Fan blade appearance intelligent monitoring method and device

Abstract

The embodiment of the invention provides an intelligent monitoring method and device for the appearance of a fan blade, and the method comprises the steps: triggering a camera device to carry out the automatic snapshot based on a preset condition, obtaining image data in a rotation period of the blade, and carrying out the preprocessing of the image data; performing multi-dimensional comparative analysis on the preprocessed image data and historical image data in a preset dynamic reference database, and outputting a comparison result; performing anomaly recognition on the comparison result by adopting an AI algorithm, and judging whether the blade has appearance anomaly or not and the anomaly level; and triggering a corresponding alarm mechanism according to the abnormal level, and performing linkage to generate an operation and maintenance work order, so as to track a maintenance process and a repair effect through the operation and maintenance work order. According to the method, traditional manual climbing inspection can be replaced, normalized and all-weather monitoring of the blade appearance is achieved, and the monitoring efficiency, the operation and maintenance response speed and the anomaly recognition accuracy are improved.

Description

Technical Field

[0001] This application relates to the field of wind power equipment condition monitoring technology, and more specifically, to a method and device for intelligent monitoring of the appearance of wind turbine blades. Background Technology

[0002] In the field of wind power generation, wind turbine blades, as the core components for energy conversion, are exposed to the complex outdoor environment for extended periods. They are susceptible to erosion from wind and sand, erosion from rain, extreme temperatures, and strong wind loads, resulting in appearance abnormalities such as cracks, corrosion, coating peeling, and edge damage. If these abnormalities are not detected and addressed in a timely manner, they can lead to a decrease in wind turbine power generation efficiency, or even cause blade breakage, turbine shutdown, and other safety accidents, resulting in huge economic losses.

[0003] Currently, the inspection of wind turbine blade appearance mainly relies on manual inspection: maintenance personnel need to use climbing equipment (such as ladders or drones) to approach the blades and inspect them by visual observation or by taking pictures with handheld devices. This method has significant drawbacks: First, the inspection cycle is long, making it difficult to achieve real-time monitoring and easily missing the best time to deal with early minor anomalies; second, manual inspection is greatly affected by the environment, and cannot be carried out in severe weather (such as heavy rain, strong winds, and dense fog), and there are safety risks associated with working at heights; third, human judgment relies on experience, is highly subjective, and is prone to omissions or misjudgments due to fatigue or lack of experience; fourth, the inspection cost is high, requiring a large investment of manpower and equipment resources, especially for remote wind farms, where the operation and maintenance efficiency is extremely low.

[0004] Therefore, how to solve the above problems is an urgent issue that needs to be addressed. Summary of the Invention

[0005] This application provides a method and device for intelligent monitoring of the appearance of wind turbine blades, aiming to improve the above-mentioned problems.

[0006] Firstly, this application provides a method for intelligent monitoring of the appearance of wind turbine blades, the method comprising: The camera device is automatically captured based on preset conditions to obtain image data within the blade rotation cycle, and the image data is preprocessed. The camera device is deployed at 20-50 meters from the wind turbine tower base and on the top of the nacelle. The preprocessed image data is compared and analyzed in multiple dimensions with historical image data in a preset dynamic benchmark database, and the comparison results are output. AI algorithms are used to identify anomalies in the comparison results and determine whether the blades have any appearance abnormalities and the level of abnormality. The corresponding alarm mechanism is triggered based on the anomaly level, and a maintenance work order is generated accordingly, so as to track the inspection process and repair effect through the maintenance work order.

[0007] In one possible embodiment, the camera device is an 8K ultra-high-definition ultra-wide-angle camera; wherein, 2-3 of the 8K ultra-high-definition ultra-wide-angle cameras are installed in layers at a height of 20-50 meters from the tower base, and the field of view of each of the 8K ultra-high-definition ultra-wide-angle cameras is 120°, and multiple of the 8K ultra-high-definition ultra-wide-angle cameras work together to form a 360° shooting range without blind spots; The 8K ultra-high-definition ultra-wide-angle camera installed on the top of the cabin integrates image stabilization and an infrared fill light module. The 8K ultra-high-definition ultra-wide-angle camera on the top of the cabin is also equipped with a miniature wind power generation unit to power the 8K ultra-high-definition ultra-wide-angle camera and the infrared fill light module.

