Automatic inspection method and device for wind power blade

Abstract

The embodiment of the invention provides an automatic inspection method and device for a wind power blade. The method comprises the steps that an unmanned aerial vehicle receives an inspection task which is issued by a central control system and carries geographic position data of a to-be-inspected fan; after the unmanned aerial vehicle arrives at the to-be-inspected fan according to the geographic position data of the to-be-inspected fan, dynamic self-adaptive inspection route planning is executed, and an exclusive inspection route for the to-be-inspected blade is generated; the unmanned aerial vehicle flies along the exclusive inspection route and synchronously performs multi-dimensional data acquisition on the to-be-inspected blade, and the unmanned aerial vehicle transmits the acquired data to the central control system through a communication module of the unmanned aerial vehicle base station; the central control system executes AI identification and analysis on the acquired data to generate a blade defect analysis result of the to-be-inspected blade; and the central control system triggers an abnormal alarm of a corresponding level according to the blade defect analysis result and executes alarm closed-loop management. The inspection efficiency can be improved while automatic inspection is realized.

Description

Technical Field

[0001] This application relates to the field of wind turbine blade inspection technology, and more specifically, to an automatic inspection method and device for wind turbine blades. Background Technology

[0002] Currently, in the field of wind turbine blade inspection, traditional manual climbing inspection and manned helicopter inspection have significant shortcomings: manual inspection is labor-intensive, has high safety risks, relies on personnel experience and is prone to missed or incorrect inspections, and takes 1-2 hours to inspect a single wind turbine, resulting in low efficiency; manned helicopter inspection is severely limited by weather, cannot operate in strong winds, rain, snow, etc., and is also very expensive (thousands of yuan per hour of flight cost), and the propeller airflow may damage the blades. Even the emerging drone inspection is mostly a segmented process of "data collection + manual analysis", only achieving automation of single links such as automatic take-off and landing, resulting in the problem of "fragmented automation", failing to form a complete closed loop from task issuance to problem resolution, and the base stations are mostly centrally deployed at the site, resulting in large coverage blind spots, and the drones have short flight time (20-30 minutes per flight), making it difficult to meet the continuous inspection needs of large-scale wind farms.

[0003] Meanwhile, the data processing and application of existing inspection technologies have obvious limitations: the collected data needs to be manually sorted and analyzed, which is inefficient and lacks historical data comparison, making it impossible to detect the development trend of minor blade damage in a timely manner; the base station design is not optimized in conjunction with the distribution density of wind turbines and has poor matching with the endurance of drones, resulting in a limited inspection range and difficulty in meeting the needs of large-scale and intelligent operation and maintenance of wind farms.

[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 an automatic inspection method and device for wind turbine blades, aiming to improve the above-mentioned problems.

[0006] Firstly, this application provides an automatic inspection method for wind turbine blades, the method comprising: The drone receives an inspection task from the central control system, which carries the geographical location data of the wind turbines to be inspected; wherein, the drone takes off from a drone base station installed at the rear of the wind turbine nacelle, with one drone for every N wind turbines. After arriving at the wind turbine to be inspected based on the geographical location data of the wind turbine to be inspected, the UAV performs dynamic adaptive inspection route planning and generates a dedicated inspection route for the blades to be inspected. The drone flies along the dedicated inspection route and simultaneously collects multi-dimensional data on the blades to be inspected. During and after the data collection process, the drone performs full-process data quality control to filter out the collected data that meets the preset standards from the multi-dimensional data. The drone transmits the collected data to the central control system through the communication module of the drone base station; The central control system performs AI recognition and analysis on the collected data to generate the blade defect analysis results of the blade to be inspected; Based on the blade defect analysis results, the central control system triggers the corresponding level of abnormal alarm and performs alarm closed-loop management.

[0007] In one possible embodiment, the central control system performs AI recognition and analysis on the collected data to generate blade defect analysis results for the blade to be inspected, including: The central control system performs preprocessing operations on the collected data to obtain preprocessed blade images; Perform blade defect identification on the preprocessed blade image to obtain the initial defect identification result corresponding to the blade; Based on the initial defect identification results, the blade defect analysis results of the blade to be inspected are generated.

[0008] In one possible embodiment, generating the blade defect analysis results of the blade to be inspected based on the initial defect identification results includes: Obtain the historical inspection data of the blade to be inspected; The initial defect identification results are compared and analyzed with the historical inspection data to generate the blade defect analysis results of the blade to be inspected. The blade defect analysis results include basic defect information, feature information and trend information.

[0009] In one possible embodiment, the central control system performs preprocessing operations on the acquired data to obtain a preprocessed blade image, including: The central control system uses a Gaussian filtering algorithm to denoise the visible light image and thermal imaging image of the leaf in the acquired data, and obtains the denoised visible light image and the denoised thermal imaging image of the leaf. Image enhancement is achieved by adjusting the brightness and contrast of the denoised visible light image and the denoised thermal image of the leaf through grayscale stretching and histogram equalization, resulting in enhanced visible light image and enhanced thermal image of the leaf. A feature point matching algorithm is used to stitch together multiple enhanced visible light images of the same leaf taken from different angles and positions into a complete leaf unfolding diagram.

