A wind turbine fault diagnosis method

Abstract

The present invention discloses a wind turbine fault diagnosis method, belonging to the field of wind turbine monitoring technology. The method comprises the following steps: S1: preprocessing a real-time wind turbine operating image to generate a standard wind turbine operating image; S2: determining the blade area of ​​the standard wind turbine operating image to generate a standard blade operating image; S3: inputting the standard blade operating image into a wind turbine fault diagnosis model to determine the wind turbine fault diagnosis result. The wind turbine fault diagnosis method collects and preprocesses the wind turbine operating image to ensure high image clarity and noise-free image; extracts the blade area of ​​the wind turbine operating image, and in subsequent steps, only processes the blade area, which can significantly reduce algorithm complexity and algorithm flow.

Description

Technical Field

[0001] The invention belongs to the technical field of wind turbine monitoring, and in particular relates to a wind turbine fault diagnosis method. Background Art

[0002] Blades are the primary components of wind turbines for absorbing wind energy. Each blade in a wind turbine is driven by a hydraulic or electric mechanism. During operation, each blade should rotate in the same direction and pitch angle. However, when blade pitch asynchrony occurs, the angles of each blade change inconsistently, affecting the turbine's output and, in severe cases, causing significant damage. Therefore, monitoring wind turbine blades for faults is crucial. Summary of the Invention

[0003] In order to solve the above problems, the present invention proposes a wind turbine fault diagnosis method.

[0004] The technical solution of the present invention is: a wind turbine fault diagnosis method comprising the following steps:

[0005] S1: collecting real-time wind turbine operation images, and preprocessing the real-time wind turbine operation images to generate standard wind turbine operation images;

[0006] S2: determining the blade area of ​​the standard wind turbine operating image and generating the standard blade operating image;

[0007] S3: Construct a wind turbine fault diagnosis model, input the standard blade operation image into the wind turbine fault diagnosis model, and determine the wind turbine fault diagnosis result.

[0008] Furthermore, S1 includes the following sub-steps:

[0009] S11, collecting real-time wind turbine operation images, performing grayscale calibration on the real-time wind turbine operation images, and obtaining calibrated wind turbine operation images;

[0010] S12, rotating the calibration wind turbine operating image to obtain a horizontal wind turbine operating image;

[0011] S13. Perform normalization and denoising processing on the horizontal wind turbine operating image to obtain a standard wind turbine operating image.

[0012] The beneficial effect of the above further scheme is: in the present invention, since the wind turbine is located outdoors, the distortion and noise of the collected wind turbine operation image are relatively large, so image preprocessing is required. In view of the fact that the wind turbine operation image may have uneven exposure, grayscale calibration is required to make the grayscale of the entire image uniform, and generate a calibrated wind turbine operation image. The calibrated wind turbine operation image may have a directional offset, and it is easy to divide the image that does not belong to the blade area into the blade area in the later stage, resulting in inaccurate cropping of the standard blade operation image. Therefore, the calibrated wind turbine operation image needs to be rotated and corrected to keep the image level.

[0013] Furthermore, in S11, the calculation formula for the grayscale value μ of the pixel in the calibrated wind turbine operation image is:

[0014]

[0015] Where H Rmax Indicates the maximum pixel value of the real-time wind turbine operation image in the R channel, H Gmax Indicates the maximum pixel value of the real-time wind turbine operating image in the G channel, H Bmax represents the maximum pixel value of the real-time wind turbine operating image in channel B, h1 represents the maximum grayscale value of the real-time wind turbine operating image, h2 represents the minimum grayscale value of the real-time wind turbine operating image, and h ave Indicates the average grayscale value of the real-time wind turbine operation image.

[0016] Furthermore, S12 includes the following sub-steps:

[0017] S121, connecting the pixel point corresponding to the maximum grayscale value and the pixel point corresponding to the minimum grayscale value in the calibration wind turbine image to form a pixel point connection line;

[0018] S122, determine whether the angle between the pixel line and the horizontal direction is less than 45 degrees, if so, proceed to S123, otherwise proceed to S124;

[0019] S123, rotating the calibration wind turbine image until the angle between the pixel line and the horizontal direction is 0°, thereby generating a horizontal wind turbine operation image;

[0020] S124. After rotating the calibration wind turbine image until the angle between the pixel line and the horizontal direction is 45°, crop the calibration wind turbine image using a cropping frame, and rotate the cropped calibration wind turbine image until the angle between the pixel line and the horizontal direction is 0°, thereby generating a horizontal wind turbine operation image.