[0008] In one possible embodiment, triggering the camera device to automatically capture images based on preset conditions includes: When the blade rotation speed stabilizes within the normal operating range of 10-20 rpm, the camera device is automatically triggered to capture images every hour, continuously taking 10-15 images during each capture to ensure coverage of the complete blade rotation cycle; or, When the wind speed exceeds 15 m / s or the change in light intensity exceeds a threshold, the capture interval of the camera device is shortened to once every 30 minutes, and exposure compensation processing is performed on the captured image data during the capture process to ensure image clarity under special conditions.

[0009] In one possible embodiment, the preprocessing of the image data includes: The Gaussian filtering algorithm is used to denoise the captured image data to obtain intermediate image data; Histogram equalization is used to enhance the contrast of blade surface details in the intermediate image data to obtain target impact data; The Canny edge detection algorithm was used to extract the contour features and surface texture features of the blades in the target influence data; The contour features and surface texture features are converted into standardized digital matrices for subsequent comparison with historical image data.

[0010] In one possible embodiment, the step of performing a multi-dimensional comparative analysis between the preprocessed image data and historical image data in a preset dynamic benchmark database, and outputting the comparison results, includes: Compare the overall outline and edge morphology of the blades in the digital matrix with those in the historical reference images to see if they are consistent. Calculate the matching degree between the digital matrix and the leaf surface texture in the historical reference image. When the texture similarity is less than 90%, it is determined that there is an anomaly in the leaf in the image data. The color difference between the digital matrix and the blade coating in the historical reference image is detected. When the color difference is greater than 5, it is determined that there is an anomaly in the blade in the image data.

[0011] In one possible embodiment, the step of using an AI algorithm to identify anomalies in the comparison results and determine whether the blade has any appearance abnormalities and the level of abnormality includes: A trained CNN model is used as the core of the AI ​​algorithm. The training samples of the CNN model include image data of various anomalies such as blade cracks, corrosion, edge damage, and foreign object adhesion. The trained CNN model has an accuracy of ≥95% in recognizing blade appearance anomalies. The CNN model is used to determine whether there are any abnormalities in the appearance of the blades based on the comparison results, and the level of abnormality is divided into three categories: minor damage, moderate damage, and severe damage.

[0012] In one possible embodiment, the method further includes: When the CNN model determines that there is an appearance abnormality in the blade, it automatically marks the abnormal area in the image and pushes the abnormal image carrying the abnormal area and the comparison result to the terminal of the operation and maintenance personnel, so that the operation and maintenance personnel can view the abnormal image and the comparison result through the terminal and manually confirm whether the abnormality is a real abnormality. If the terminal returns a false anomaly, the abnormal image is input into the CNN model for iterative optimization, thereby further improving the recognition accuracy of the CNN model.

[0013] In one possible embodiment, triggering the corresponding alarm mechanism based on the anomaly level includes: When the abnormality level is the minor damage, a level 1 alarm is triggered: an alarm prompt window pops up on the remote monitoring platform, and the abnormality location, abnormality type and severity information are automatically recorded. A monthly statistical report is generated based on the monthly abnormality data for maintenance personnel to check regularly. When the abnormality level is moderate damage or severe damage, a level 2 or level 3 alarm is triggered: in addition to popping up an alarm prompt window on the remote monitoring platform, the operation and maintenance manager is also notified via SMS and telephone voice, and the GPS location information of the wind turbine to which the abnormal blade belongs and the abnormal image are pushed simultaneously.

[0014] In one possible embodiment, the maintenance work order includes the turbine number, location, abnormality type, and suggested maintenance plan of the abnormal blade.