[0010] In one possible embodiment, blade defect identification is performed on the preprocessed blade image to obtain an initial defect identification result corresponding to the blade, including: The unfolded image of the blade is input into a pre-trained YOLOv8 model, which outputs the current surface defect identification result corresponding to the blade. The current surface defect identification result includes the defect type, defect location coordinates, and defect size parameters; and, Temperature anomaly region identification is performed on the enhanced thermal imaging image. When the temperature of a certain area of ​​the blade is detected to be greater than or equal to a preset temperature threshold, the area is marked as a temperature anomaly region and associated with the visible light image corresponding to the temperature anomaly region.

[0011] In one possible embodiment, the step of identifying temperature anomaly regions in the enhanced thermal imaging image, wherein when the temperature of a certain area of ​​the blade is detected to be greater than or equal to a preset temperature threshold, the area is marked as a temperature anomaly region, and the visible light image corresponding to the temperature anomaly region is associated, includes: Traverse all pixels of the enhanced leaf thermal imaging image; Extract the temperature value corresponding to each pixel; Filter out target pixels whose temperature values ​​are greater than or equal to a preset temperature threshold; The continuous region formed by the target pixels is marked as a temperature anomaly region; Obtain the coordinate information of the temperature anomaly region in the enhanced blade thermal imaging image; Obtain the spatial position parameters recorded by the UAV when collecting the multi-dimensional data; Based on the spatial location parameters, the coordinate information is mapped onto the unfolded image of the blade to obtain the corresponding area of ​​the temperature anomaly region on the visible light image of the blade.

[0012] In one possible embodiment, the historical inspection data of the blade to be inspected is compared and analyzed with the initial defect identification results to generate blade defect analysis results for the blade to be inspected, including: Acquire historical inspection data of the blade to be inspected for the past 1 month, 3 months and 6 months. The historical inspection data includes historical surface defect identification results and historical temperature anomaly area identification results at the corresponding time nodes. Compare the current surface defect identification results with the historical surface defect identification results, and calculate the defect size change rate; Compare the temperature anomaly area with the historical temperature anomaly area. If the current temperature is higher than or equal to the historical average temperature for the same period by a preset temperature, it is determined that the internal damage of the temperature anomaly area has deteriorated. Based on the current surface defect identification results, the temperature anomaly region, the defect size change rate, and the internal damage deterioration results, a blade defect analysis result including basic defect information, feature information, and trend information is generated.

[0013] In one possible embodiment, after the UAV arrives at the wind turbine to be inspected based on the geographical location data of the turbine, it performs dynamic adaptive inspection route planning to generate a dedicated inspection route for the blades to be inspected, including: After arriving at the wind turbine to be inspected based on the geographical location data of the wind turbine to be inspected, the UAV performs a preset radius surround scan on the blades of the wind turbine to be inspected, and collects the three-dimensional contour data of the blades of the wind turbine to be inspected through lidar to construct a real-time three-dimensional model of the blades. The real-time three-dimensional model of the blade is compared with the pre-stored standard model of the blade to identify whether the blade of the wind turbine to be inspected has bending deformation with a deflection deviation greater than or equal to the first threshold or surface protrusions with a diameter greater than or equal to the second threshold. If the bending deformation or surface protrusion exists, adjust the flight distance and shooting interval of the inspection route, shortening the flight distance from the first preset range in meters to the second preset range, and shortening the shooting interval to half of the original, so as to serve as a dedicated inspection route for the blade to be inspected.

[0014] In one possible embodiment, after the UAV arrives at the wind turbine to be inspected based on the geographical location data of the turbine, it performs dynamic adaptive inspection route planning to generate a dedicated inspection route for the blades to be inspected, including: After the drone arrives at the wind turbine to be inspected based on the geographical location data of the wind turbine to be inspected, it retrieves the historical alarm records of the blades to be inspected. If there is a defect area with a Level 1 or Level 2 alarm in the historical alarm record, a 360° surround shot is taken around the defect area, and the local high temperature tracking mode of the thermal imaging camera is activated as a dedicated inspection route for the blade to be inspected.

[0015] Secondly, this application provides an automatic inspection device for wind turbine blades, comprising: a drone and a central control system, wherein: The drone is used to receive inspection tasks from the central control system, which carry geographical location data of the wind turbines to be inspected; wherein, the drone takes off from a drone base station installed at the rear of the wind turbine nacelle, with one drone for every N wind turbines. The drone is also used to perform dynamic adaptive inspection route planning after arriving at the wind turbine to be inspected based on the geographical location data of the wind turbine to be inspected, and to generate a dedicated inspection route for the blades to be inspected. The drone is also used to fly along the dedicated inspection route and simultaneously collect multi-dimensional data on the blades to be inspected. During and after the data acquisition process, the drone is also used to perform full-process data quality control and filter out the acquired data that meets the preset standards. The drone is also used to transmit the collected data to the central control system through the communication module of the drone base station; The central control system is used to perform AI recognition and analysis on the collected data to generate blade defect analysis results for the blades to be inspected; based on the blade defect analysis results, it triggers corresponding level of abnormal alarms and performs alarm closed-loop management.

