Unmanned aerial vehicle carries intelligent sensor fan inspection route planning method and system

Abstract

The application discloses a fan inspection route planning method and system of an unmanned aerial vehicle carrying an intelligent sensor, relates to the technical field of inspection planning, and comprises the following steps: performing grid processing on fan blades, and calculating the prior defect probability of each blade grid; performing detection balance decision by using the prior defect probability and shutdown cost; if the detection balance decision is a start-up detection decision, acquiring the blade time sequence motion track of the fan blade, combining the prior defect probability to configure a time sequence target detection position; acquiring the structure data of the fan blade, performing wind field disturbance perception of blade rotation, and establishing a disturbance perception factor; performing follow-up mode and static suspension mode decision by using the disturbance perception factor and the time sequence target detection position, configuring an inspection route, and performing inspection management. The application solves the technical problems of non-intelligent route planning, low detection efficiency and great influence of wind field disturbance in the fan blade inspection process in the prior art, and achieves the technical effects of improving inspection accuracy and stability.

Description

Technical Field

[0001] This invention relates to the field of inspection planning technology, specifically to a method and system for planning wind turbine inspection routes using unmanned aerial vehicles equipped with intelligent sensors. Background Technology

[0002] Wind turbine blade inspections typically rely on manual labor or drones operating along pre-defined routes. These fixed routes lack dynamic sensing capabilities regarding blade structural characteristics and operational status, making it impossible to adaptively adjust to real-time blade movement and wind field changes during inspection. Wind disturbances also negatively impact drone attitude stability and inspection accuracy, leading to incomplete or significantly inaccurate inspection data. Furthermore, repetitive route planning results in low inspection efficiency, hindering the achievement of high-precision, high-efficiency, and intelligent inspection requirements. Summary of the Invention

[0003] This application provides a method and system for wind turbine inspection route planning using UAVs equipped with intelligent sensors, which is used to address the technical problems of unintelligent route planning, low detection efficiency, and significant impact from wind field disturbances during wind turbine blade inspection in the prior art.

[0004] In view of the above problems, this application provides a method and system for wind turbine inspection route planning using UAVs equipped with intelligent sensors.

[0005] The first aspect of this application provides a method for planning wind turbine inspection routes using a drone equipped with intelligent sensors, the method comprising:

[0006] The process involves meshing the wind turbine blades and calculating the prior defect probability for each blade mesh. This prior defect probability is constructed using historical data, thermal image data, and long-range visual data. A detection balance decision is made using the prior defect probability and downtime cost. If the detection balance decision is to initiate a detection process, the temporal motion trajectory of the wind turbine blades is acquired, and a temporal target detection position is configured based on the temporal motion trajectory and the prior defect probability. Structural data of the wind turbine blades is acquired, and wind field disturbance perception based on the structural data and the temporal motion trajectory is performed to establish a disturbance perception factor. A follow-up mode and static suspension mode decision are executed using the disturbance perception factor and the temporal target detection position. The decision results are used to configure inspection routes and perform wind turbine blade inspection management.

[0007] A second aspect of this application provides a wind turbine inspection route planning system for unmanned aerial vehicles equipped with intelligent sensors, the system comprising:

[0008] The system includes a gridding module for performing gridding of the wind turbine blades and calculating the prior defect probability for each blade grid. This prior defect probability is constructed using historical data, thermal image data, and long-range visual data. A decision-making module is used to make a detection balance decision based on the prior defect probability and downtime costs. A detection position configuration module is used to acquire the temporal motion trajectory of the wind turbine blades if the detection balance decision is an on-time detection decision, and configure the temporal target detection position based on the temporal motion trajectory and the prior defect probability. A disturbance perception module is used to acquire the structural data of the wind turbine blades, perceive wind field disturbances caused by blade rotation based on the structural data and the temporal motion trajectory, and establish a disturbance perception factor. An inspection management module is used to execute follow-up mode and static suspension mode decisions using the disturbance perception factor and the temporal target detection position, configure inspection routes based on the decision results, and perform inspection management of the wind turbine blades.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] This application performs gridded processing of wind turbine blades, calculates the prior defect probability of each blade grid, and constructs the prior defect probability using historical data, thermal image data, and long-range visual data. It then uses the prior defect probability and downtime cost to make a detection balance decision. If the detection balance decision is to enable detection, it acquires the temporal motion trajectory of the wind turbine blades and configures the temporal target detection position based on the temporal motion trajectory and the prior defect probability. It acquires the structural data of the wind turbine blades and uses the structural data and the temporal motion trajectory to perceive wind field disturbances caused by blade rotation, establishing a disturbance perception factor. It then uses the disturbance perception factor and the temporal target detection position to execute follow-up mode and static suspension mode decisions, configures the inspection route based on the decision results, and performs wind turbine blade inspection management. This invention solves the technical problems of unintelligent route planning, low detection efficiency, and significant susceptibility to wind field disturbances in the prior art during wind turbine blade inspection. By introducing prior defect probability and disturbance perception factor for intelligent route planning and dynamic decision-making, it achieves the technical effect of improving inspection accuracy and stability. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A schematic diagram of the wind turbine inspection route planning method using an unmanned aerial vehicle equipped with intelligent sensors, provided in an embodiment of this application.