[0015] Secondly, this application provides an intelligent monitoring device for the appearance of wind turbine blades, the device comprising: The camera module is used to trigger the camera device to automatically capture images based on preset conditions, acquire image data within the blade rotation cycle, and preprocess the image data. The camera device is deployed at 20-50 meters from the wind turbine tower base and on the top of the nacelle. The data analysis module is used to perform multi-dimensional comparative analysis between the preprocessed image data and historical image data in the preset dynamic benchmark database, and output the comparison results. The AI ​​analysis module is used to use AI algorithms to identify anomalies in the comparison results and determine whether the blade has any appearance abnormalities and the level of abnormality. The alarm module is used to trigger the corresponding alarm mechanism according to the anomaly level and generate maintenance work orders in conjunction with it, so as to track the inspection process and repair effect according to the maintenance work orders, thereby forming a closed-loop management.

[0016] The present application provides a method and device for intelligent monitoring of wind turbine blade appearance. This application automatically captures image data within the blade's rotation cycle by triggering a camera device based on preset conditions. The image data is preprocessed, with the camera device deployed 20-50 meters above the wind turbine tower base and on the top of the nacelle. The preprocessed image data is then compared and analyzed in multiple dimensions with historical image data in a preset dynamic benchmark database, outputting the comparison results. An AI algorithm is used to identify anomalies in the comparison results, determining whether the blade exhibits any appearance abnormalities and their severity. Based on the anomaly severity, a corresponding alarm mechanism is triggered, and a maintenance work order is generated to track the repair process and its effectiveness, thus forming a closed-loop management system. This replaces traditional manual inspections from elevated positions, enabling routine, all-weather monitoring of wind turbine blade appearance, improving monitoring efficiency, maintenance response speed, and the accuracy of anomaly identification. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of an electronic device provided in the first embodiment of this application; Figure 2 A flowchart illustrating an intelligent monitoring method for the appearance of wind turbine blades provided in the second embodiment of this application; Figure 3 A schematic diagram of the functional modules of a wind turbine blade appearance intelligent monitoring device provided in the third embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] First Embodiment Figure 1 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. In this application, it can be... Figure 1 The schematic diagram shown illustrates an example of an electronic device 100 used to implement the intelligent monitoring method and apparatus for the appearance of wind turbine blades according to embodiments of this application.

[0021] like Figure 1 The diagram shows the structure of an electronic device 100, which includes one or more processors 102, one or more storage devices 104, and input devices 106. These components are interconnected via a bus system and / or other forms of connection mechanisms (not shown). It should be noted that... Figure 1 The components and structure of the electronic device 100 shown are merely exemplary and not limiting; the electronic device may have, as needed. Figure 1 The components shown may also have Figure 1 Other components and structures not shown.

[0022] The processor 102 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 100 to perform desired functions.

[0023] The storage device 104 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 102 may execute the program instructions to implement the functions described in the embodiments of this application below and / or other desired functions. Various application programs and various data may also be stored in the computer-readable storage medium, such as various data used and / or generated by the application programs.

[0024] The input device 106 may be a device used by a user to input commands, and may include one or more of the following: keyboard, mouse, microphone, and touch screen.

[0025] Second embodiment: Reference Figure 2 The flowchart shown illustrates a method for intelligent monitoring of wind turbine blade appearance. This method specifically includes the following steps: Step S201: Based on preset conditions, the camera device is automatically triggered to capture images within the blade rotation cycle, and the image data is preprocessed. The camera equipment is deployed at a height of 20-50 meters from the base of the wind turbine tower and on the top of the nacelle.

[0026] In one implementation, the camera device is an 8K ultra-high-definition ultra-wide-angle camera; wherein, 2-3 of the 8K ultra-high-definition ultra-wide-angle cameras are installed in layers at a height of 20-50 meters from the tower base, each of the 8K ultra-high-definition ultra-wide-angle cameras has a field of view coverage of 120°, and multiple 8K ultra-high-definition ultra-wide-angle cameras work together to form a 360° shooting range without blind spots; the 8K ultra-high-definition ultra-wide-angle camera installed on the top of the nacelle integrates image stabilization function and infrared fill light module, and the 8K ultra-high-definition ultra-wide-angle camera on the top of the nacelle is also equipped with a miniature wind power generation power supply unit to power the 8K ultra-high-definition ultra-wide-angle camera and the infrared fill light module.

