blade body and calculate the distance difference between the three blades to determine impeller imbalance. Simultaneously, the trajectory distance results can be combined with master control information to eliminate abnormal points, thereby improving the accuracy of the early warning.

Description

Technical Field

[0001] This invention relates to the field of wind turbine monitoring technology, and in particular to a method for monitoring impeller imbalance based on machine vision. Background Technology

[0002] As wind turbine capacity and rotor diameter continue to increase, the risks and hazards associated with rotor imbalance are growing. Aerodynamic and mass imbalances caused by blade icing, fouling, lightning strikes, and other factors exacerbate system fatigue and ultimate loads, leading to increased turbine vibration, reduced turbine lifespan, and even major safety accidents. Therefore, identifying and evaluating the causes, locations, and severity of rotor imbalance-related failures, and implementing targeted maintenance decisions to extend equipment lifespan, reduce maintenance time, improve maintenance quality, save maintenance costs, and extend the continuous trouble-free operation time of wind turbines, becomes crucial.

[0003] Currently existing methods for testing and monitoring impeller imbalance include: 1. Strain gauge monitoring; 2. Ground clearance measurement. The strain gauge monitoring method involves installing strain gauges on the blades to measure the load strain, identify load imbalance, and thus provide early warnings. This method is relatively direct, but the blades are slender structures, requiring strain gauges to be attached internally, which is difficult to install. Furthermore, a large number of strain gauges are needed, and the blades have particularly high lightning protection requirements, posing significant challenges to this method. The ground clearance measurement method uses ground-mounted cameras to measure the continuous clearance values ​​of the unit, calculating the clearance values ​​for each blade. Significant differences in clearance values ​​between different blades indicate impeller imbalance. However, this method cannot provide real-time monitoring and early warning, and it has certain requirements regarding camera testing distance, making it unsuitable for units in special terrains or regions. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a machine vision-based method for monitoring impeller imbalance. It uses machine vision to acquire real-time images of the tower and blades from the nacelle location. Through image processing, it obtains the equivalent points of the blade tips in each frame, and then plots the trajectories traced by different blades during impeller rotation. The distance between the trajectories of each blade is used to determine whether a set limit has been exceeded, triggering an alarm. This method offers real-time monitoring and is easy to implement. When integrated with the unit's main control system, it can also provide impeller imbalance monitoring and alarm functions.

[0005] The technical solution adopted by this invention to solve its technical problem is: a method for monitoring impeller imbalance based on machine vision is provided, comprising the following steps:

[0006] S1. Install the camera at the lower front end of the wind turbine nacelle;

[0007] S2. The camera acquires video data and transmits it to the processor via a real-time video stream.

[0008] S3. The processor processes each frame of the video data to determine impeller imbalance.

[0009] Step S3 specifically includes the following processes:

[0010] S3.1 Initialize the record list and start timing;

[0011] S3.2 For each frame of the video data, the leaf outline is identified and the leaf tip is found through machine learning methods or frame difference, noise reduction and grayscale processing, and the equivalent points of the leaf tip are formed and recorded in the record list.

[0012] S3.3 For the equivalent points of the leaf tips in the record list, fit straight lines according to the least squares method to obtain the trajectory lines of each leaf.

[0013] S3.4. Determine the validity of the trajectory lines of each blade. If the determination is valid, proceed to step S3.6; otherwise, discard the trajectory lines.

[0014] S3.5 Select the area within a certain number of pixels before and after the blades at a specific point, calculate the distance between the trajectory lines of each blade, obtain the distance between each blade, and take the maximum distance as the distance difference △.

[0015] S3.6 Determine whether the distance difference △ exceeds the preset value. If so, add the distance difference △ to the record list. If the length of the record list is greater than the limit length, an alarm will be triggered, indicating that there is impeller imbalance, which requires the user to investigate, verify and resolve.

[0016] S3.7 Determine whether the number of records in the record list has reached the warning number or exceeded the timer period. If either condition is met, clear the record list and return to step S3.1; otherwise, proceed to step S3.2.