[0013] Figure 2 A schematic diagram of the structure of a wind turbine inspection route planning system equipped with intelligent sensors for unmanned aerial vehicles provided in this application embodiment.

[0014] Figure labeling: 11. Grid processing module, 12. Decision module, 13. Detection location configuration module, 14. Disturbance sensing module, 15. Inspection management module. Detailed Implementation

[0015] This application provides a method and system for wind turbine inspection route planning using UAVs equipped with intelligent sensors. It addresses the technical problems of unintelligent route planning, low detection efficiency, and significant susceptibility to wind field disturbances in existing wind turbine blade inspection processes. By introducing prior defect probability and disturbance perception factors for intelligent route planning and dynamic decision-making, it achieves the technical effect of improving inspection accuracy and stability.

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0017] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0018] Example 1, as Figure 1 As shown, this application provides a method for planning wind turbine inspection routes using an unmanned aerial vehicle (UAV) equipped with intelligent sensors. The method includes:

[0019] Step S100: Perform meshing processing on the wind turbine blades and calculate the prior defect probability of each blade mesh. The prior defect probability is constructed using historical data, thermal image data, and long-distance visual data.

[0020] In this embodiment, when performing meshing processing on the wind turbine blades, the complete shape data of the wind turbine blades is first obtained using a three-dimensional laser scanning modeling method, with a sampling accuracy set to 0.5 mm to ensure high-fidelity reconstruction of surface details. By performing surface fitting and equidistant division processing on the obtained three-dimensional point cloud data, the blade surface is divided into several regular mesh units with a side length of approximately 10 cm, and each mesh unit has a unique spatial coordinate number.

[0021] When calculating the prior defect probability of the blade mesh, the defect distribution characteristics in historical operating data are first extracted based on statistical analysis methods. By analyzing maintenance and inspection records from the past three years, the frequency of failures in each blade region is calculated. For example, the defect probability of the blade leading edge region is 0.10. Subsequently, infrared thermal imaging analysis is used to process the thermal shadow data of the wind turbine under operating conditions, detecting abnormal areas with a temperature gradient greater than 2.5℃ / cm, and setting their corresponding thermal anomaly probability to 0.12; if the temperature gradient is less than or equal to 2.5℃ / cm, the thermal anomaly probability of that area is set to 0. Finally, image recognition methods are used to analyze long-distance visual data, identifying feature areas with crack lengths greater than 5 mm and grayscale contrast higher than 0.4, and setting their visual defect probability to 0.15; if the crack length is less than or equal to 5 mm or the grayscale contrast is less than or equal to 0.4, the visual defect probability is set to 0.

[0022] After obtaining three types of basic data, they are uniformly normalized, and the prior defect probability of each grid is calculated using a weighted average method. Historical data has a weight of 0.50, thermal anomaly data has a weight of 0.30, and long-range visual data has a weight of 0.20. The calculation formula is: Prior defect probability = 0.50 × Historical defect probability + 0.30 × Thermal anomaly probability + 0.20 × Visual defect probability. Using this formula, the prior defect probability of each blade grid is obtained.

[0023] Step S200: Make a detection balance decision using the prior defect probability and downtime cost.

[0024] Furthermore, the method provided in the application embodiments, which utilizes the prior defect probability and downtime cost to make a detection balance decision, further includes:

[0025] A risk energy distribution model for the blade region is constructed based on the prior defect probability. The risk energy characterizes the potential failure accumulation trend of each region of the blade under continuous wind turbine operation. Real-time operating parameters and downtime cost data of the wind turbine are obtained, and the energy decay function of the blade's current operating state is calculated. The risk energy distribution model and the energy decay function are subjected to dynamic game analysis to establish a detection balance spectrum. The detection balance spectrum characterizes the detection benefit boundary under different blade states. A detection balance decision is made based on the detection balance spectrum.

[0026] In this embodiment, when making a detection balance decision using prior defect probability and downtime cost, a risk energy distribution model for the blade region is first constructed based on the prior defect probability. This model is then calculated using the energy accumulation method. Taking the gridded blade as the object, the prior defect probability of each grid is multiplied by the equivalent load energy density of that grid under the current operating conditions to obtain the risk energy value. Spatial interpolation is then performed on the entire blade grid to form a continuous risk energy distribution. The equivalent load energy density is obtained by dividing the mechanical power converted from real-time rotational speed and main shaft torque by the effective stress area, ensuring that the risk energy distribution model can quantitatively characterize the failure accumulation trend under continuous wind turbine operation. This process completes the construction of the risk energy distribution model.