[0027] Understandably, in this implementation, by setting up multiple cameras at the tower base with a 120° field of view in coordination, the problem of existing fixed cameras having "large blind spots and inability to cover the entire angle of the blades" can be completely solved, ensuring that every part of the blades can be clearly captured during rotation, avoiding omissions caused by missed shots. Furthermore, the image stabilization function of the nacelle camera eliminates the impact of wind turbine vibration on image clarity, the infrared supplementary lighting module solves the problem of image blurring in nighttime / low-light environments, and the micro-wind power generation reduces dependence on the main power supply of the wind turbine, improving the stability and endurance of the equipment in complex environments such as sandstorms, high temperatures, and low light. In addition, by selecting 8K ultra-high-definition resolution, early anomalies such as tiny scratches (e.g., scratches <5cm) and minor corrosion on the blade surface can be clearly presented, overcoming the deficiency of low-definition cameras "unable to identify minor damage," and buying time for early anomaly handling.

[0028] As one implementation method, the automatic image capture triggered by the camera device based on preset conditions includes: acquiring wind turbine blade rotation speed data in real time; when the blade rotation speed is stable within the normal operating range of 10-20 revolutions per minute, triggering the camera device to automatically capture images once every hour, continuously taking 10-15 images during each capture to ensure coverage of the complete rotation cycle of the blades; or, when the wind speed is detected to exceed 15 meters per second or the change in light intensity is detected to be greater than a threshold, shortening the capture interval of the camera device to once every 30 minutes, and performing exposure compensation processing on the captured image data during the capture process to ensure image clarity under special conditions.

[0029] It is understood that in this embodiment, the blade speed can be obtained from the control system used to control the wind turbine by communicating with the wind turbine control system.

[0030] Of course, in practical use, rotational speed can also be detected by acquiring images; no specific limitations are made here.

[0031] Understandably, in this embodiment, by capturing images at 1-hour intervals when the rotation speed is stable, the monitoring frequency per hour is ensured while avoiding data redundancy caused by frequent captures. In special environments, the interval is shortened to ensure image coverage during critical periods, solving the problem that "fixed-interval captures cannot adapt to environmental changes." Furthermore, capturing 10-15 images each time covers the entire rotation cycle, avoiding missed shots due to blade rotation. Exposure compensation processing addresses overexposure / underexposure issues in strong / weak light conditions, ensuring image details are discernible and providing a high-quality data foundation for subsequent preprocessing and comparative analysis.

[0032] As one implementation method, the preprocessing of the image data includes: using a Gaussian filtering algorithm to denoise the captured image data to obtain intermediate image data; using histogram equalization technology to enhance the contrast of leaf surface details in the intermediate image data to obtain target influence data; using the Canny edge detection algorithm to extract the contour features and surface texture features of the leaves in the target influence data; and converting the contour features and surface texture features into a standardized digital matrix for subsequent comparison with historical image data.

[0033] Understandably, denoising can effectively remove noise interference such as sand and raindrops, thereby improving data quality. Furthermore, by converting leaf features into a digital matrix, the comparison between current images and historical baseline images is upgraded from "visual comparison" to "quantitative comparison," avoiding the subjectivity of manual visual comparison and thus improving the objectivity and accuracy of comparative analysis.

[0034] Step S202: Perform multi-dimensional comparative analysis between the preprocessed image data and historical image data in the preset dynamic benchmark database, and output the comparison results.

[0035] The preset dynamic benchmark database is established in advance. For example, the preset dynamic benchmark database is established by capturing images of the blades continuously for a certain period of time (such as three days) during the first run and selecting undamaged images as the initial benchmark data; thereafter, the benchmark database is updated once a month, and the latest images without abnormalities in that month are included in the benchmark database.

[0036] It should be understood that the above are merely examples and not limitations.

[0037] As one implementation, step S203 includes: comparing whether the overall outline and edge shape of the blade in the digital matrix are consistent with those in the historical reference image; calculating the matching degree of the surface texture of the blade in the digital matrix and the historical reference image, and determining that the blade in the image data is abnormal when the texture similarity is less than 90%; detecting the color difference between the digital matrix and the blade coating in the historical reference image, and determining that the blade in the image data is abnormal when the color difference is greater than 5.

