A fan brake early warning method based on infrared images

Abstract

This invention provides a wind turbine braking early warning method based on infrared images. In this invention, infrared images of the wind turbine during operation are collected, and the infrared images are clustered to obtain the pixels of each cluster. Then, the temperature of each cluster is obtained through an image temperature prediction model. The highest temperature is found, and when the highest temperature is higher than the temperature threshold, the wind turbine enters the braking process. This invention utilizes the characteristic that images can capture the temperature of a region of the wind turbine, thereby realizing the temperature monitoring of a region of the wind turbine. This solves the problems of small monitoring range, complex wiring, and complicated installation that exist in the method of using fixed-point monitoring thermometers for high temperature early warning of wind turbines.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically, to a wind turbine braking early warning method based on infrared images. Background Technology

[0002] During high-speed rotation, the fan will generate high temperatures. If it is not braked and cooled down in time, it will damage the fan itself. However, the fan is large in size, and it is difficult to monitor the fan temperature comprehensively using a fixed-point monitoring thermometer. If multiple fixed-point monitoring thermometers are used to monitor multiple parts of the fan, they can only monitor the temperature of one point in each part. In addition, using multiple thermometers will result in complicated wiring and installation. Summary of the Invention

[0003] To address the aforementioned shortcomings in existing technologies, this invention provides a wind turbine brake early warning method based on infrared images, which solves the problems of small monitoring range, complex wiring, and complex installation associated with using fixed-point monitoring thermometers for wind turbine high-temperature early warning.

[0004] To achieve the aforementioned objectives, the present invention employs the following technical solution: a wind turbine braking early warning method based on infrared images, comprising:

[0005] S1. Acquire infrared images during the operation of the fan;

[0006] S2. Perform clustering processing on the infrared image to obtain the pixels of each cluster;

[0007] S3. Based on the pixel values ​​of each cluster of pixels, the temperature of each cluster is obtained using the image temperature prediction model;

[0008] S4. Based on the temperature of each cluster, find the highest temperature;

[0009] S5. If the maximum temperature is higher than the temperature threshold, the fan will enter the braking process.

[0010] Furthermore, step S2 includes the following sub-steps:

[0011] S21. Find the maximum and minimum pixel values ​​from the infrared image;

[0012] S22. Based on the maximum and minimum pixel values, set the cluster center sequence {I}. i}, where I i For the i-th cluster center, when i=1, I1 equals the minimum pixel value, and when i=N, I... N Equals the maximum pixel value, where N is the number of cluster centers;

[0013] S23. Calculate the distance between the pixel value of all pixels in the infrared image and the center of each cluster.

[0014] S24. When there is a common cluster center, all pixels whose distance is less than the distance threshold are grouped into one cluster to obtain the pixels of each cluster.

[0015] Furthermore, in S22, except for I1 and I N The formula for calculating the cluster centers is:

[0016] I i =I1+(i-1)·[(I N -I1) / (N-1)]

[0017] Among them, I i Let I1 be the i-th cluster center and I1 be the 1-th cluster center. N Let N be the Nth cluster center, where N is the number of cluster centers.

[0018] Furthermore, the expression for the image temperature prediction model in S3 is: a first prediction sub-model and a second prediction sub-model.

[0019] Furthermore, the first prediction sub-model is:

[0020]

[0021] Where g is the output of the first prediction sub-model, arctan is the arctangent function, w1 is the weight of the first prediction sub-model, b1 is the bias of the first prediction sub-model, and x is the pixel feature value of each cluster.

[0022] Furthermore, the second prediction sub-model is:

[0023] y = tanh(w²g + b²)

[0024] Where y is the output of the second predictor sub-model, tanh is the hyperbolic tangent function, w2 is the weight of the second predictor sub-model, b2 is the bias of the second predictor sub-model, and g is the output of the first predictor sub-model.

[0025] Furthermore, step S3 includes the following sub-steps:

[0026] S31. Calculate the pixel feature value of each cluster based on the pixel value of each cluster's pixels;

[0027] S32. Input the pixel feature values ​​of each cluster into the image temperature prediction model to obtain the temperature of each cluster.