[0027] Next, real-time operating parameters and downtime cost data of the wind turbine are acquired. Specifically, the SCADA system reads real-time operating parameters such as rotor speed, main shaft torque, power generation, wind speed, blade pitch angle, and nacelle attitude angle. Based on the power curve and electricity price information, the power generation control system calculates the economic loss per unit downtime, obtaining downtime cost data. For example, under the conditions of a rated power of 2MW and an electricity price of 0.45 yuan per kilowatt-hour, the economic loss for a 1-hour downtime is approximately 900 yuan. Then, when calculating the energy decay function of the blades in their current operating state, an exponential decay fitting method is used to fit the risk energy time series. By using the least squares method to solve the risk energy data points of the most recent 24 hours, a functional form is established. ,in, Let k be the initial risk energy, and k be the energy decay coefficient. The decay coefficient describes the natural decay of risk energy over time or its reduction through routine online maintenance under continuous wind turbine operation. When k is small (e.g., <0.02), it indicates that the risk energy is reduced slowly, blade fatigue is accumulating, and shutdown for inspection should be prioritized; when k is large (e.g., >0.05), it indicates that the energy decay is significant, and the turbine can maintain its operating status.

[0028] The risk energy distribution model and the energy decay function are then subjected to dynamic game analysis. Cost-benefit analysis is used for evaluation during this process. The equilibrium point of the detection is determined by comparing the risk energy reduction benefits with the economic losses from downtime under different detection timings. The risk energy reduction benefit is obtained by comparing the risk energy difference in high-risk areas before and after detection. For example, if the risk energy in the high-risk area of ​​the blade decreases by 20% after detection, it indicates that the detection has achieved a corresponding reduction benefit. When the risk energy reduction benefit is higher than the downtime loss, the detection is considered economically reasonable. By calculating the equilibrium point under different operating conditions over time, a curve showing the change in the detection benefit boundary is plotted, thus establishing a detection equilibrium map to characterize the balance between detection benefits and costs under different blade conditions.

[0029] Finally, a detection balance decision is made based on the detection balance map. The detection trigger interval is identified based on the temporal changes of the detection benefit boundary. When the detection trigger interval covers the target high-risk area, a shutdown detection decision is output, and a static inspection by the UAV is performed to obtain high-precision detection results. When the detection trigger interval does not cover the target high-risk area, a startup detection decision is output, maintaining the wind turbine's normal operation and conducting online monitoring. The target high-risk area refers to the region in the risk energy distribution model where the risk energy value exceeds a set threshold. These regions are typically located at the blade leading edge, blade tip, and areas of concentrated stress.

[0030] Step S300: If the detection balance decision is the start-up detection decision, then obtain the blade temporal motion trajectory of the wind turbine blade, and configure the temporal target detection position according to the blade temporal motion trajectory and the prior defect probability.

[0031] In this embodiment, when the detection balance decision is the start-up detection decision, the timing motion trajectory of the wind turbine blades is first acquired. During this process, the rotation angle, speed, and blade spatial position information of the wind turbine in operation are synchronously collected through a speed sensor and an angle encoder, and the collected data are processed for time synchronization and spatial coordinate calculation to obtain the timing motion trajectory of the wind turbine blades.

[0032] Next, the temporal target detection positions are configured based on the blade's temporal motion trajectory and prior defect probabilities. In this process, the prior defect probabilities of all grids on the blade surface are spatially mapped. The prior defect probability corresponding to each grid is matched with the spatial coordinates in the blade's temporal motion trajectory to determine the temporal position relationship of each grid within the rotation cycle. By calculating the blade's angular velocity, the relative position of the UAV, and the observation range of the detection equipment, the detectable time window and spatial coordinates of each blade grid during rotation are determined sequentially, thus configuring corresponding temporal detection points for all blade regions. Subsequently, the detection time, detection coordinates, and corresponding prior defect probability parameters of each grid are integrated to form the temporal target detection positions.

[0033] Step S400: Obtain the structural data of the wind turbine blades, and based on the structural data and the time-series motion trajectory of the blades, perform wind field disturbance perception of blade rotation, and establish a disturbance perception factor.

[0034] In this embodiment, the structural data of the wind turbine blades, including geometric and material parameters, is first obtained from pre-stored wind turbine blade design data. The geometric parameters include length, chord length, and torsion angle distribution; the material parameters include density, elastic modulus, and damping coefficient.

[0035] Next, wind field disturbance perception during blade rotation is performed based on structural data and the blade's time-series motion trajectory. In this process, fluid dynamics analysis methods are used to model the interaction between the blade and the airflow during rotation. The blade's geometry and rotational trajectory are input into a computational fluid dynamics (CFD) model to solve for the relative wind speed, angle of attack, and local aerodynamic changes at different rotation angles. Numerical simulations are used to obtain the air pressure and velocity distribution fields on the blade surface, which are then corrected using real-time wind speed and pressure data collected by anemometers and pressure sensors. This yields the wind field disturbance distribution characteristics during the blade's rotation cycle, including dynamic changes in airflow velocity fluctuations, pressure pulsations, and vortex formation intensity.

[0036] After completing the wind field disturbance sensing, a disturbance sensing factor is established. This process involves first normalizing the aerodynamic fluctuation data at different angular positions within the blade rotation cycle, extracting the average disturbance intensity and disturbance energy density. Then, using the ratio of disturbance energy at each moment to the periodic average energy, the disturbance change ratio within the blade rotation cycle is calculated. Finally, the disturbance ratios at each position are weighted and averaged to obtain the disturbance sensing factor.