[0016] The automatic inspection method and device for wind turbine blades provided in this application, by designing the base station spacing in conjunction with the wind turbine distribution density, allows drones to charge within the base station in a timely manner and continue working after charging, effectively solving the problem of insufficient range and better meeting the needs of large-scale and intelligent operation and maintenance of wind farms. Furthermore, after the drone arrives at the wind turbine to be inspected based on its geographical location data, it performs dynamic adaptive inspection route planning to generate a dedicated inspection route for the blades to be inspected. The drone flies along the dedicated inspection route, simultaneously collecting multi-dimensional data from the blades to be inspected. During and after data collection, the drone performs full-process data quality control, filtering out collected data that meets preset standards from the multi-dimensional data. The drone transmits the collected data to the central control system through the communication module of the drone base station. The central control system performs AI recognition and analysis on the collected data, generating blade defect analysis results for the blades to be inspected. Based on the blade defect analysis results, the central control system triggers corresponding level of abnormal alarms and performs alarm closed-loop management. This forms a complete automated system from task assignment to problem solving, enabling the data collected by drones to be analyzed intelligently without manual processing, thus improving inspection efficiency. 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 A flowchart of an automatic inspection method for wind turbine blades provided in the first embodiment of this application; Figure 2 This is a schematic diagram of the functional modules of an automatic wind turbine blade inspection device provided in the second 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 Reference Figure 1 The flowchart shown represents an automatic inspection method for wind turbine blades, which specifically includes the following steps: Step S201: The UAV receives an inspection task from the central control system, which carries the geographical location data of the wind turbine to be inspected.

[0021] The drones take off from a drone base station installed at the rear of the wind turbine nacelle, with one drone station provided for every N wind turbines.

[0022] Optionally, N can be 3, a number greater than 3, or a number less than 3, such as 2. No specific restrictions are imposed here.

[0023] As one implementation method, the drone can be equipped with miniature wind speed and direction sensors, air pressure sensors, and a miniature microphone array to collect more data.

[0024] It is understandable that by combining the distribution density of wind turbines to design the spacing between base stations, and by ensuring that each base station can charge the drones, the drones' endurance can be effectively improved, thereby increasing the drones' inspection range and efficiency.

[0025] It is understood that in this embodiment, the central control system forwards tasks to the drones through drone base stations. That is, the central control system sends inspection tasks to designated drone base stations, which then distribute the inspection tasks to the drones within them via their built-in short-range wireless communication modules (such as Bluetooth). The specific implementation process is not specifically limited in this application.

[0026] It should be understood that the above are merely examples and not limitations.

[0027] Understandably, the inspection task includes wind turbine information (such as wind turbine number, coordinates of the drone base station, number and number of blades to be inspected, etc.), task execution parameters (such as planned take-off time, inspection time for a single blade, data acquisition frequency, etc.), and safety constraints (such as distance from the blade, coordinates of no-fly zones, minimum remaining power threshold, etc.).

[0028] Of course, in actual use, inspection tasks may include other information besides those mentioned above, or even less information than those mentioned above. No specific limitations are made here.

[0029] In step S202, after the UAV arrives at the wind turbine to be inspected based on the geographical location data of the wind turbine to be inspected, it performs dynamic adaptive inspection route planning and generates a dedicated inspection route for the blades to be inspected.

[0030] As one implementation method, step S202 includes: after the UAV arrives at the wind turbine to be inspected based on the geographical location data of the wind turbine to be inspected, it performs a preset radius surround scan on the blades of the wind turbine to be inspected, and collects the three-dimensional contour data of the blades of the wind turbine to be inspected through LiDAR to construct a real-time three-dimensional model of the blades; the real-time three-dimensional model of the blades is compared with a pre-stored standard model of the blades to identify whether the blades of the wind turbine to be inspected have bending deformation with a deflection deviation greater than or equal to a first threshold or surface protrusions with a diameter greater than or equal to a second threshold; if the bending deformation or the surface protrusions exist, the flight distance and shooting interval of the inspection route are adjusted, shortening the flight distance from a first preset range of meters to a second preset range, and shortening the shooting interval to half of the original, so as to serve as a dedicated inspection route for the blades to be inspected.

[0031] Optionally, the preset radius can be 100 meters.

[0032] Optionally, the first threshold is 5cm.

[0033] Optionally, the second threshold is 10cm.

[0034] Optionally, the first preset range is 5-8 meters, and the second preset range is 3-5 meters.

[0035] It should be understood that the drone is also equipped with a lidar system.

[0036] It is understandable that after the drone reaches the designated location, it first collects blade data to build a real-time 3D model, and then compares it with the pre-stored standard blade model to predict whether the blade has been damaged. This allows for dynamic adjustment of the current inspection route, avoiding the problem of poor inspection results caused by directly using the inspection route issued by the central control system.