[0021] The beneficial effect of the above further scheme is: in the present invention, when horizontally adjusting the calibration wind turbine image, it is necessary to first determine whether the angle is less than 45°. If it is less than 45°, only a simple rotation is required. If it is greater than or equal to 45°, the calibration wind turbine image needs to be cropped to avoid the influence of redundant pixels.

[0022] Furthermore, S2 includes the following sub-steps:

[0023] S21. Randomly select a pixel point from the standard wind turbine operating image as the center point, and calculate characteristic distance values ​​between the remaining pixel points and the center point;

[0024] S22, determining pixel characteristic values ​​of the standard wind turbine operating image based on characteristic distance values ​​between the remaining pixel points and the center point;

[0025] S23 . Determine a blade region of the standard wind turbine generator set operation image according to pixel feature values ​​of the standard wind turbine generator set operation image, and generate a standard blade operation image.

[0026] The beneficial effect of the above further scheme is: in the present invention, a pixel point of the standard wind turbine operation image is randomly selected as the center point, which can truly reflect the characteristics of each pixel point, and then the contour of the blade area is extracted based on the mathematical operation of the characteristic distance value and the pixel characteristic value. The standard blade operation image thus generated can fully consider the distribution of pixel points, and its contour extraction is accurate without redundant blank areas.

[0027] Furthermore, the calculation formula of the characteristic distance value S between the pixel point and the center point is:

[0028]

[0029] Where A0 represents the area occupied by the center point in the standard wind turbine operation image, A i represents the area occupied by the i-th pixel in the standard wind turbine operation image, a represents the length of the standard wind turbine operation image, b represents the width of the standard wind turbine operation image, and li0 represents the Euclidean distance between the i-th pixel and the center point.

[0030] Furthermore, in S22, the calculation formula of the pixel characteristic value Z of the standard wind turbine operating image is:

[0031]

[0032] Where N represents the number of pixels in the standard wind turbine operating image, S i represents the characteristic distance between the i-th pixel and the center point in the standard wind turbine operation image, S ave Represents the mean of all feature distance values ​​in the standard wind turbine operation image.

[0033] Furthermore, in S23 , a specific method for determining the blade area of ​​the standard wind turbine operating image is: taking a set of pixel points corresponding to grayscale values ​​greater than the pixel characteristic value as the blade area.

[0034] Furthermore, the wind turbine fault diagnosis model includes an input layer, a first encoding block, a second encoding block, a first decoding block, a second decoding block, a first convolution block, a second convolution block, a first splicing block, a second splicing block, a fully connected block, and an output layer;

[0035] The input layer is used to input the standard blade operation image into the wind turbine fault diagnosis model; the first output end of the input layer is connected to the input end of the first coding block, and the second output end of the input layer is connected to the input end of the second coding block;

[0036] The output end of the first encoding block is connected to the input end of the first decoding block; the output end of the second encoding block is connected to the input end of the second decoding block;

[0037] The output end of the first decoding block and the output end of the first convolution block are respectively connected to the first input end and the second input end of the first splicing block in a one-to-one correspondence; the output end of the second decoding block and the output end of the second convolution block are respectively connected to the first input end and the second input end of the second splicing block in a one-to-one correspondence;

[0038] The output end of the first splicing block and the output end of the second splicing block are respectively connected to the first input end and the second input end of the fully connected block in a one-to-one correspondence;

[0039] The output end of the fully connected block is connected to the input end of the output layer; the output layer is used to output the fault diagnosis results of the wind turbine.

[0040] The beneficial effect of the above further scheme is: in the present invention, the input layer inputs the standard blade running image into multiple coding blocks for encoding to obtain a coding feature map, and then jump-connects the coding feature map with the convolution block, processes the coding feature map and then splices it; the fully connected block integrates the results of splicing each splicing block and outputs the result.

[0041] Furthermore, the loss function X of the fully connected layer is expressed as:

[0042]

[0043] Where Y1 represents the output feature of the first splicing block, Y2 represents the output feature of the second splicing block, b represents the weight coefficient of the fully connected layer, sigmoid(·) represents the S-type activation function, K represents the number of neurons in the fully connected layer, and λ k represents the weight of the kth neuron in the fully connected layer.