[0037] Step S500: Utilize the disturbance sensing factor and the time-series target detection position to perform follow mode and static suspension mode decision-making, configure the inspection route using the decision results, and perform inspection management of the wind turbine blades.

[0038] In this embodiment, when using the disturbance sensing factor and the time-series target detection location to decide between follow mode and hover mode, the degree of wind field disturbance is first judged based on the value of the disturbance sensing factor. Using the disturbance sensing factor as the core criterion, when the disturbance sensing factor is less than 0.5, it indicates that the airflow in the blade rotation area is stable and the disturbance amplitude is low. In this case, the UAV can maintain a relatively stable flight attitude during blade rotation and execute the follow mode decision. When the disturbance sensing factor is greater than or equal to 0.5, it indicates that the wind field around the blade fluctuates significantly and the disturbance energy is high. If the UAV continues to perform follow-up detection, it is easily affected by the airflow, and the hover mode decision is executed to complete the detection task in a hovering manner.

[0039] When executing the follow-mode decision, a spatial trajectory sequence with time stamps is generated based on the time-series target detection position. This spatial trajectory sequence is used to construct a safe following curve for the UAV, and the curve is positionally corrected based on real-time changes in disturbance sensing factors to establish a dynamically corrected flight path. The dynamically corrected flight path is optimized through a path planning optimization algorithm to obtain a calibrated following path. Finally, an inspection flight path is configured based on the calibrated following path, enabling the UAV to achieve continuous and synchronous dynamic detection during blade rotation.

[0040] When executing static hovering mode decisions, the dynamic passage angle range of the blade tip and middle is extracted based on the blade's temporal motion trajectory to establish a temporal channel model of the blade, mapping the blade's rotational motion into a time-space channel to predict the blade's spatial position at each moment. The safe static hovering range of the UAV is determined by combining disturbance sensing factors, constructing a set of UAV static hovering points, and using prior defect probabilities to perform confidence-weighted selection from this set to determine the target static hovering detection points. An inspection route is generated based on the target static hovering detection points to guide the UAV in completing intermittent inspections of the blade surface within a stable airflow range.

[0041] Furthermore, the method provided in the application embodiment, which uses the decision results to configure the inspection route and perform inspection management of the wind turbine blades, also includes:

[0042] Configure a spatial trajectory sequence with temporal identifiers based on the time-series target detection location; if the decision result is a follow-up mode decision, then use the spatial trajectory sequence as a reference to construct a safe following curve for the UAV; perform position correction of the safe following curve based on the disturbance perception factor to establish a dynamically corrected flight path; perform path planning optimization of the dynamically corrected flight path, and use the path planning optimization result to establish a calibrated following path; configure an inspection flight path using the calibrated following path to perform inspection management of the wind turbine blades.

[0043] In this embodiment, when configuring a spatial trajectory sequence with temporal identifiers based on the temporal target detection positions, the previously determined temporal target detection positions are first arranged in chronological order, and the time node, spatial coordinates, and attitude angle parameters of each detection position are paired to form a trajectory point set with time synchronization characteristics. A trajectory interpolation algorithm is then used to perform curve fitting on the trajectory point set to generate a continuous spatial trajectory sequence, and a time identifier is added to each trajectory point in the sequence, thereby forming a spatial motion trajectory with temporal characteristics.

[0044] When the decision outcome is a follower-mode decision, a safe following curve for the UAV is constructed using the spatial trajectory sequence as a reference. During the construction process, the safe distance between the UAV and the blade surface is calculated by analyzing the speed and acceleration of the blade rotation and the UAV's maximum attitude adjustment rate. Using the Bezier curve fitting method, a safe following curve for the UAV is generated while ensuring the safety of the observation angle and distance, enabling it to maintain its observation attitude synchronously with the blade rotation without entering the aerodynamic interference zone.

[0045] Subsequently, the position of the safe following curve is corrected based on the disturbance sensing factor to establish a dynamically corrected flight path. During this process, changes in the disturbance sensing factor are monitored in real time. When wind disturbance increases, the relative distance parameter between the UAV and the blades is dynamically adjusted, and the flight path point position is corrected through an adaptive correction algorithm to expand the safety boundary. When the disturbance sensing factor decreases, the relative distance is automatically shortened to improve detection resolution. After continuous correction, a dynamically corrected flight path that adapts to changes in wind disturbance is obtained.

[0046] After establishing the dynamically corrected flight path, path planning and optimization are performed. The path is optimized using an improved A* algorithm or genetic optimization algorithm, aiming to minimize range energy consumption and flight attitude changes while ensuring trajectory smoothness and temporal continuity. After optimization, a corresponding calibration and following path is generated, which balances detection perspective, flight safety, and path smoothness.

[0047] Finally, the calibrated following path is used to configure the inspection route, forming a complete inspection task path for the UAV in following mode. Based on the inspection route, tasks are assigned and attitude control is performed, and the UAV performs inspection and management of the wind turbine blades along the calibrated following path.