[0037] It should be understood that the pre-stored blade standard model is a three-dimensional model built based on historical data, and the specific construction process is not specifically limited in this application.

[0038] As another implementation, step S202 includes: after the UAV arrives at the wind turbine to be inspected based on the geographical location data of the wind turbine to be inspected, it retrieves the historical alarm records of the blade to be inspected; if there is a defect area with a level 1 or level 2 alarm in the historical alarm record, it performs 360° surround shooting around the defect area and starts the local high temperature tracking mode of the thermal imaging camera as a dedicated inspection route for the blade to be inspected.

[0039] In step S203, the UAV flies along the dedicated inspection route and simultaneously collects multi-dimensional data on the blade to be inspected.

[0040] During and after the data acquisition process, the UAV performs full-process data quality control, filtering out the acquired data that meets preset standards from the multi-dimensional data.

[0041] Optionally, multidimensional data includes visible light images and thermal images of the leaves.

[0042] It should be understood that thermal imaging images can be acquired by a pre-configured thermal imaging camera on a drone, and no specific limitations are made here.

[0043] Optionally, the acquisition frequencies of visible light images and thermal imaging images of the leaves can be different. For example, when acquiring visible light images of the leaves, one image can be taken every 2 meters, while when acquiring thermal imaging images, one image can be taken every 5 meters, and the temperature data at the shooting location can be recorded simultaneously.

[0044] Of course, this multi-dimensional data can also include blade surface hardness, roughness, coating adhesion data, blade surface acoustic signals, and airflow velocity and pressure difference data around the blade. For example, miniature contact sensors mounted on the drone can be used to collect blade surface hardness, roughness, and coating adhesion data; miniature microphone arrays mounted on the drone can be used to collect blade surface acoustic signals and convert them into a spectrum; and miniature wind speed and direction sensors and pressure sensors mounted on the drone can be used to collect airflow velocity and pressure difference data around the blade.

[0045] It should be understood that the above are merely examples and not limitations.

[0046] Understandably, end-to-end data quality control refers to pre-acquisition calibration, in-acquisition monitoring, and post-acquisition verification. For example, before acquisition, acquisition parameters are preset according to environmental conditions; the color reproduction of industrial cameras is calibrated using a standard color chart (deviation ≤5%), and the resolution is calibrated using a resolution test chart (able to distinguish 0.1mm lines); the thermal imaging camera is calibrated using a standard blackbody radiation source (temperature measurement deviation ≤0.5℃); and zero-point calibration is performed on each sensor. During acquisition, image sharpness (edge ​​sharpness value <0.3 indicates blur) and exposure are evaluated in real time (overly dark / overly bright pixel ratio ≥30% indicates anomaly). If three consecutive images are abnormal, parameters are adjusted and the image is retaken; if sensor data is out of range, a second acquisition is triggered (with a 5-second interval); if the deviation is ≥10%, the system is shut down for maintenance and a backup drone is dispatched; after acquisition, the actual data is compared with the preset acquisition quantity, and a supplementary acquisition task is generated if any data is missing; thermal imaging and visible light data are correlated and analyzed. If there are no corresponding surface defects in high-temperature areas above 85℃, a logical anomaly is determined, and a re-acquisition is ordered. This ensures that acquired data that meets the criteria of "clear images, valid data, and logical consistency" is ultimately selected.

[0047] Understandably, using drones to automatically filter data during the data collection process can effectively improve data validity and reduce the transmission of invalid data, thereby reducing the waste of transmission resources and improving inspection efficiency.

[0048] In step S204, the UAV transmits the collected data to the central control system through the communication module of the UAV base station.

[0049] Understandably, after collecting data, the drone will return to the drone base station and transmit the collected data that meets the preset standards to the drone base station through a short-range wireless communication module. After receiving the data, the drone base station can further encrypt the data and then transmit it to the central control system through its communication module.

[0050] Of course, it can also forward the data transmitted by the drone directly to the central control system without encryption. No specific limitations are specified here.

[0051] In step S205, the central control system performs AI recognition and analysis on the collected data to generate the blade defect analysis results of the blade to be inspected.

[0052] The blade defect analysis results include basic defect information, characteristic information, and trend information.

[0053] As one implementation method, step S205 includes: the central control system performing preprocessing operations on the collected data to obtain a preprocessed blade image; performing blade defect identification on the preprocessed blade image to obtain an initial defect identification result corresponding to the blade; and generating a blade defect analysis result for the blade to be inspected based on the initial defect identification result.

[0054] Optionally, the central control system performs preprocessing operations on the acquired data to obtain preprocessed blade images, including: the central control system uses a Gaussian filtering algorithm to denoise the visible light image and thermal imaging image of the blade in the acquired data to obtain a denoised visible light image and a denoised thermal imaging image of the blade; the brightness and contrast of the denoised visible light image and the denoised thermal imaging image of the blade are adjusted by grayscale stretching and histogram equalization to achieve image enhancement, resulting in an enhanced visible light image and an enhanced thermal imaging image of the blade; and a feature point matching algorithm is used to stitch together multiple enhanced visible light images of the blade acquired from different angles and positions of the same blade into a complete blade unfolding image.

