A wind power equipment analysis system based on AI vision assistance

Abstract

This invention provides an AI-based visually assisted wind turbine analysis system, comprising an image acquisition module, an image processing module, a feature extraction module, and a fault diagnosis module. The image acquisition module is used to acquire image information of the wind turbine. The image processing module preprocesses the acquired image information to improve image quality. The feature extraction module extracts features from the preprocessed image. The fault diagnosis module uses an AI algorithm to perform fault diagnosis based on the feature extraction results. This system can capture images of wind turbines, classify them, output features, and then perform intelligent analysis based on the feature output to obtain accurate diagnostic results.

Description

Technical Field

[0001] The present invention relates to the field of wind power generation, and in particular to a wind power equipment analysis system based on AI vision assistance. Background Art

[0002] With the rapid development of the wind power industry, the reliability and operational efficiency of wind power equipment have become bottlenecks in the industry's development. Traditional wind power equipment monitoring and maintenance methods have problems such as low efficiency and high labor costs. Therefore, the development of an AI-based vision-assisted wind power equipment analysis system is of great significance for improving the reliability and operational efficiency of wind power equipment.

[0003] The foregoing discussion of the background art is intended only to facilitate an understanding of the present invention. This discussion does not acknowledge or admit that any of the material referred to is part of the common general knowledge.

[0004] Numerous wind turbine equipment analysis systems have been developed. After extensive research and reference, we discovered existing analysis systems, such as the one disclosed in publication number CN112879219B. These systems generally include the following steps: 1) acquiring a blade profile image; 2) extracting blade edge information at horizontal positions within the blade profile image, and calculating the distance between two adjacent blade edges on the same horizontal line based on this blade edge information; 3) calculating the deviation rate between each pair of adjacent blades based on the distance between the two adjacent blade edges on the same horizontal line; and 4) determining whether the deviation rate between each pair of adjacent blades is greater than or equal to a preset threshold. However, this system can only analyze blades in wind turbines, is not universal, and uses a single analysis method, resulting in analysis results for only one type of fault. Summary of the Invention

[0005] The purpose of the present invention is to address the existing deficiencies and propose a wind power equipment analysis system based on AI vision assistance.

[0006] The present invention adopts the following technical solutions:

[0007] An AI-assisted vision-assisted wind power equipment analysis system includes an image acquisition module, an image processing module, a feature extraction module, and a fault diagnosis module;

[0008] The image acquisition module is used to acquire image information of the wind power equipment, the image processing module preprocesses the acquired image information to improve image quality, the feature extraction module extracts features from the preprocessed image, and the fault diagnosis module uses an AI algorithm to perform fault diagnosis based on the feature extraction results;

[0009] The image processing module includes a grayscale extraction unit, a first storage unit, a filtering processing unit, and a second storage unit. The grayscale extraction unit is used to extract grayscale information from the captured image. The first storage unit is used to store the extracted grayscale information. The filtering processing unit performs filtering processing on the grayscale information. The second storage unit is used to store the filtered grayscale information.

[0010] The feature extraction module includes an edge processing unit, a classification unit, and an internal feature unit. The edge processing unit processes the grayscale information in the second storage unit to obtain edge information. The classification unit classifies the wind turbines in the image according to the edge information. The internal feature unit performs feature processing on the grayscale information within the edge information and outputs a feature matrix.

[0011] The diagnostic model used by the fault diagnosis module includes multiple unit processors. Each unit processor obtains an element from a specific position in the feature matrix and performs a specific calculation. The specific position and specific calculation method are recorded in the unit processor. Different unit processors record different specific positions and specific calculation methods. The unit processor outputs 0 or 1. The diagnostic model processes the output values ​​of all unit processors to obtain a result value Rs, and obtains the corresponding diagnostic result according to the interval in which the result value falls.

[0012] Furthermore, the image acquisition module includes a camera, an image transmission device, a power supply device and a controller. The camera is used to directly capture images of the wind turbine. The image transmission device is used to transmit the images captured by the camera to a back-end server. The power supply device is used to provide power support to the camera and the image transmission device. The controller is used to control and manage the camera and the image transmission device.