[0048] Furthermore, in the method provided in the application embodiments, configuring the inspection route using the calibrated following path further includes:

[0049] The sampling density of the wind turbine blade's position trajectory is adaptively configured using the prior defect probability to establish non-uniform inspection trajectory points; path trajectory compensation is performed on the calibration and following path based on the non-uniform inspection trajectory points to establish a first compensation result; an attitude stabilization control model is constructed based on the first compensation result, and the attitude stabilization control model dynamically adjusts the UAV's attitude angle and heading speed according to the transient airflow distribution and disturbance perception factor of the wind turbine blade rotation to establish a dynamic attitude correction result; the dynamic attitude correction result is fused with the first compensation result to establish an inspection route.

[0050] In this embodiment, when adaptively configuring the sampling density of wind turbine blade position tracks using prior defect probabilities, a probability density mapping-based method is employed. First, the prior defect probabilities on the blade surface are normalized, limiting their distribution range to between 0 and 1. Then, a linear mapping algorithm establishes the correspondence between the prior defect probabilities and the track sampling density, with the sampling density expressed as the number of track points per unit area. When the prior defect probability is greater than 0.8, the sampling density is set to 9 track points per square meter; when the prior defect probability is between 0.5 and 0.8, the sampling density is 6 track points per square meter; and when the prior defect probability is less than 0.5, the sampling density is 3 track points per square meter. Spatial interpolation of the sampling density distribution generates a sampling density distribution map covering the entire blade surface. Based on this sampling density distribution map, the spacing between the UAV track points is adaptively adjusted, thereby generating non-uniform inspection track points.

[0051] Next, when compensating the path trajectory of the calibration and following path based on the non-uniform inspection trajectory points, a least-squares curve fitting method is used. The calibration and following path is paired with the non-uniform inspection trajectory points, and the spatial deviation between the trajectory points and the calibration path is calculated. Least-squares optimization is then used to minimize the sum of squares of the overall path deviation. This method smoothly corrects the geometric offset caused by the uneven distribution of trajectory points, generating a continuous and smooth path curve. The path obtained after compensation is the first compensation result.

[0052] Subsequently, when constructing the attitude stabilization control model based on the first compensation result, a PID control method was adopted. The transient airflow distribution and disturbance perception factor during the wind turbine blade rotation process were used as real-time input parameters. Wind speed, airflow direction, and attitude angles were acquired in real time through airflow sensors and an inertial measurement unit (IMU) installed on the UAV. The PID controller, using the attitude angle error as input, outputs control quantities for the UAV's pitch, roll, and yaw angles, and performs feedback calculations at 0.1-second intervals to achieve dynamic attitude correction. When the disturbance perception factor is less than 0.4, the attitude stabilization control model maintains a normal heading speed; when the disturbance perception factor is greater than 0.7, the heading speed is automatically reduced by 20% to maintain flight stability. The control parameters output in this process are the dynamic attitude correction results.

[0053] Finally, the dynamic attitude correction result is fused with the first compensation result. This process begins by acquiring time-series wind data and weather forecast data, performing transient wind fluctuation analysis, and establishing a time-series wind transient response node. Then, the tolerance space is configured using the time-series wind transient response node to correct the first compensation result, yielding the second compensation result. Finally, the second compensation result is fused with the dynamic attitude correction result to establish the final inspection route, which is used by the UAV to perform inspection and management of the wind turbine blades.

[0054] Furthermore, in the method provided in the application embodiments, fusing the dynamic attitude correction result with the first compensation result to establish an inspection route further includes:

[0055] Acquire time-series wind data and weather forecast data; perform transient wind fluctuation analysis based on the time-series wind data and weather forecast data, and establish a time-series wind transient response node; after configuring the tolerance space using the time-series wind transient response node, correct the first compensation result and establish a second compensation result; fuse the second compensation result and the dynamic attitude correction result to establish an inspection route.

[0056] In this embodiment, to obtain time-series wind data and weather forecast data, wind speed and direction sensors installed on the top of the wind turbine nacelle and at the root of the blades are used to continuously collect real-time aerodynamic parameters such as wind speed, wind direction, and air pressure at a sampling interval of 1 second, forming a time-series wind data set covering the entire operation of the wind turbine. Simultaneously, weather forecast data covering the area where the wind turbine is located is obtained from a meteorological monitoring system, including parameters such as average wind speed, gust frequency, wind direction change rate, and airflow shear intensity for the next 10 minutes. Based on aerodynamic relationships, the kinetic energy density at each sampling time is calculated using the formula: kinetic energy density = 0.5 × ρ × v 2 Where ρ is taken as 1.225 kg / m 3 v is the wind speed measured in real time.

[0057] Next, when performing transient wind force fluctuation analysis based on time-series wind force datasets and weather forecast data, the Fast Fourier Transform (FFT) method is used to perform spectral analysis on the air kinetic energy density sequence to identify the main frequency components and fluctuation amplitude characteristics of the energy distribution. Combined with the time-varying sequence of wind direction angle, an energy threshold determination method is used to extract time points where wind speed changes are significant. When the air kinetic energy density exceeds 1.5 times the average wind energy density and the wind direction change rate exceeds 10° / s, this time point is determined as the transient response node of the time-series wind force.