[0055] Optionally, performing blade defect recognition on the preprocessed blade image to obtain the initial defect recognition result corresponding to the blade includes: inputting the unfolded image of the blade into a pre-trained YOLOv8 model and outputting the current surface defect recognition result corresponding to the blade, wherein the current surface defect recognition result includes defect type, defect location coordinates, and defect size parameters; and performing temperature anomaly region recognition on the enhanced thermal imaging image, wherein when the temperature of a certain region of the blade is detected to be greater than or equal to a preset temperature threshold, the region is marked as a temperature anomaly region and associated with the visible light image corresponding to the temperature anomaly region.

[0056] The YOLOv8 target detection model was trained using over 100,000 labeled images covering various typical defects in wind turbine blades, including cracks, corrosion, coating peeling, lightning damage, blade edge breakage, and surface deposits.

[0057] For example, the training process of this YOLOv8 object detection model is as follows: 1. First, construct a defect image dataset. The dataset contains blade defect images under different climatic conditions (high temperature, low temperature, high humidity, salt spray), different wind turbine types (onshore wind turbines, offshore wind turbines, high-altitude wind turbines), and different operating years (new turbines of 1-3 years, mid-term turbines of 4-10 years, and old turbines of ≥10 years). 2: Next, the images in the dataset are labeled: Defect areas are labeled with rectangular boxes, and attribute information such as defect type, defect level, and defect size is also labeled. 3: Divide the dataset into training, validation, and test sets in a 7:2:1 ratio; 4: The transfer learning method is adopted. Based on the pre-trained YOLOv8 base model, the model is fine-tuned using the training set. The learning rate is set to 0.001, the number of iterations is 100 rounds, and the batch size is 16. 5. Adjust the model hyperparameters in real time using the validation set to ensure that the model's defect identification accuracy on the validation set is ≥97%; 6. Finally, use the test set to evaluate the performance of the trained model: if the defect recognition accuracy is ≥98% and the recall is ≥96% on the test set, the model training is considered complete and the model parameters are saved, which means the training of the YOLOv8 object detection model is complete.

[0058] It should be understood that the above are merely examples and not limitations.

[0059] Optionally, the step of performing temperature anomaly region identification on the enhanced thermal imaging image, whereby when the temperature of a certain area of ​​the blade is detected to be greater than or equal to a preset temperature threshold, the area is marked as a temperature anomaly region, and the visible light image corresponding to the temperature anomaly region is associated, includes: traversing all pixels of the enhanced thermal imaging image of the blade; extracting the temperature value corresponding to each pixel; filtering out target pixels with temperature values ​​greater than or equal to the preset temperature threshold; marking the continuous area formed by the target pixels as a temperature anomaly region; obtaining the coordinate information of the temperature anomaly region in the enhanced thermal imaging image of the blade; obtaining the spatial position parameters recorded by the UAV when collecting the multi-dimensional data; and mapping the coordinate information to the unfolded image of the blade based on the spatial position parameters to obtain the corresponding area of ​​the temperature anomaly region on the visible light image of the blade.

[0060] Optionally, the preset temperature threshold is 80℃.

[0061] It is understandable that, in the process of extracting the temperature value, the grayscale stretching parameter of the thermal imaging image can be used to call the "grayscale value-temperature value" calibration mapping table of the thermal imaging camera.

[0062] It should be noted that this mapping table is a table that establishes a one-to-one correspondence between different grayscale values ​​and corresponding temperature values ​​by calibrating the thermal imaging camera in advance using a standard blackbody radiation source at multiple temperature points in the range of -40℃ to 120℃.

[0063] Of course, in actual use, the mapping table can also be stored locally in advance, that is, the table can be created in advance so that it can be called directly when needed. No specific limitations are made here.

[0064] Optionally, generating the blade defect analysis results for the blade to be inspected based on the initial defect identification results includes: Obtain historical inspection data of the blade to be inspected; compare and analyze the initial defect identification result with the historical inspection data to generate blade defect analysis results of the blade to be inspected, the blade defect analysis results including basic defect information, feature information and trend information.

[0065] Optionally, historical inspection data can be stored locally or on a server, recording the inspection data of the corresponding blades up to the current moment. No specific limitations are specified here.

[0066] Optionally, the historical inspection data of the blade to be inspected is compared and analyzed with the initial defect identification results to generate a blade defect analysis result for the blade to be inspected. This includes: acquiring historical inspection data of the blade to be inspected for the past 1 month, 3 months, and 6 months, wherein the historical inspection data includes historical surface defect identification results and historical temperature anomaly region identification results at the corresponding time points; comparing the current surface defect identification results with the historical surface defect identification results to calculate the defect size change rate; comparing the temperature anomaly region with the historical temperature anomaly region, and if the current temperature is higher than or equal to the historical average temperature for the same period by a preset temperature, determining that the internal damage of the temperature anomaly region has deteriorated; and generating a blade defect analysis result including defect basic information, feature information, and trend information based on the current surface defect identification result, the temperature anomaly region, the defect size change rate, and the internal damage deterioration result.