[0038] Understandably, quantitative identification based on texture and color dimensions can provide clear standards for anomaly detection, avoid subjectivity based on experience, and improve the consistency and accuracy of initial anomaly detection.

[0039] Step S203: Use AI algorithm to identify anomalies in the comparison results and determine whether the blade has any appearance abnormalities and the level of abnormality.

[0040] Optionally, the core of the AI ​​algorithm employs a trained CNN model. The training samples for this CNN model include image data of various anomalies such as blade cracks, corrosion, edge damage, and foreign object attachment. The trained CNN model achieves an accuracy rate of ≥95% in identifying blade appearance anomalies.

[0041] It should be noted that this application does not specifically limit the training process of the CNNC model. For example, the model architecture of this CNN model is ResNet-50, with an input size of 512×512 and an output of 5 types of anomaly probabilities (cracks, corrosion, coating peeling, edge damage, and foreign object attachment); the training data for this CNNC model can be 120,000 labeled images (30,000 images of cracks, 0.5-50cm; 25,000 images of corrosion, 1-1000cm). 2 25,000 sheets of coating peeled off, 5-2000cm 2The dataset consisted of 20,000 images with edge damage and 20,000 images with foreign objects attached, divided into a 7:2:1 ratio (training: validation: testing). The inference parameters of this CNNC model were selected using an NVIDIA Jetson AGX Xavier edge gateway, with an inference speed of approximately 0.5 seconds per image and a test accuracy of 96.5%.

[0042] It should be understood that the above are merely examples and not limitations.

[0043] Understandably, in step S203, the comparison result is input into the CNN model, thereby using the CNN model to determine whether there are any abnormalities in the appearance of the blade, and the level of abnormality is divided into three categories: minor damage, moderate damage and severe damage.

[0044] Understandably, utilizing CNN models for automatic identification solves the problems of low efficiency and high false negative / false positive rates associated with manual identification, enabling rapid and accurate anomaly identification. Furthermore, by classifying anomalies according to their severity, it avoids the waste of resources caused by treating all anomalies the same (e.g., minor damage does not require emergency shutdown), guiding maintenance personnel to prioritize handling moderate / severe damage, improving the efficiency of maintenance resource allocation, and reducing unnecessary downtime losses.

[0045] In one possible embodiment, after step S203, the method further includes: When the CNN model determines that there is an appearance abnormality in the blade, it automatically marks the abnormal area in the image and pushes the abnormal image carrying the abnormal area and the comparison result to the terminal of the maintenance personnel. This allows the maintenance personnel to view the abnormal image and the comparison result through the terminal and manually confirm whether the abnormality is a real abnormality. If the result returned by the terminal is a false abnormality, the abnormal image is input into the CNN model for iterative optimization of the CNN model to further improve the recognition accuracy of the CNN model.

[0046] Understandably, by adding a manual verification step after the initial AI identification, misjudgments caused by AI's limited sample size (such as misidentifying surface stains on blades as corrosion) can be corrected, ensuring the accuracy of anomaly detection and avoiding ineffective maintenance due to misjudgments. Furthermore, by feeding misjudged cases back to the CNN model, the training samples are continuously enriched, allowing the CNN model's recognition capabilities to improve with use. This effectively solves the problem of "AI models being unable to adapt to new scenarios after training," ensuring recognition accuracy over long-term use.

[0047] Step S204: Trigger the corresponding alarm mechanism according to the anomaly level and generate an operation and maintenance work order so as to track the inspection process and repair effect through the operation and maintenance work order.

[0048] The maintenance work order includes the wind turbine number, location, abnormality type, and suggested maintenance plan of the abnormal blade.

[0049] As one implementation method, the alarm mechanism triggered according to the anomaly level includes: when the anomaly level is minor damage, triggering a level one alarm: only an alarm prompt window pops up on the remote monitoring platform, and at the same time automatically records the location, type and severity of the anomaly, and generates a monthly statistical report based on the monthly anomaly data for maintenance personnel to check regularly; when the anomaly level is moderate damage or severe damage, triggering a level two or three alarm: in addition to popping up an alarm prompt window on the remote monitoring platform, the maintenance manager is also notified via SMS or telephone voice, and the GPS location information of the wind turbine to which the abnormal blade belongs and the abnormal image are pushed simultaneously.