[0013] Furthermore, the internal feature unit calls the corresponding feature algorithm according to the type of device, and constructs a feature matrix based on the feature algorithm. The element in the i-th row and j-th column of the feature matrix is ​​recorded as GL(i, j). The matrix elements are obtained by processing according to the following formula:

[0014]

[0015] Among them, R is the pixel coordinate set inside the edge information, d is the feature distance, G0 is the feature grayscale difference, and the feature distance and feature grayscale difference are different in different feature algorithms;

[0016] For the elements in the matrix that cannot be obtained by the above formula, fill them with 0;

[0017] Furthermore, the internal feature unit performs a zero division operation on the feature matrix, and the zero division operation includes the following steps:

[0018] S31, determine whether the elements of the first row or the last row are all 0, if yes, go to step S32, otherwise directly exit the operation;

[0019] S32: Determine whether the elements in the first column or the last column are all 0. If yes, proceed to step S33; otherwise, exit the operation directly.

[0020] S33, delete the first row or last row elements that are all 0, delete the first column or last column elements that are all 0, and return to step S31;

[0021] Furthermore, the diagnostic model calculates the result value Rs according to the following formula:

[0022]

[0023] Among them, unit is the number of the unit processor, Out(unit) is the output value of the corresponding unit processor, and μ(unit) is the associated parameter of the corresponding unit processor;

[0024] The fault diagnosis module includes an execution unit and a training unit. The training unit is trained by feeding a large number of feature matrices to optimize and adjust μ(unit), and the execution unit performs diagnostic processing based on the optimized μ(unit).

[0025] The beneficial effects achieved by the present invention are:

[0026] This system improves image quality through filtering preprocessing, laying the foundation for subsequent analysis. This system classifies devices in the image through edge detection. When obtaining the feature matrix, the algorithms corresponding to different classifications contain different parameters, making the obtained feature matrix unique. This system processes the feature matrix through AI algorithms to obtain different diagnostic results. The mapping relationship between the diagnostic results and the feature matrix is ​​not obtained through an ordinary logical relationship, but is obtained by feeding a large amount of training data and is comprehensive.

[0027] To further understand the features and technical contents of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are only for reference and illustration and are not intended to limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic diagram of the overall structural framework of the present invention;

[0029] Figure 2 This is a schematic diagram of the image acquisition module of the present invention;

[0030] Figure 3 This is a schematic diagram of the image processing module of the present invention;

[0031] Figure 4 This is a schematic diagram of the feature extraction module of the present invention;

[0032] Figure 5 This is a schematic diagram of the fault diagnosis module of the present invention. DETAILED DESCRIPTION

[0033] The following is an explanation of the embodiments of the present invention through specific embodiments. Those skilled in the art can understand the advantages and effects of the present invention from the contents disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. In addition, the drawings of the present invention are only simple schematic illustrations and are not depicted in actual size. It is stated in advance. The following embodiments will further explain the relevant technical content of the present invention in detail, but the disclosed content is not intended to limit the scope of protection of the present invention.

[0034] Example 1.

[0035] This embodiment provides a wind power equipment analysis system based on AI vision assistance, combined with Figure 1 , including image acquisition module, image processing module, feature extraction module and fault diagnosis module;

[0036] The image acquisition module is used to acquire image information of the wind power equipment, the image processing module preprocesses the acquired image information to improve image quality, the feature extraction module extracts features from the preprocessed image, and the fault diagnosis module uses an AI algorithm to perform fault diagnosis based on the feature extraction results;

[0037] The image processing module includes a grayscale extraction unit, a first storage unit, a filtering processing unit, and a second storage unit. The grayscale extraction unit is used to extract grayscale information from the captured image. The first storage unit is used to store the extracted grayscale information. The filtering processing unit performs filtering processing on the grayscale information. The second storage unit is used to store the filtered grayscale information.

[0038] The feature extraction module includes an edge processing unit, a classification unit, and an internal feature unit. The edge processing unit processes the grayscale information in the second storage unit to obtain edge information. The classification unit classifies the wind turbines in the image according to the edge information. The internal feature unit performs feature processing on the grayscale information within the edge information and outputs a feature matrix.