[0067] Optionally, the current surface defect identification result is compared with the historical surface defect identification result to calculate the defect size change rate, including: using a pixel-level comparison algorithm to compare the defect position coordinates and defect size parameters in the current surface defect identification result with the defect parameters in the historical data one by one, thereby calculating the defect size change rate and position offset; if a defect has a size extension ratio ≥150% or a position offset ≥5cm within 3 months, the defect is determined to be a "deteriorating defect".

[0068] Optionally, the preset temperature can be 10°C.

[0069] Optionally, the basic defect information includes, but is not limited to, a unique identifier, defect type, and defect location coordinates; for example, the unique identifier includes the fan number - blade number - defect sequence number (e.g., #15 fan - blade number 2 - 001).

[0070] It should be noted that the defect number can be generated according to a preset method, or according to the chronological order or the order in which the defects appeared. No specific limitation is made here.

[0071] Optionally, the feature information includes, but is not limited to, current quantization parameters, physical property data, and environmental correlation data; for example, the current quantization parameters may be defect size parameters (such as "crack length 6.8cm, width 0.3cm") and abnormal temperatures in the temperature anomaly area (such as "temperature anomaly area temperature 82℃, surrounding area 36℃").

[0072] Optionally, trend information may include, but is not limited to, development forecasts, risk warnings, and inspection recommendations.

[0073] It is understandable that trend information is generated based on predictions made from current and historical situations, and no specific limitations are imposed here.

[0074] For example, regarding the prediction of defect size, if the current defect size is 6.8cm, then the defect size will be 9.5cm after 3 months, and 12.2cm after 6 months.

[0075] It should be noted that predictions can be made by setting up a linear growth model, but other methods can also be used.

[0076] It should be understood that the above are merely examples and not limitations.

[0077] In step S206, the central control system triggers an alarm of the corresponding level and performs alarm closed-loop management based on the blade defect analysis results.

[0078] Optionally, the level can be divided into multiple levels, such as level one, level two, or level three.

[0079] For example, a Level 1 anomaly alarm (i.e., an emergency alarm) is triggered when the defect analysis results show that the blade has a through crack with a length ≥10cm, severe corrosion with an area ≥0.5㎡, an internal hot spot with a temperature ≥80℃ and an area ≥0.2㎡, or when the defect size increases by ≥5cm within 1 month. Level 2 anomaly alarm (i.e., critical alarm): triggered when defect analysis results show that the blade has a non-penetrating crack with a length of 3-10cm, localized corrosion with an area of ​​0.1-0.5㎡, localized hot spots with a temperature of 60-80℃, or when the defect size increases by 3-5cm within 3 months; Level 3 anomaly alarm (i.e., alert alarm): triggered when the defect analysis results show that the blade has a fine crack with a length of <3cm, a slight coating peeling with an area of ​​<0.1㎡, a slight temperature anomaly with a temperature of <60℃, or the defect has not changed significantly in size within 6 months.

[0080] Understandably, different alarm methods are triggered for different levels of anomalies, for example: When a Level 1 alarm is triggered: the central control system immediately issues an audible and visual alarm with flashing red lights and a continuous buzzer; alarm information is sent to the operation and maintenance manager and regional supervisor via SMS and a dedicated operation and maintenance APP, including the wind turbine number, defect location, defect type, defect size, and suggested processing time limit (within 1 hour); the wind turbine location is highlighted in red on the central control system map interface, and a pop-up window displays the defect image and analysis report; When a Level 2 alarm is triggered: the central control system emits an audible and visual alarm with flashing yellow lights and intermittent buzzer sounds (once every 3 seconds); alarm information is sent to maintenance personnel via a dedicated maintenance APP, including the wind turbine number, defect details, and suggested processing time (within 72 hours); the wind turbine location is marked in yellow on the central control system map interface, and clicking the mark allows viewing defect analysis data; When a Level 3 alarm is triggered: the central control system only records the defect information in the "Defects to be Monitored" list and does not display the audible and visual alarm on the central control screen; during the next inspection of the wind turbine, the defect area will be automatically set as a key inspection area, and the number of images collected and the sensor sampling frequency will be increased.

[0081] It should be understood that the above are merely examples and not limitations.

[0082] Understandably, by triggering different alarm modes for different levels of anomalies, differentiated information push can be achieved to ensure accurate allocation of operation and maintenance resources (such as level 1 alarm dispatch supervisor, level 2 alarm dispatch on-site personnel), avoid "resource waste or response delay", effectively improve operation and maintenance efficiency, ensure that wind turbines can operate efficiently, and reduce the adverse effects of wind turbines operating "with defects".