[0050] Understandably, tiered alarms prevent the distraction of maintenance personnel caused by "all anomalies triggering the same alarm." Moderate / severe anomalies are notified through multiple channels, including SMS and telephone, and GPS location is pushed to ensure that maintenance personnel are informed immediately and can arrive on-site quickly, reducing the risk of anomalies escalating. In addition, closed-loop management from work order generation and allocation to maintenance tracking and repair verification can effectively ensure that anomaly handling is recorded, tracked, and verified, avoiding situations of "reporting without repair" and "repairing without verification," thereby improving the standardization and effectiveness of maintenance management.

[0051] Third embodiment: See Figure 3 The device shown is an intelligent monitoring device for the appearance of wind turbine blades. The device includes: a camera module 510, a data analysis module 520, an AI analysis module 530, and an alarm module 540. The specific functions of each module are as follows: The camera module 510 is used to trigger the camera device to automatically capture images based on preset conditions, acquire image data within the blade rotation cycle, and preprocess the image data. The camera device is deployed at 20-50 meters from the wind turbine tower base and on the top of the nacelle. Data analysis module 520 is used to perform multi-dimensional comparative analysis of preprocessed image data with historical image data in a preset dynamic benchmark database and output the comparison results. AI analysis module 530 is used to use AI algorithms to identify anomalies in the comparison results and determine whether the blade has any appearance abnormalities and the level of abnormality. The alarm module 540 is used to trigger the corresponding alarm mechanism according to the abnormality level and generate an operation and maintenance work order so as to track the inspection process and repair effect through the operation and maintenance work order.

[0052] It should be noted that the specific implementation of each of the above modules can be found in the corresponding descriptions in the method embodiments, and will not be repeated here.

[0053] Furthermore, this embodiment also provides a computer-readable storage medium storing a computer program, which, when run by a processing device, executes the steps of any of the intelligent monitoring methods for the appearance of wind turbine blades provided in Embodiment 2 above.

[0054] The computer program product of the intelligent monitoring method for the appearance of wind turbine blades provided in this application includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0055] It should be noted that if the above functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0056] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

Claims

1. A method for intelligent monitoring of the appearance of wind turbine blades, characterized in that, The method includes: The camera device is automatically captured based on preset conditions to obtain image data within the blade rotation cycle, and the image data is preprocessed. The camera device is deployed at 20-50 meters from the wind turbine tower base and on the top of the nacelle. The preprocessed image data is compared and analyzed in multiple dimensions with historical image data in a preset dynamic benchmark database, and the comparison results are output. AI algorithms are used to identify anomalies in the comparison results and determine whether the blade has any appearance abnormalities and the level of abnormality. The corresponding alarm mechanism is triggered based on the anomaly level, and a maintenance work order is generated accordingly, so as to track the inspection process and repair effect through the maintenance work order.

2. The method according to claim 1, characterized in that, The camera equipment is an 8K ultra-high-definition ultra-wide-angle camera; wherein, 2-3 of the 8K ultra-high-definition ultra-wide-angle cameras are installed in layers at a height of 20-50 meters from the base of the tower. Each of the 8K ultra-high-definition ultra-wide-angle cameras has a field of view coverage of 120°, and multiple of the 8K ultra-high-definition ultra-wide-angle cameras work together to form a 360° shooting range without blind spots. The 8K ultra-high-definition ultra-wide-angle camera installed on the top of the cabin integrates image stabilization and an infrared fill light module. The 8K ultra-high-definition ultra-wide-angle camera on the top of the cabin is also equipped with a miniature wind power generation unit to power the 8K ultra-high-definition ultra-wide-angle camera and the infrared fill light module.

3. The method according to claim 1, characterized in that, The automatic snapshot triggering of the camera device based on preset conditions includes: When the blade rotation speed stabilizes within the normal operating range of 10-20 revolutions per minute, the camera device is automatically triggered to capture images every hour, continuously taking 10-15 images during each capture to ensure coverage of the complete blade rotation cycle; or, When the wind speed exceeds 15 m / s or the change in light intensity exceeds a threshold, the capture interval of the camera device is shortened to once every 30 minutes, and exposure compensation processing is performed on the captured image data during the capture process to ensure image clarity under special conditions.