[0039] The diagnostic model used by the fault diagnosis module includes multiple unit processors. Each unit processor obtains an element from a specific position in the feature matrix and performs a specific calculation. The specific position and specific calculation method are recorded in the unit processor. Different unit processors record different specific positions and specific calculation methods. The unit processor outputs 0 or 1. The diagnostic model processes the output values ​​of all unit processors to obtain a result value Rs, and obtains the corresponding diagnostic result according to the interval in which the result value falls.

[0040] The image acquisition module includes a camera, an image transmission device, a power supply device and a controller. The camera is used to directly capture images of the wind turbine. The image transmission device is used to transmit the images captured by the camera to the back-end server. The power supply device is used to provide power support to the camera and the image transmission device. The controller is used to control and manage the camera and the image transmission device.

[0041] The internal feature unit calls the corresponding feature algorithm according to the type of device and constructs a feature matrix based on the feature algorithm. The element in the i-th row and j-th column of the feature matrix is ​​recorded as GL(i, j). The matrix elements are obtained by processing according to the following formula:

[0042]

[0043] Among them, R is the pixel coordinate set inside the edge information, d is the feature distance, G0 is the feature grayscale difference, and the feature distance and feature grayscale difference are different in different feature algorithms;

[0044] For the elements in the matrix that cannot be obtained by the above formula, fill them with 0;

[0045] The internal feature unit performs a zero division operation on the feature matrix, and the zero division operation includes the following steps:

[0046] S31, determine whether the elements of the first row or the last row are all 0, if yes, go to step S32, otherwise directly exit the operation;

[0047] S32: Determine whether the elements in the first column or the last column are all 0. If yes, proceed to step S33; otherwise, exit the operation directly.

[0048] S33, delete the first row or last row elements that are all 0, delete the first column or last column elements that are all 0, and return to step S31;

[0049] The diagnostic model calculates the result value Rs according to the following formula:

[0050]

[0051] Among them, unit is the number of the unit processor, Out(unit) is the output value of the corresponding unit processor, and μ(unit) is the associated parameter of the corresponding unit processor;

[0052] The fault diagnosis module includes an execution unit and a training unit. The training unit is trained by feeding a large number of feature matrices to optimize and adjust μ(unit), and the execution unit performs diagnostic processing based on the optimized μ(unit).

[0053] Example 2.

[0054] This embodiment includes all the contents of the first embodiment and provides an AI-assisted wind power equipment analysis system, including an image acquisition module, an image processing module, a feature extraction module, and a fault diagnosis module;

[0055] The image acquisition module is used to acquire image information of the wind power equipment, the image processing module preprocesses and filters the acquired image information to improve image quality, the feature extraction module extracts features from the preprocessed image, and the fault diagnosis module uses an AI algorithm to perform fault diagnosis based on the feature extraction results;

[0056] Combine Figure 2 The image acquisition module includes a camera, an image transmission device, a power supply device and a controller. The camera is installed at different locations such as wind turbine blades, towers, and nacelles to directly capture images of wind turbines. The image transmission device is used to transmit images captured by the camera to a back-end server. The power supply device is used to provide power support to the camera and the image transmission device. The power supply device can obtain and store electrical energy by connecting to the output terminal of the wind turbine. The controller is used to control and manage the camera and the image transmission device, and the controller is configured or directly operated through the back-end server.

[0057] Combine Figure 3 The image processing module includes a grayscale extraction unit, a first storage unit, a filtering processing unit, and a second storage unit. The grayscale extraction unit is used to extract grayscale information from the captured image. The first storage unit is used to store the extracted grayscale information. The filtering processing unit performs filtering processing on the grayscale information. The second storage unit is used to store the filtered grayscale information.