[0028] Furthermore, the formula for calculating the pixel feature value of each cluster in S31 is as follows:

[0029]

[0030] Where X is the pixel feature value of each cluster. θ is the average pixel value for each cluster, and θ is the scaling factor.

[0031] Furthermore, the formula for the proportionality coefficient θ is:

[0032]

[0033] Among them, X max The maximum pixel value of each pixel within each cluster, X min The minimum pixel value of a pixel within each cluster. The average pixel value for each cluster.

[0034] The technical solution of this invention has at least the following advantages and beneficial effects: In this invention, infrared images of the wind turbine's operation process are collected, clustered to obtain the pixels of each cluster, and then the temperature of each cluster is obtained through an image temperature prediction model. The highest temperature is found, and when the highest temperature is higher than the temperature threshold, the wind turbine enters the braking process. This invention utilizes the characteristics of a region of the wind turbine to collect images, thereby realizing temperature monitoring of a region of the wind turbine. This solves the problems of small monitoring range, complex wiring, and complex installation that exist in the method of using fixed-point monitoring thermometers for high temperature early warning of wind turbines. Attached Figure Description

[0035] Figure 1 This is a flowchart of a wind turbine braking early warning method based on infrared images. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0037] like Figure 1 As shown, a wind turbine braking early warning method based on infrared images includes:

[0038] S1. Acquire infrared images during the operation of the fan;

[0039] S2. Perform clustering processing on the infrared image to obtain the pixels of each cluster;

[0040] S2 includes the following steps:

[0041] S21. Find the maximum and minimum pixel values ​​from the infrared image;

[0042] S22. Based on the maximum and minimum pixel values, set the cluster center sequence {I}. i}, where I i For the i-th cluster center, when i=1, I1 equals the minimum pixel value, and when i=N, I... N Equals the maximum pixel value, where N is the number of cluster centers;

[0043] In S22, except for I1 and I N The formula for calculating the cluster centers is:

[0044] I i =I1+(i-1)·[(I N -I1) / (N-1)]

[0045] Among them, I i Let I1 be the i-th cluster center and I1 be the 1-th cluster center. N Let N be the Nth cluster center, where N is the number of cluster centers;

[0046] S23. Calculate the distance between the pixel value of all pixels in the infrared image and the center of each cluster.

[0047] In this embodiment, the distance can be the absolute value of the difference between two pixel values.

[0048] S24. When there is a common cluster center, all pixels whose distance is less than the distance threshold are grouped into one cluster to obtain the pixels of each cluster.

[0049] In this invention, the maximum and minimum pixel values ​​are found in the infrared image to determine the range of pixel values ​​in the infrared image. Based on the range of pixel values, the classification level is determined, the cluster center is found, and the distance between each pixel value and each cluster center is calculated. Pixel values ​​less than a threshold are grouped into one cluster.

[0050] S3. Based on the pixel values ​​of each cluster of pixels, the temperature of each cluster is obtained using the image temperature prediction model;

[0051] The expression for the image temperature prediction model in S3 is: the first prediction sub-model and the second prediction sub-model.

[0052] The first prediction sub-model is:

[0053]

[0054] Where g is the output of the first prediction sub-model, arctan is the arctangent function, w1 is the weight of the first prediction sub-model, b1 is the bias of the first prediction sub-model, and x is the pixel feature value of each cluster.

[0055] The second prediction sub-model is:

[0056] y = tanh(w²g + b²)

[0057] Where y is the output of the second predictor sub-model, tanh is the hyperbolic tangent function, w2 is the weight of the second predictor sub-model, b2 is the bias of the second predictor sub-model, and g is the output of the first predictor sub-model.

[0058] S3 includes the following steps:

[0059] S31. Calculate the pixel feature value of each cluster based on the pixel value of each cluster's pixels;

[0060] S32. Input the pixel feature values ​​of each cluster into the image temperature prediction model to obtain the temperature of each cluster.