[0083] Second embodiment: See Figure 2 The device shown is an automatic inspection device for wind turbine blades, which includes: a central control system 100, a drone 200, and a drone base station 300.

[0084] The drone base station 200 is deployed at the rear of the wind turbine nacelle. For example, one drone base station 200 is provided for every N wind turbines.

[0085] Optionally, N can be 3, 2, or 4; no specific limitation is made here.

[0086] It should be noted that the data exchange between the central control system 100 and the UAV 200 is transmitted through the communication module installed in the UAV base station 300.

[0087] The drone 200 is used to perform the following steps: After receiving the inspection task from the central control system, which carries the geographical location data of the wind turbines to be inspected, the drone takes off from the drone base station installed at the rear of the wind turbine nacelle, with one drone for every N wind turbines. After arriving at the wind turbine to be inspected based on the geographical location data of the wind turbine to be inspected, dynamic adaptive inspection route planning is performed to generate a dedicated inspection route for the blades to be inspected. The system flies along the dedicated inspection route and simultaneously collects multi-dimensional data on the blades to be inspected. During and after the data collection process, full-process data quality control is performed to filter out the collected data that meets the preset standards from the multi-dimensional data. The collected data is transmitted to the central control system through the communication module of the UAV base station.

[0088] The central control system 100 is used to perform the following steps: AI recognition and analysis are performed on the collected data to generate the blade defect analysis results of the blade to be inspected; Based on the blade defect analysis results, trigger the corresponding level of abnormal alarm and execute alarm closed-loop management.

[0089] It should be noted that the specific process of the steps performed by the drone 200 and the central control system 100 can be found in the description of the method embodiment, and will not be repeated here.

[0090] 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 automatic wind turbine blade inspection methods provided in Embodiment 1 above.

[0091] The computer program product of the automatic inspection method and device for 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.

[0092] 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.

[0093] 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 automatically inspecting wind turbine blades, characterized in that, The method comprises: The unmanned aerial vehicle receives an inspection task carrying geographic position data of a wind turbine to be inspected issued by a central control system; wherein the unmanned aerial vehicle takes off from an unmanned aerial vehicle base station installed at the tail of a wind turbine cabin and equipped with one unmanned aerial vehicle every N wind turbines; After the unmanned aerial vehicle arrives at the wind turbine to be inspected according to the geographic position data of the wind turbine to be inspected, the unmanned aerial vehicle performs dynamic adaptive inspection route planning and generates a dedicated inspection route for the blade to be inspected; The unmanned aerial vehicle flies along the dedicated inspection route and synchronously collects multi-dimensional data of the blade to be inspected; wherein during and after the data collection process, the unmanned aerial vehicle performs full-process data quality control and filters out collected data meeting preset standards from the multi-dimensional data; The unmanned aerial vehicle transmits the collected data to the central control system through a communication module of the unmanned aerial vehicle base station; The central control system performs AI recognition and analysis on the collected data and generates blade defect analysis results of the blade to be inspected; The central control system triggers an abnormal alarm of a corresponding level and performs alarm closed-loop management according to the blade defect analysis results.

2. The method of claim 1, wherein, The central control system performs AI recognition and analysis on the collected data and generates blade defect analysis results of the blade to be inspected, comprising: The central control system performs a pretreatment operation on the collected data to obtain pretreated blade images; The central control system performs blade defect recognition on the pretreated blade images to obtain initial defect recognition results corresponding to the blade; The central control system generates blade defect analysis results of the blade to be inspected based on the initial defect recognition results.

3. The method of claim 2, wherein, The central control system generates blade defect analysis results of the blade to be inspected based on the initial defect recognition results, comprising: The central control system acquires historical inspection data of the blade to be inspected; The central control system compares and analyzes the initial defect recognition results with the historical inspection data to generate blade defect analysis results of the blade to be inspected, wherein the blade defect analysis results comprise defect basic information, feature information and trend information.

4. The method of claim 3, wherein, The central control system performs a pretreatment operation on the collected data to obtain pretreated blade images, comprising: The central control system uses a Gaussian filter algorithm to perform noise reduction processing on blade visible light images and thermal imaging images in the collected data to obtain noise-reduced blade visible light images and noise-reduced thermal imaging images; The central control system adjusts the brightness and contrast of the noise-reduced blade visible light images and the noise-reduced thermal imaging images through gray scale stretching and histogram equalization to realize image enhancement, thereby obtaining enhanced blade visible light images and enhanced thermal imaging images; The central control system uses a feature point matching algorithm to stitch multiple enhanced blade visible light images collected at different angles and different positions of the same blade into a complete blade development map.

5. The method of claim 4, wherein, The central control system performs blade defect recognition on the pretreated blade images to obtain initial defect recognition results corresponding to the blade, comprising: The central control system inputs the blade development map into a pre-trained YOLOv8 model to output current surface defect recognition results corresponding to the blade, wherein the current surface defect recognition results comprise defect types, defect position coordinates and defect size parameters; and, The enhanced thermal imaging image is subjected to temperature abnormal area identification, and when it is detected that the temperature of a certain area of the blade is greater than or equal to a preset temperature threshold, the area is marked as a temperature abnormal area, and the visible light image corresponding to the temperature abnormal area is associated.