[0044] The beneficial effects of the present invention are: the wind turbine fault diagnosis method collects and preprocesses the wind turbine operation image to ensure that the image is high in clarity and free of noise interference; the blade area of ​​the wind turbine operation image is extracted, and only the blade area is processed in subsequent steps, which can greatly reduce the algorithm complexity and algorithm flow; on the other hand, the wind turbine fault diagnosis model is constructed through a neural network, which can improve the applicability of the method, has high diagnostic accuracy, and has strong fault tolerance, can timely discover the operation fault of the wind turbine and ensure the safe operation of the wind turbine. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is a flow chart of a wind turbine fault diagnosis method;

[0046] Figure 2 This is the structural diagram of the wind turbine fault diagnosis model. DETAILED DESCRIPTION

[0047] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0048] like Figure 1 As shown, the present invention provides a wind turbine fault diagnosis method, comprising the following steps:

[0049] S1: collecting real-time wind turbine operation images, and preprocessing the real-time wind turbine operation images to generate standard wind turbine operation images;

[0050] S2: determining the blade area of ​​the standard wind turbine operating image and generating the standard blade operating image;

[0051] S3: Construct a wind turbine fault diagnosis model, input the standard blade operation image into the wind turbine fault diagnosis model, and determine the wind turbine fault diagnosis result.

[0052] In this embodiment of the present invention, S1 includes the following sub-steps:

[0053] S11, collecting real-time wind turbine operation images, performing grayscale calibration on the real-time wind turbine operation images, and obtaining calibrated wind turbine operation images;

[0054] S12, rotating the calibration wind turbine operating image to obtain a horizontal wind turbine operating image;

[0055] S13. Perform normalization and denoising processing on the horizontal wind turbine operating image to obtain a standard wind turbine operating image.

[0056] Because wind turbines are located outdoors, the captured images of wind turbine operation are subject to significant distortion and noise, necessitating image preprocessing. To address the potential for uneven exposure in wind turbine operation images, grayscale calibration is required to achieve uniform grayscale across the entire image, generating a calibrated wind turbine operation image. Calibrated wind turbine operation images may exhibit directional offsets, making it easy to later classify images that do not belong to the blade area as belonging to the blade area, resulting in inaccurate cropping of the standard blade operation image. Therefore, rotation correction is required to maintain a level image.

[0057] In the embodiment of the present invention, in S11, the calculation formula for the grayscale value μ of the pixel in the calibrated wind turbine operation image is:

[0058]

[0059] Where H Rmax Indicates the maximum pixel value of the real-time wind turbine operation image in the R channel, H Gmax Indicates the maximum pixel value of the real-time wind turbine operating image in the G channel, H Bmax represents the maximum pixel value of the real-time wind turbine operating image in channel B, h1 represents the maximum grayscale value of the real-time wind turbine operating image, h2 represents the minimum grayscale value of the real-time wind turbine operating image, and h ave Indicates the average grayscale value of the real-time wind turbine operation image.

[0060] In this embodiment of the present invention, S12 includes the following sub-steps:

[0061] S121, connecting the pixel point corresponding to the maximum grayscale value and the pixel point corresponding to the minimum grayscale value in the calibration wind turbine image to form a pixel point connection line;

[0062] S122, determine whether the angle between the pixel line and the horizontal direction is less than 45 degrees, if so, proceed to S123, otherwise proceed to S124;

[0063] S123, rotating the calibration wind turbine image until the angle between the pixel line and the horizontal direction is 0°, thereby generating a horizontal wind turbine operation image;

[0064] S124. After rotating the calibration wind turbine image until the angle between the pixel line and the horizontal direction is 45°, crop the calibration wind turbine image using a cropping frame, and rotate the cropped calibration wind turbine image until the angle between the pixel line and the horizontal direction is 0°, thereby generating a horizontal wind turbine operation image.

[0065] When horizontally adjusting the calibration wind turbine image, it is necessary to first determine whether the angle is less than 45°. If it is less than 45°, a simple rotation is sufficient. If it is greater than or equal to 45°, the calibration wind turbine image needs to be cropped to avoid the influence of redundant pixels.