[0017] Step S3.3, fitting the straight line using the least squares method, specifically includes the following process: Let the straight line to be fitted be expressed as: y = a + bx, where the parameters to be solved, a, represent the intercept and b, represent the slope; estimate the parameters to be solved, a and b, using the least squares method, and substitute the equivalent point data of the leaf tip into the equation. To minimize its value; where (x i y i ( ) represents the equivalent point data at the leaf tip; the best estimate of the parameter 'a' to be solved is obtained by solving for it. The best estimate of b

[0018] Step S3.4, which assesses the validity of the trajectory lines of each blade, specifically includes the following process: calculating the correlation coefficient. in Using the correlation coefficient to fit the linear relationship and the coefficient of determination r 2 Alternatively, the slope 'a' can be used to determine the validity of the trajectory lines of each blade.

[0019] After obtaining the distance difference Δ in step S3.5, the stability of the operating condition is judged through the following process: access the main control data of the wind turbine, obtain the data on the variation range of average wind speed, impeller speed, power and blade pitch angle, select the parameters as constraints according to the characteristics of the unit, and judge the stability of the operating condition.

[0020] In step S3.6, the impeller azimuth angle or blade number information is obtained by accessing the main control data of the wind turbine, and the unbalanced blades of the impeller are located.

[0021] The beneficial effects of this invention based on its technical solution are as follows:

[0022] This invention provides a machine vision-based impeller imbalance monitoring method. It deploys a video blade monitoring system that uses image recognition to conveniently, effectively, and accurately capture blade trajectories, thereby identifying impeller imbalance phenomena, providing real-time monitoring and timely alarms, and preventing damage to the unit caused by impeller imbalance. This system enables automated and autonomous early warning and protection without human intervention. The image recognition accuracy is at the centimeter level, with high precision and resolution. The judgment logic is simple and effective, and the system can acquire impeller imbalance logic and integrated unit control information, improving the accuracy of early warnings. Attached Figure Description

[0023] Figure 1 This is a three-dimensional schematic diagram of the equivalent point at the leaf tip.

[0024] Figure 2 This is a schematic diagram of the equivalent point imaging at the leaf tip.

[0025] Figure 3 This is a schematic diagram of the blade trajectory.

[0026] Figure 4 This is a schematic diagram of trajectory distance calculation. Detailed Implementation

[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0028] This invention provides a method for monitoring impeller imbalance based on machine vision, comprising the following steps:

[0029] S1. Install the camera at the lower front of the wind turbine nacelle.

[0030] S2. The camera acquires video data and transmits it to the processor in real time via video streaming.

[0031] S3. The processor processes each frame of the video data to determine impeller imbalance, specifically including the following processes:

[0032] S3.1 Initialize the record list and start timing.

[0033] S3.2 For each frame of the video data, machine learning methods or frame differencing, denoising, and grayscale processing are used. For example... Figure 1 As shown, the camera captures the blades and tower in three-dimensional space, forming a two-dimensional image, such as... Figure 2 As shown, the equivalent points of the leaf tip can be obtained from the two-dimensional image through professional image processing algorithms, and the equivalent points of the leaf tip can be recorded in the record list.

[0034] S3.3. For the equivalent points of the leaf tips in the record list, fit straight lines using the least squares method to obtain the trajectory lines of each leaf, such as... Figure 3 As shown. The process of fitting a straight line using the least squares method specifically includes the following steps:

[0035] Let the line to be fitted be represented as: y = a + bx, where the parameters to be solved, a, represent the intercept and b, represent the slope. Estimate the parameters a and b using the least squares method, and substitute the equivalent point data from the leaf tip into the equation. To minimize its value; where (x i y i ( ) represents the equivalent point data at the leaf tip; the best estimate of the parameter 'a' to be solved is obtained by solving for it. The best estimate of b

[0036] S3.4. Determine the validity of the trajectory lines of each blade. If the determination is valid, proceed to step S3.6; otherwise, discard the trajectory lines.

[0037] The validity assessment of the trajectory lines of each blade specifically includes the following process: calculating the correlation coefficient. in Using the correlation coefficient to fit the linear relationship and the coefficient of determination r 2 Alternatively, the slope 'a' can be used to determine the validity of the trajectory lines of each blade: if the slopes of each blade are significantly different, indicating a large deviation, it suggests a problem with the identified trajectory; and if r 2 The smaller ones are considered to have poor linear fitting performance and are therefore removed, further reducing the probability of false alarms in the system.