6. The method of claim 5, wherein, The enhanced thermal imaging image is subjected to temperature abnormal area identification, and when it is detected that the temperature of a certain area of the blade is greater than or equal to a preset temperature threshold, the area is marked as a temperature abnormal area, and the visible light image corresponding to the temperature abnormal area is associated. All pixel points of the enhanced blade thermal imaging image are traversed; The temperature value corresponding to each pixel point is extracted; Target pixel points with a temperature value greater than or equal to a preset temperature threshold are screened out; A continuous area formed by the target pixel points is marked as a temperature abnormal area; Coordinate information of the temperature abnormal area in the enhanced blade thermal imaging image is obtained; Spatial position parameters recorded by the unmanned aerial vehicle when collecting the multi-dimensional data are obtained; Based on the spatial position parameters, the coordinate information is mapped to the blade development diagram to obtain a corresponding area of the temperature abnormal area on the blade visible light image.

7. The method according to claim 5 or 6, characterized in that, The historical inspection data of the blade to be inspected and the initial defect identification result are compared and analyzed to generate a blade defect analysis result of the blade to be inspected, including: Obtain the historical inspection data of the blade to be inspected within 1 month, 3 months and 6 months, and the historical inspection data includes the historical surface defect identification result and the historical temperature abnormal area identification result at the corresponding time node; Compare the current surface defect identification result with the historical surface defect identification result to calculate the defect size change rate; Compare the temperature abnormal area with the historical temperature abnormal area, and if the current temperature is greater than or equal to the preset temperature than the historical average temperature of the same period, it is determined that the internal damage of the temperature abnormal area is deteriorating; According to the current surface defect identification result, the temperature abnormal area, the defect size change rate and the internal damage deterioration result, a blade defect analysis result including defect basic information, feature information and trend information is generated.

8. The method of claim 1, wherein, After the unmanned aerial vehicle arrives at the wind turbine to be inspected according to the geographic location data of the wind turbine to be inspected, a dynamic adaptive inspection flight path is planned, and a dedicated inspection flight path for the blade to be inspected is generated, including: After the unmanned aerial vehicle arrives at the wind turbine to be inspected according to the geographic location data of the wind turbine to be inspected, a preset radius around scanning is performed on the blades of the wind turbine to be inspected, three-dimensional profile data of the blades of the wind turbine to be inspected is collected by a laser radar, and a real-time three-dimensional model of the blades is constructed; The real-time three-dimensional model of the blades is compared with a pre-stored standard model of the blades to identify whether the blades of the wind turbine to be inspected have a bending deformation with a deflection deviation greater than or equal to a first threshold value or a surface protrusion with a diameter greater than or equal to a second threshold value; If the bending deformation or the surface protrusion exists, the flight distance and the shooting interval of the inspection flight path are adjusted, the flight distance is shortened from a first preset range to a second preset range, and the shooting interval is shortened to half of the original, so as to serve as a dedicated inspection flight path for the blade to be inspected.

9. The method of claim 1, wherein, The unmanned aerial vehicle reaches the wind turbine to be inspected according to the geographic position data of the wind turbine to be inspected, performs dynamic adaptive inspection route planning, and generates a dedicated inspection route for the blade to be inspected, including: The unmanned aerial vehicle reaches the wind turbine to be inspected according to the geographic position data of the wind turbine to be inspected, and calls the historical alarm record of the blade to be inspected; If there is a defect area of 1st or 2nd level alarm in the historical alarm record, 360° surround shooting is performed around the defect area, and a local high-temperature tracking mode of the thermal imaging camera is started to serve as the dedicated inspection route of the blade to be inspected.

10. An automatic inspection device for wind turbine blades, characterized in that The device comprises an unmanned aerial vehicle and a central control system, wherein: The unmanned aerial vehicle is configured to receive an inspection task carrying geographic position data of a wind turbine to be inspected issued by the central control system, and take off from a UAV base station installed at the tail of a wind turbine cabin and equipped with one UAV every N wind turbines. The unmanned aerial vehicle is further configured to reach the wind turbine to be inspected according to the geographic position data of the wind turbine to be inspected, perform dynamic adaptive inspection route planning, and generate a dedicated inspection route for the blade to be inspected. The unmanned aerial vehicle is further configured to fly along the dedicated inspection route and synchronously collect multi-dimensional data of the blade to be inspected. During and after the data collection, the unmanned aerial vehicle is further configured to perform full-process data quality control and screen collected data meeting preset standards. The unmanned aerial vehicle is further configured to transmit the collected data to the central control system through a communication module of the UAV base station. The central control system is configured to perform AI identification and analysis on the collected data, generate a blade defect analysis result of the blade to be inspected, trigger an abnormal alarm of a corresponding level according to the blade defect analysis result, and perform alarm closed-loop management.

Images