[0058] The process of filtering processing by the filtering processing unit includes the following steps:

[0059] S1. Select a filtering point and mark its coordinates as (x0, y0);

[0060] S2. Determine the filtering range [x0-r, x0+r], [y0-r, y0+r] based on the filtering point, and record the actual number of pixels n in the filtering range. The n pixels in the filtering range constitute the filter kernel. When the filtering point is in the middle of the image, n is (2r+1). 2 When the filter point is at the edge, the filter range exceeds the boundary, resulting in n being less than (2r+1) 2 ;

[0061] S3. Calculate the weight value H(i, j) of each pixel within the filtering range according to the following formula:

[0062]

[0063] Among them, (i, j) is the coordinate of the pixel point in the filter kernel, and the pixel point on the coordinate of the filter kernel origin is the filter point. At the same time, i and j may satisfy 0<i+x0≤x max ,0<j+y0≤y max , x max is the maximum horizontal coordinate of the image pixel, y max is the maximum vertical coordinate of the image pixel, and σ is the filter standard deviation, which is set by the staff;

[0064] S4. Normalize the weight value according to the following formula:

[0065]

[0066] Among them, (i′, j′) is the coordinate of the normalized pixel point in the filter kernel, which has the same value range as (i, j);

[0067] S5. Calculate the new grayscale value Gn(x0, y0) of the filter point according to the following formula:

[0068]

[0069] Wherein, Go(x, y) represents the original grayscale value of the pixel with coordinates (x, y);

[0070] Combine Figure 4 The feature extraction module includes an edge processing unit, a classification unit, and an internal feature unit. The edge processing unit processes the grayscale information in the second storage unit to obtain edge information, wherein the edge information is the coordinate information of the edge pixel points. The classification unit classifies the wind turbines in the image according to the edge information. The internal feature unit performs feature processing on the grayscale information within the edge information.

[0071] The process of the classification unit classifying wind power equipment according to edge information includes the following steps:

[0072] S21. Calculate the distance between every two edge pixels and record the maximum distance as d max ;

[0073] S22. Taking an edge pixel as an object, sort the distances associated with it from small to large, and calculate the classification feature value of the edge pixel according to the following formula:

[0074]

[0075] Where n1 is a value less than λ·d max The number of distances, n2 is not less than λ·d max The number of distances, λ is the classification feature coefficient;

[0076] S23, sort the classification feature values ​​of all edge pixels from small to large, and divide these classification feature values ​​into k intervals, and count the number of classification feature values ​​in each interval {m k}, and obtain the classification feature vector Fv:

[0077] Fv=(m1, m2, m3,···,m k );

[0078] S24, comparing the calculated classification feature vector with the stored classification feature vectors of each wind power device to determine the device type corresponding to the edge information;

[0079] The internal feature unit calls a corresponding feature algorithm according to the type of the device, and processes the grayscale information inside the edge information using the feature algorithm;

[0080] The internal feature unit constructs a feature matrix based on the feature algorithm. The element in the i-th row and j-th column of the feature matrix is ​​recorded as GL(i, j). The matrix elements are obtained by processing according to the following formula:

[0081]

[0082] Among them, R is the pixel coordinate set inside the edge information, d is the feature distance, G0 is the feature grayscale difference, and the feature distance and feature grayscale difference are different in different feature algorithms;

[0083] For the elements in the matrix that cannot be obtained by the above formula, fill them with 0;

[0084] The internal feature unit performs a zero division operation on the feature matrix, and the zero division operation includes the following steps:

[0085] S31, determine whether the elements of the first row or the last row are all 0, if yes, go to step S32, otherwise directly exit the operation;

[0086] S32: Determine whether the elements in the first column or the last column are all 0. If yes, proceed to step S33; otherwise, exit the operation directly.

[0087] S33, delete the first row or last row elements that are all 0, delete the first column or last column elements that are all 0, and return to step S31;

[0088] The internal feature unit sends the new feature matrix to the fault diagnosis module;

[0089] Combine Figure 5 The fault diagnosis module includes an execution unit and a training unit. The training unit is trained by feeding a large number of feature matrices to optimize the matching accuracy between the feature matrices and the diagnosis results. The execution unit performs diagnostic processing on the feature matrices based on the optimized parameters to obtain a diagnosis result.