[0066] In this embodiment of the present invention, S2 includes the following sub-steps:

[0067] S21. Randomly select a pixel point from the standard wind turbine operating image as the center point, and calculate characteristic distance values ​​between the remaining pixel points and the center point;

[0068] S22, determining pixel characteristic values ​​of the standard wind turbine operating image based on characteristic distance values ​​between the remaining pixel points and the center point;

[0069] S23 . Determine a blade region of the standard wind turbine generator set operation image according to pixel feature values ​​of the standard wind turbine generator set operation image, and generate a standard blade operation image.

[0070] Randomly selecting a pixel point in the standard wind turbine operation image as the center point can truly reflect the characteristics of each pixel point. Then, based on the mathematical operation of the characteristic distance value and the pixel characteristic value, the contour of the blade area is extracted. The standard blade operation image thus generated can fully consider the distribution of pixel points, and its contour extraction is accurate without redundant blank areas.

[0071] In the embodiment of the present invention, the calculation formula of the characteristic distance value S between the pixel point and the center point is:

[0072]

[0073] Where A0 represents the area occupied by the center point in the standard wind turbine operation image, A i represents the area occupied by the i-th pixel in the standard wind turbine operation image, a represents the length of the standard wind turbine operation image, b represents the width of the standard wind turbine operation image, and li0 represents the Euclidean distance between the i-th pixel and the center point.

[0074] In the embodiment of the present invention, in S22, the calculation formula of the pixel characteristic value Z of the standard wind turbine operating image is:

[0075]

[0076] Where N represents the number of pixels in the standard wind turbine operating image, S i represents the characteristic distance between the i-th pixel and the center point in the standard wind turbine operation image, S ave Represents the mean of all feature distance values ​​in the standard wind turbine operation image.

[0077] In the embodiment of the present invention, in S23 , the specific method for determining the blade area of ​​the standard wind turbine operating image is: taking a set of pixel points corresponding to grayscale values ​​greater than the pixel characteristic value as the blade area.

[0078] In the embodiment of the present invention, Figure 2 As shown, the wind turbine fault diagnosis model includes an input layer, a first encoding block, a second encoding block, a first decoding block, a second decoding block, a first convolution block, a second convolution block, a first splicing block, a second splicing block, a fully connected block and an output layer;

[0079] The input layer is used to input the standard blade operation image into the wind turbine fault diagnosis model; the first output end of the input layer is connected to the input end of the first coding block, and the second output end of the input layer is connected to the input end of the second coding block;

[0080] The output end of the first encoding block is connected to the input end of the first decoding block; the output end of the second encoding block is connected to the input end of the second decoding block;

[0081] The output end of the first decoding block and the output end of the first convolution block are respectively connected to the first input end and the second input end of the first splicing block in a one-to-one correspondence; the output end of the second decoding block and the output end of the second convolution block are respectively connected to the first input end and the second input end of the second splicing block in a one-to-one correspondence;

[0082] The output end of the first splicing block and the output end of the second splicing block are respectively connected to the first input end and the second input end of the fully connected block in a one-to-one correspondence;

[0083] The output end of the fully connected block is connected to the input end of the output layer; the output layer is used to output the fault diagnosis results of the wind turbine.

[0084] The input layer inputs the standard blade running image into multiple encoding blocks for encoding to obtain the encoding feature map, and then jump-connects the encoding feature map with the convolution block, processes the encoding feature map and then splices it; the fully connected block integrates the results of the splicing of each splicing block and outputs the result.

[0085] In this embodiment of the present invention, the loss function X of the fully connected layer is expressed as:

[0086]

[0087] Where Y1 represents the output feature of the first splicing block, Y2 represents the output feature of the second splicing block, b represents the weight coefficient of the fully connected layer, sigmoid(·) represents the S-type activation function, K represents the number of neurons in the fully connected layer, and λ k represents the weight of the kth neuron in the fully connected layer.