[0058] After configuring the tolerance space using the time-series wind transient response nodes, the first compensation result is corrected. In this process, firstly, with each time-series wind transient response node as the center, the airflow disturbance influence radius Δr = v′ × t is calculated based on the wind speed change amplitude corresponding to that node, where v′ is the wind speed change value at that node, and t is the UAV attitude adjustment response time. Secondly, the disturbance influence radius of each node is spatially mapped onto the original UAV trajectory, generating a circular local buffer zone around each trajectory point to describe the permissible range of airflow fluctuations for UAV flight safety. Then, adjacent buffer zones are spatially stitched and curve-fitted according to the node time sequence to construct a tolerance space covering the entire detection flight path. Finally, the tolerance space radius is adjusted according to the root mean square value of the wind speed fluctuation: the radius is set to 0.3m when the root mean square value is less than 1.5m / s, and to 0.8m when the root mean square value is greater than 3.0m / s. Spline interpolation is used to smooth the spatial boundary. This method completes the airflow adaptive correction of the first compensation result, yielding the second compensation result.

[0059] Finally, the second compensation result and the dynamic attitude correction result are fused. A weighted fusion method is used to synchronize the two sets of data in time and match their spatial coordinates. The path coordinates in the second compensation result are the main input, and the attitude angle adjustment and heading speed change in the dynamic attitude correction result are the weighted inputs. The weight ratio is set to 0.6 for path information and 0.4 for attitude information. Through fusion calculation, an inspection route that comprehensively considers flight path smoothness and attitude stability is formed.

[0060] Furthermore, the method provided in the application embodiments also includes:

[0061] If the decision result is a static suspension mode decision, then the dynamic passing angle interval of the blade tip and middle is extracted according to the blade's temporal motion trajectory, and a temporal channel model of the blade is established. The temporal channel model models the blade's rotational motion as a time-space channel to predict the blade's temporal spatial position. Based on the temporal channel model and the disturbance sensing factor, a static suspension safety interval is determined, and a set of UAV static suspension points is constructed. The prior defect probability is used to perform confidence-weighted selection on the UAV static suspension point set to determine the target static suspension detection point. The inspection route is configured using the target static suspension detection point to perform the inspection management of the wind turbine blades.

[0062] In this embodiment, if the decision result is a static suspension mode decision, the dynamic clearance angle range of the blade tip and middle is first extracted based on the blade's temporal motion trajectory. A rotary encoder is used to collect the corresponding sequence of blade rotation angles and time, with a sampling frequency of 100 Hz. The instantaneous angle change of the blade within one rotation cycle is calculated by integrating the angular velocity. Combining the blade length, installation tilt angle, and propeller pitch angle, the starting and ending ranges of the clearance angle of the blade tip and middle within the rotation cycle are obtained using inverse kinematics calculation methods, thus yielding the dynamic clearance angle range of the blade tip and middle.

[0063] After extracting the dynamic passing angle range, a time-series channel model of the blade is established. This model models the blade rotational motion as a time-space channel that maps time and space, used to predict the blade's spatial position at any given time. Specifically, by employing a time-space coupled modeling method, using rotation angle and time as input variables, parameters such as blade length, pitch angle, and rotation radius are substituted into the coordinate transformation equation to calculate the three-dimensional spatial coordinates x(t), y(t), and z(t) of the blade tip and center at time t. Through continuous time interpolation, the spatial trajectory sequence of the blade within a complete rotation cycle is obtained, thus establishing a time-space channel reflecting the entire blade rotation process, achieving accurate prediction of the blade's time-series spatial position, and obtaining the blade's time-series channel model.

[0064] Next, the safe hovering range is determined based on the time-series channel model and the disturbance sensing factor. In this process, a dynamic wind field overlap analysis method is used to match the spatial position of the blade at different times with the airflow disturbance intensity corresponding to the disturbance sensing factor. By calculating the minimum safe distance between the blade spatial envelope and the UAV hovering area, and determining whether the disturbance sensing factor is less than 0.3 at the corresponding time, if the minimum safe distance is greater than 1.2 meters and the disturbance sensing factor meets the threshold condition, the spatial position is determined as the safe hovering range. All spatial positions meeting the conditions within one rotation cycle are filtered for temporal continuity to extract spatial points that are stable in time, ultimately obtaining the UAV hovering point set.

[0065] Subsequently, a confidence-weighted selection is performed on the set of UAV hovering points using prior defect probabilities. In this process, a linear weighted scoring method is employed. For each hovering point, the prior defect probability of the corresponding blade grid is first read, and then combined with the spatial distance from the UAV to the center point of that grid for scoring calculation. To ensure comparability of data with different dimensions under the same scoring system, the spatial distance is normalized to the range of 0 to 1. The smaller the normalized distance, the closer its value is to 1, indicating better detection conditions. Then, by linearly weighting the prior defect probability and spatial distance, assigning a weight of 0.7 to the prior defect probability and a weight of 0.3 to the spatial distance, the detection confidence value for each hovering point is obtained. When the prior defect probability is high and the spatial distance is small, the detection confidence value increases accordingly. The detection confidence values ​​of all hovering points are calculated and ranked, and the hovering point with the highest confidence is selected as the target hovering detection point.