[0038] S3.5. Select a region within a specified number of pixels before and after a specific point for each blade, calculate the spacing between the blade trajectory lines, obtain the spacing between each blade, and take the maximum distance as the distance difference Δ. Figure 4As shown. Specifically, the stability of the operating conditions is determined through the following process: Access the wind turbine's main control data to obtain data on the average wind speed, impeller speed, power, and pitch angle variation range. Based on the unit's characteristics, select these parameters as constraints to determine the stability of the operating conditions. For example, a pitch angle variation <1°, wind speed fluctuation <3m / s, power fluctuation <200kW, and impeller speed variation <5%.

[0039] S3.6. Determine if the distance difference △ exceeds the preset value. If so, add the distance difference △ to the record list. If the record list length exceeds the limit, trigger an alarm, indicating impeller imbalance, requiring user investigation, verification, and resolution. If connected to the fan main control data, the impeller azimuth angle or blade number information can also be obtained to locate the unbalanced blades.

[0040] S3.7 Determine whether the number of records in the record list has reached the warning number or exceeded the timer period. If either condition is met, clear the record list and return to step S3.1; otherwise, proceed to step S3.2.

[0041] This invention provides a method for monitoring impeller imbalance based on machine vision. The method or device can identify the trajectory of the blade tip or blade body and calculate the distance difference between each blade to determine the impeller imbalance based on machine vision or other measurement methods. At the same time, the trajectory distance results can be combined with the master control information to remove abnormal points and improve the accuracy of early warning.

Claims

1. A method for monitoring impeller imbalance based on machine vision, characterized in that... Includes the following steps: S1. Install the camera at the lower front end of the wind turbine nacelle; S2. The camera acquires video data and transmits it to the processor via a real-time video stream. S3. The processor processes each frame of the video data to perform impeller imbalance early warning judgment, specifically... Includes the following processes: S3.1 Initialize the record list and start timing; S3.2 For each frame of the video data, the leaf outline is identified and the leaf tip is found through machine learning methods or frame difference, noise reduction and grayscale processing, and the equivalent points of the leaf tip are formed and recorded in the record list. S3.

3. For the equivalent points of the leaf tips in the record list, fit a straight line using the least squares method to obtain the trajectory line of each leaf; specifically, fit the straight line using the least squares method... The process includes the following steps: Let the line to be fitted be represented as: y = a + bx, where the parameters to be solved, a, represent the intercept and b, represent the slope; estimate the parameters to be solved, a and b, using the least squares method, and substitute the equivalent point data of the leaf tip into the equation. To minimize its value; where (x i ,y i ( ) represents the equivalent point data at the leaf tip; the best estimate of the parameter 'a' to be solved is obtained by solving for it. The best estimate of b ; S3.

4. Validate the trajectory lines of each blade. If valid, proceed to step S3.5; otherwise, discard the trajectory lines. The validity determination of the trajectory lines of each blade includes the following process: calculating the correlation coefficient. ,in The effectiveness of the trajectory lines of each blade is judged using the correlation coefficient r or the slope b. S3.5 Select the area within a certain number of pixels before and after the blades at a specific point, calculate the distance between the trajectory lines of each blade, obtain the distance between each blade, and take the maximum distance as the distance difference △. S3.6 Determine whether the distance difference △ exceeds the preset value. If so, add the distance difference △ to the record list. If the length of the record list is greater than the limit length, an alarm will be triggered, indicating that there is impeller imbalance, which requires the user to investigate, verify and resolve. S3.7 Determine whether the number of records in the record list has reached the warning number or exceeded the timer period. If either condition is met, clear the record list and return to step S3.1; otherwise, proceed to step S3.

2.

2. The method for monitoring impeller imbalance based on machine vision according to claim 1, characterized in that: Step S3.5 obtains the distance difference. Then, the stability of the operating condition is judged through the following process: access the main control data of the wind turbine, obtain the data on the variation range of average wind speed, impeller speed, power and blade pitch angle, select the parameters as constraints according to the characteristics of the unit, and judge the stability of the operating condition.

3. The method for monitoring impeller imbalance based on machine vision according to claim 1, characterized in that: In step S3.6, the impeller azimuth angle or blade number information is obtained by accessing the main control data of the wind turbine, and the unbalanced blades of the impeller are located.

Images