[0088] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. A wind turbine fault diagnosis method, characterized in that: The following steps are involved: S1: collecting real-time wind turbine operation images, and preprocessing the real-time wind turbine operation images to generate standard wind turbine operation images; S2: determining the blade area of ​​the standard wind turbine operating image and generating the standard blade operating image; S3: constructing a wind turbine fault diagnosis model, inputting the standard blade operation image into the wind turbine fault diagnosis model, and determining the wind turbine fault diagnosis result; The S1 comprises the following sub-steps: S11, collecting real-time wind turbine operation images, performing grayscale calibration on the real-time wind turbine operation images, and obtaining calibrated wind turbine operation images; S12, rotating the calibration wind turbine operating image to obtain a horizontal wind turbine operating image; S13, performing normalization and denoising processing on the horizontal wind turbine operating image to obtain a standard wind turbine operating image; In the above S11, the calculation formula for the grayscale value μ of the pixel in the calibrated wind turbine operating image is: Where H Rmax Indicates the maximum pixel value of the real-time wind turbine operation image in the R channel, H Gmax Indicates the maximum pixel value of the real-time wind turbine operating image in the G channel, H Bmax represents the maximum pixel value of the real-time wind turbine operating image in channel B, h1 represents the maximum grayscale value of the real-time wind turbine operating image, h2 represents the minimum grayscale value of the real-time wind turbine operating image, and h ave Indicates the average grayscale value of the real-time wind turbine operation image; The S2 includes the following sub-steps: S21. Randomly select a pixel point from the standard wind turbine operating image as the center point, and calculate characteristic distance values ​​between the remaining pixel points and the center point; S22, determining pixel characteristic values ​​of the standard wind turbine operating image based on characteristic distance values ​​between the remaining pixel points and the center point; S23, determining a blade region of the standard wind turbine operating image according to pixel feature values ​​of the standard wind turbine operating image, and generating a standard blade operating image; The calculation formula of the characteristic distance value S between the pixel point and the center point is: Where A0 represents the area occupied by the center point in the standard wind turbine operation image, A i represents the area occupied by the i-th pixel in the standard wind turbine operating image, a represents the length of the standard wind turbine operating image, b represents the width of the standard wind turbine operating image, l i0 Represents the Euclidean distance between the i-th pixel and the center point.

2. The wind turbine fault diagnosis method according to claim 1, characterized in that: The S12 includes the following sub-steps: S121, connecting the pixel point corresponding to the maximum grayscale value and the pixel point corresponding to the minimum grayscale value in the calibration wind turbine operation image to form a pixel point connection line; S122, determine whether the angle between the pixel line and the horizontal direction is less than 45 degrees, if so, proceed to S123, otherwise proceed to S124; S123, rotating the calibration wind turbine operation image until the angle between the pixel line and the horizontal direction is 0°, thereby generating a horizontal wind turbine operation image; S124. After rotating the calibration wind turbine operation image until the angle between the pixel point line and the horizontal direction is 45°, use the cropping frame to crop the calibration wind turbine operation image, and rotate the cropped calibration wind turbine operation image until the angle between the pixel point line and the horizontal direction is 0°, thereby generating a horizontal wind turbine operation image.

3. The wind turbine fault diagnosis method according to claim 1, characterized in that: In the above S23 , the specific method for determining the blade area of ​​the standard wind turbine operating image is: taking a set of pixel points corresponding to grayscale values ​​greater than the pixel characteristic value as the blade area.

4. The wind turbine fault diagnosis method according to claim 1, characterized in that: The wind turbine fault diagnosis model includes an input layer, a first encoding block, a second encoding block, a first decoding block, a second decoding block, a first convolution block, a second convolution block, a first splicing block, a second splicing block, a fully connected block and an output layer; The input layer is used to input the standard blade operation image into the wind turbine fault diagnosis model; the first output end of the input layer is connected to the input end of the first coding block, and the second output end thereof is connected to the input end of the second coding block; An output terminal of the first encoding block is connected to an input terminal of the first decoding block; An output terminal of the second encoding block is connected to an input terminal of the second decoding block; The output end of the first decoding block and the output end of the first convolution block are respectively connected to the first input end and the second input end of the first splicing block in a one-to-one correspondence; The output end of the second decoding block and the output end of the second convolution block are respectively connected to the first input end and the second input end of the second splicing block in a one-to-one correspondence; The output end of the first splicing block and the output end of the second splicing block are respectively connected to the first input end and the second input end of the fully connected block in a one-to-one correspondence; The output end of the fully connected block is connected to the input end of the output layer; the output layer is used to output the fault diagnosis result of the wind turbine generator set.

5. The wind turbine fault diagnosis method according to claim 4, characterized in that: The expression of the loss function X of the fully connected block is: Where Y1 represents the output feature of the first splicing block, Y2 represents the output feature of the second splicing block, b represents the weight coefficient of the fully connected layer, sigmoid(·) represents the S-type activation function, K represents the number of neurons in the fully connected layer, and λ k represents the weight of the kth neuron in the fully connected layer.

Images