[0090] The diagnostic model used by the execution unit and the training unit includes multiple unit processors. Each unit processor obtains an element from a specific position in the feature matrix and performs a specific calculation process. The specific position and the specific calculation method are recorded in the unit processor. Different unit processors record different specific positions and specific calculation methods. The unit processor outputs 0 or 1. The diagnostic model calculates the result value Rs according to the following formula:

[0091]

[0092] Among them, unit is the number of the unit processor, Out(unit) is the output value of the corresponding unit processor, and μ(unit) is the associated parameter of the corresponding unit processor;

[0093] Different diagnostic results have different result intervals. When Rs falls into the corresponding result interval, it means that the corresponding diagnostic result is obtained;

[0094] The training unit increases the probability that the result value falls into the correct interval by adjusting μ(unit);

[0095] The execution unit uses a set of μ (unit) with the highest accuracy to calculate the result value, and obtains a corresponding diagnosis result according to the interval into which the result value falls.

[0096] The contents disclosed above are only preferred feasible embodiments of the present invention and do not limit the scope of protection of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention description and drawings are included in the scope of protection of the present invention. In addition, the elements therein can be updated as technology develops.

Claims

1. A wind power equipment analysis system based on AI vision assistance, characterized in that: It includes image acquisition module, image processing module, feature extraction module and fault diagnosis module; The image acquisition module is used to acquire image information of the wind power equipment, the image processing module preprocesses the acquired image information to improve image quality, the feature extraction module extracts features from the preprocessed image, and the fault diagnosis module uses an AI algorithm to perform fault diagnosis based on the feature extraction results; The image processing module includes a grayscale extraction unit, a first storage unit, a filtering processing unit, and a second storage unit. The grayscale extraction unit is used to extract grayscale information from the captured image. The first storage unit is used to store the extracted grayscale information. The filtering processing unit performs filtering processing on the grayscale information. The second storage unit is used to store the filtered grayscale information. The feature extraction module includes an edge processing unit, a classification unit, and an internal feature unit. The edge processing unit processes the grayscale information in the second storage unit to obtain edge information. The classification unit classifies the wind turbines in the image according to the edge information. The internal feature unit performs feature processing on the grayscale information within the edge information and outputs a feature matrix. The diagnostic model used by the fault diagnosis module includes multiple unit processors. Each unit processor obtains elements from a specific position in the feature matrix and performs specific calculation processing. The specific position and specific calculation method are recorded in the unit processor. Different unit processors record different specific positions and specific calculation methods. The unit processor outputs 0 or 1. The diagnostic model obtains the result value based on the output values ​​of all unit processors. , according to the interval in which the result value falls, the corresponding diagnosis result is obtained; The internal feature unit calls the corresponding feature algorithm according to the type of device and constructs a feature matrix based on the feature algorithm. The element in the i-th row and j-th column of the feature matrix is ​​recorded as , the matrix elements are processed according to the following formula: ； Among them, R is the pixel coordinate set inside the edge information, d is the feature distance, is the feature grayscale difference. The feature distance and feature grayscale difference in different feature algorithms are different. For elements in the matrix that cannot be obtained by the above formula, fill them with 0.

2. The AI ​​visually assisted wind power equipment analysis system according to claim 1, characterized in that: The image acquisition module includes a camera, an image transmission device, a power supply device and a controller. The camera is used to directly capture images of wind power equipment. The image transmission device is used to transmit the images captured by the camera to a back-end server. The power supply device is used to provide power support to the camera and the image transmission device. The controller is used to control and manage the camera and the image transmission device.

3. The AI ​​visually assisted wind power equipment analysis system according to claim 2, characterized in that: The internal feature unit performs a zero division operation on the feature matrix, and the zero division operation includes the following steps: S31, determine whether the elements of the first row or the last row are all 0, if yes, go to step S32, otherwise directly exit the operation; S32: Determine whether the elements in the first column or the last column are all 0. If yes, proceed to step S33; otherwise, exit the operation directly. S33. Delete the first row or last row elements that are all 0, delete the first column or last column elements that are all 0, and return to step S31.

4. The AI ​​visually assisted wind power equipment analysis system according to claim 3, characterized in that: The diagnostic model calculates the result value according to the following formula : ； Among them, unit is the number of the unit processor, Out(unit) is the output value of the corresponding unit processor, and μ(unit) is the associated parameter of the corresponding unit processor; The fault diagnosis module includes an execution unit and a training unit. The training unit is trained by feeding a large number of feature matrices to optimize and adjust μ(unit), and the execution unit performs diagnostic processing based on the optimized μ(unit).

Images