[0066] Finally, the inspection route is configured using the target static suspension detection points to perform inspection management of the wind turbine blades. In this process, a detection time window sequence is first set according to the rotation cycle of the wind turbine blades, and dynamic data acquisition is triggered as the blades pass through each time window. Then, the UAV's static suspension attitude self-balancing control is performed based on real-time wind field disturbance signals to generate time-stable observation attitude compensation, ensuring observation stability. Finally, this time-stable observation attitude compensation is used to control the UAV to synchronously sample within multiple rotation cycles, forming a dynamic defect dataset of the blades. Based on this dataset, inspection management is performed to achieve continuous monitoring and data updates of the blade's operating status.

[0067] Furthermore, in the method provided in the application embodiments, the method of configuring inspection routes using the target static suspension detection points to perform inspection management of wind turbine blades further includes:

[0068] The detection time window sequence is configured according to the rotation period of the wind turbine blade and the target static suspension detection point, and dynamic trigger acquisition is performed when the blade passes through each detection time window; self-balancing control of static suspension attitude is performed according to real-time wind field disturbance signal, and time-series stable observation attitude compensation is established; the time-series stable observation attitude compensation is used to control the UAV to perform synchronous sampling in multiple rotation periods, establish dynamic defect dataset of blade, and perform inspection management according to the dynamic defect dataset.

[0069] In this embodiment, when configuring a detection time window sequence based on the rotation cycle of the wind turbine blades and the target static suspension detection point, and performing dynamic trigger acquisition as the blade passes, a phase-locked triggering method is adopted. The sequence of changes in rotational speed and angle over time is obtained through a main shaft angular velocity sensor. One rotation cycle is divided into several detection time windows according to fixed phase intervals, and each target static suspension detection point is associated with a corresponding phase. When the phase count reaches a preset threshold, the acquisition device is triggered to operate within that detection time window, realizing dynamic trigger acquisition as the blade passes, and obtaining the detection time window sequence and the dynamically triggered acquisition data bound to it.

[0070] Next, when performing self-balancing control of the static hovering attitude based on real-time wind disturbance signals and establishing time-stable observation attitude compensation, a PID control method is adopted. Wind speed changes and attitude angle error signals are acquired in real time through a wind speed sensor and an inertial measurement unit. The errors are input into the proportional loop, integral loop, and derivative loop to calculate the control quantity, which is continuously applied to the pitch angle, roll angle, and yaw angle channels. This allows the UAV to counteract the attitude deviation caused by disturbances while in a static hovering state. The attitude correction amount within each detection time window is recorded in chronological order as a compensation table, resulting in time-stable observation attitude compensation.

[0071] Subsequently, when using time-stable observation attitude compensation control to perform synchronous sampling and establish a dynamic defect dataset of the blades in multiple rotation cycles, a time synchronization method was adopted. The detection time window sequence was used as the time reference to align the trigger time of camera exposure laser ranging with infrared acquisition. At the same time, the attitude compensation table of the corresponding time window was loaded before triggering to ensure geometric consistency of imaging. The above alignment and acquisition operations were repeated for several consecutive rotation cycles. Multi-cycle samples of the same position and phase were archived into the same record to obtain a dynamic defect dataset of the blades covering multiple rotation cycles.

[0072] Finally, inspection management is performed based on the dynamic defect dataset. This process begins by identifying the defect levels of the wind turbine blades using the dynamic defect dataset. Classification is based on characteristic parameters such as defect area, abnormal temperature amplitude, and texture change rate, generating defect warning signals. Subsequently, wind turbine anomalies are reported based on these warning signals. When the warning signal reaches a preset warning threshold, a shutdown re-acquisition task is automatically established to perform higher-precision data compensation acquisition under shutdown conditions. After acquisition, the new detection data is compared and analyzed with the original dynamic defect data, and the original defect warning signals are corrected and updated to ensure the accuracy and reliability of the warning results. This achieves closed-loop management and dynamic monitoring of the wind turbine blade operating status.

[0073] Furthermore, in the method provided in the application embodiments, the inspection management based on the dynamic defect dataset further includes:

[0074] The defect level of the wind turbine blades is identified using the dynamic defect dataset, and a defect early warning signal is established. The wind turbine anomaly is reported based on the defect early warning signal. If the defect early warning signal meets the preset early warning threshold, a shutdown re-acquisition task is established. After compensation acquisition based on the shutdown re-acquisition task, the defect early warning signal is corrected and reported.