4. The method according to claim 1, characterized in that, The preprocessing of the image data includes: The Gaussian filtering algorithm is used to denoise the captured image data to obtain intermediate image data; Histogram equalization is used to enhance the contrast of blade surface details in the intermediate image data to obtain target impact data; The Canny edge detection algorithm was used to extract the contour features and surface texture features of the blades in the target influence data; The contour features and surface texture features are converted into standardized digital matrices for subsequent comparison with historical image data.

5. The method according to claim 4, characterized in that, The process involves performing a multi-dimensional comparative analysis of the preprocessed image data with historical image data in a preset dynamic benchmark database, and outputting the comparison results, including: Compare the overall outline and edge morphology of the blades in the digital matrix with those in the historical reference images to see if they are consistent. Calculate the matching degree between the digital matrix and the leaf surface texture in the historical reference image. When the texture similarity is less than 90%, it is determined that there is an anomaly in the leaf in the image data. The color difference between the digital matrix and the blade coating in the historical reference image is detected. When the color difference is greater than 5, it is determined that there is an anomaly in the blade in the image data.

6. The method according to claim 5, characterized in that, The step of using AI algorithms to identify anomalies in the comparison results and determining whether the blades have any appearance abnormalities and the level of abnormality includes: A trained CNN model is used as the core of the AI ​​algorithm. The training samples of the CNN model include image data of various anomalies such as blade cracks, corrosion, edge damage, and foreign object adhesion. The trained CNN model has an accuracy of ≥95% in recognizing blade appearance anomalies. The CNN model is used to determine whether there are any abnormalities in the appearance of the blades based on the comparison results, and the level of abnormality is divided into three categories: minor damage, moderate damage, and severe damage.

7. The method according to claim 6, characterized in that, The method further includes: When the CNN model determines that there is an appearance abnormality in the blade, it automatically marks the abnormal area in the image and pushes the abnormal image carrying the abnormal area and the comparison result to the terminal of the operation and maintenance personnel, so that the operation and maintenance personnel can view the abnormal image and the comparison result through the terminal and manually confirm whether the abnormality is a real abnormality. If the terminal returns a false anomaly, the abnormal image is input into the CNN model for iterative optimization, thereby further improving the recognition accuracy of the CNN model.

8. The method according to claim 7, characterized in that, The step of triggering the corresponding alarm mechanism based on the anomaly level includes: When the abnormality level is the minor damage, a level 1 alarm is triggered: an alarm prompt window pops up on the remote monitoring platform, and the abnormality location, abnormality type and severity information are automatically recorded. A monthly statistical report is generated based on the monthly abnormality data for maintenance personnel to check regularly. When the abnormality level is moderate damage or severe damage, a level 2 or level 3 alarm is triggered: in addition to popping up an alarm prompt window on the remote monitoring platform, the operation and maintenance manager is also notified via SMS and telephone voice, and the GPS location information of the wind turbine to which the abnormal blade belongs and the abnormal image are pushed simultaneously.

9. The method according to claim 8, characterized in that, The maintenance work order includes the wind turbine number, location, abnormality type, and suggested maintenance plan of the abnormal blade.

10. A smart monitoring device for the appearance of wind turbine blades, characterized in that, The device includes: The camera module is used to trigger the camera device to automatically capture images based on preset conditions, acquire image data within the blade rotation cycle, and preprocess the image data. The camera device is deployed at 20-50 meters from the wind turbine tower base and on the top of the nacelle. The data analysis module is used to perform multi-dimensional comparative analysis between the preprocessed image data and historical image data in the preset dynamic benchmark database, and output the comparison results. The AI ​​analysis module is used to use AI algorithms to identify anomalies in the comparison results and determine whether the blade has any appearance abnormalities and the level of abnormality. The alarm module is used to trigger the corresponding alarm mechanism according to the anomaly level and generate an operation and maintenance work order so as to track the inspection process and repair effect through the operation and maintenance work order.

Images