[0061] The formula for calculating the pixel feature value of each cluster in S31 is as follows:

[0062]

[0063] Where X is the pixel feature value of each cluster. θ is the average pixel value for each cluster, and θ is the scaling factor.

[0064] The formula for the proportionality coefficient θ is:

[0065]

[0066] Among them, X max The maximum pixel value of each pixel within each cluster, X min The minimum pixel value of a pixel within each cluster. The average pixel value for each cluster.

[0067] After clustering, the pixel values ​​of pixels within a cluster are not significantly different, but there are fluctuations. Therefore, this invention uses a scaling factor to enhance the average value, so that the pixel feature value can express the highest temperature of the region.

[0068] S4. Based on the temperature of each cluster, find the highest temperature;

[0069] S5. If the maximum temperature is higher than the temperature threshold, the fan will enter the braking process.

[0070] The technical solution of this invention has at least the following advantages and beneficial effects: In this invention, infrared images of the wind turbine's operation process are collected, clustered to obtain the pixels of each cluster, and then the temperature of each cluster is obtained through an image temperature prediction model. The highest temperature is found, and when the highest temperature is higher than the temperature threshold, the wind turbine enters the braking process. This invention utilizes the characteristics of a region of the wind turbine to collect images, thereby realizing temperature monitoring of a region of the wind turbine. This solves the problems of small monitoring range, complex wiring, and complex installation that exist in the method of using fixed-point monitoring thermometers for high temperature early warning of wind turbines.

[0071] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A wind turbine braking early warning method based on infrared images, characterized in that, include: S1. Acquire infrared images during the operation of the fan; S2. Perform clustering processing on the infrared image to obtain the pixels of each cluster; S3. Based on the pixel values ​​of each cluster of pixels, the temperature of each cluster is obtained using the image temperature prediction model; S4. Based on the temperature of each cluster, find the highest temperature; S5. If the maximum temperature exceeds the temperature threshold, the fan will enter the braking process. S2 includes the following steps: S21. Find the maximum and minimum pixel values ​​from the infrared image; S22. Set the cluster center sequence based on the maximum and minimum pixel values. ,in, For the first Cluster centers, in hour, Equal to the minimum pixel value, in hour, Equal to the maximum pixel value, The number of cluster centers; S23. Calculate the distance between the pixel value of all pixels in the infrared image and the center of each cluster. S24. When there is a common cluster center, all pixels whose distance is less than the distance threshold are grouped into one cluster to obtain the pixels of each cluster. In S22, except and The formula for calculating the cluster centers is: in, For the first Cluster centers, As the first cluster center, For the first Cluster centers, The number of cluster centers; The expression for the image temperature prediction model in S3 is: the first prediction sub-model and the second prediction sub-model; The first prediction sub-model is: in, This is the output of the first prediction sub-model. It is the arctangent function. The weights of the first prediction sub-model. The bias of the first prediction sub-model. The pixel feature values ​​for each cluster; The second prediction sub-model is: in, This is the output of the second prediction sub-model. It is the hyperbolic tangent function. The weights of the second prediction sub-model, The bias for the second prediction sub-model. This is the output of the first prediction sub-model.

2. The wind turbine braking early warning method based on infrared images according to claim 1, characterized in that, S3 includes the following steps: S31. Calculate the pixel feature value of each cluster based on the pixel value of each cluster's pixels; S32. Input the pixel feature values ​​of each cluster into the image temperature prediction model to obtain the temperature of each cluster.

3. The wind turbine braking early warning method based on infrared images according to claim 2, characterized in that, The formula for calculating the pixel feature value of each cluster in S31 is as follows: in, For each cluster, the pixel feature value, The average pixel value for each cluster. This is the proportionality coefficient.

4. The wind turbine braking early warning method based on infrared images according to claim 3, characterized in that, The proportionality coefficient The formula is: in, The maximum pixel value of each pixel within each cluster. The minimum pixel value of a pixel within each cluster. The average pixel value for each cluster.

Images