[0075] In this embodiment, when identifying the defect level of wind turbine blades using a dynamic defect dataset, a multi-source data fusion method is first used to spatially register infrared thermal image data and visible light images to ensure that the two types of data correspond to the same blade surface position in the same coordinate system. Then, the temperature change rate of the blade under adjacent rotation cycles is extracted using infrared differential analysis, and features such as texture energy, contrast, and entropy values ​​of the blade surface are extracted using gray-level co-occurrence matrix texture analysis to form a blade defect feature set. Next, a fuzzy clustering algorithm is used to classify and analyze the defect feature set, and a defect warning signal intensity value is calculated for each blade grid. The defect warning signal intensity value ranges from 0 to 1, comprehensively considering three indicators: temperature anomaly amplitude, texture change rate, and surface deformation. The weight of temperature anomaly amplitude is 0.5, the weight of texture change rate is 0.3, and the weight of surface deformation is 0.2. A defect warning signal intensity value greater than 0.4 is identified as a mild defect, greater than 0.7 as a moderate defect, and greater than 0.9 as a severe defect. Through the above steps, the defect warning signal intensity value of each blade area is obtained, and a defect warning signal is established accordingly.

[0076] Subsequently, when reporting wind turbine anomalies based on defect warning signals, a dynamic threshold triggering method is employed to monitor the changes in the intensity of defect warning signals in each blade area in real time. When the intensity value of the defect warning signal in a certain blade area exceeds the preset warning threshold of 0.8, an abnormal event is automatically triggered, and a wind turbine anomaly reporting result is generated. This wind turbine anomaly reporting result includes the blade number, detection time, detection location, and defect warning signal intensity value, and is used for real-time risk assessment and maintenance scheduling of wind turbine operation status.

[0077] Once the defect warning signal strength value meets the preset warning threshold, a shutdown re-acquisition task is established. A shutdown acquisition command is issued using a task scheduling algorithm, and under wind turbine shutdown conditions, a UAV performs high-precision short-range compensation acquisition. The UAV acquires infrared thermal images, visible light images, and depth information of the blade surface at a 10-centimeter grid step, obtaining high-resolution compensation data. After acquisition, an image registration algorithm is used to spatially align the compensated acquisition data with the original dynamic defect dataset, and residual analysis is used to correct the boundary and temperature distribution of the defect area. Using the corrected data, the defect warning signal strength value is recalculated, updating and calibrating the original signal. When the corrected defect warning signal strength value is below 0.6, the area is marked as a stable observation state; when the corrected signal strength value is still above 0.8, the abnormal reporting state is maintained. Through these steps, the corrected defect warning signal reporting is finally achieved, yielding the corrected defect warning signal reporting result.

[0078] Furthermore, the method provided in the application embodiments, which utilizes the prior defect probability and downtime cost to make a detection balance decision, further includes:

[0079] If the detection balance decision is a shutdown detection decision, then a static inspection trajectory under the detection density constraint is configured according to the spatial distribution of the wind turbine blades after shutdown and the prior defect probability; the static inspection trajectory is used as the target trajectory, and the UAV's follow-up control optimization is performed, and the follow-up control optimization result is used to perform inspection management.

[0080] In this embodiment, when the detection balance decision is a shutdown detection decision, a static inspection trajectory under detection density constraints is first configured based on the spatial distribution of the wind turbine blades after shutdown and the prior defect probability. The spatial attitude data of the blades in the shutdown state is acquired using a 3D laser scanning system, the 3D coordinate point cloud of the blades is extracted, and the blade shape is reconstructed using a surface fitting method. Combining the prior defect probability distribution of the blades, a detection density constraint method is adopted, setting a higher sampling density in areas with a higher prior defect probability and a lower sampling density in areas with a lower prior defect probability, thus achieving adaptive configuration of detection resources. Specifically, when the prior defect probability is greater than 0.7, the sampling interval is set to 5 cm; when the prior defect probability is between 0.4 and 0.7, the sampling interval is set to 10 cm; and when the prior defect probability is less than 0.4, the sampling interval is set to 20 cm. This method obtains a non-uniform sampling path covering the entire blade surface, forming a static inspection trajectory.

[0081] Next, using the static inspection trajectory as the target trajectory, the UAV's following control optimization is performed. During this process, by calculating the spatial deviation between the UAV's current position and the target trajectory, a minimum mean square error (MMS) control method is used to correct the position and attitude in real time, enabling the UAV to fly smoothly along the set trajectory during the inspection. Then, a genetic algorithm is used to optimize the flight speed and turning radius, reducing energy consumption, minimizing attitude fluctuations, and improving trajectory tracking accuracy. After optimization, the optimal flight parameters and trajectory control results for the UAV under static inspection conditions are obtained, forming the following control optimization result.

[0082] Finally, the follow-up control optimization results are used to perform inspection management. Based on the optimized flight parameters and static inspection trajectory, the order of inspection tasks is automatically planned, and the UAV is controlled to synchronously acquire infrared thermal images and visible light images of each detection point in sequence, so as to realize high-precision inspection operation based on shutdown detection decision.

[0083] In summary, the embodiments of this application have at least the following technical effects:

[0084] This application performs gridded processing of wind turbine blades, calculates the prior defect probability of each blade grid, and constructs the prior defect probability using historical data, thermal image data, and long-range visual data. It then uses the prior defect probability and downtime cost to make a detection balance decision. If the detection balance decision is to enable detection, it acquires the temporal motion trajectory of the wind turbine blades and configures the temporal target detection position based on the temporal motion trajectory and the prior defect probability. It acquires the structural data of the wind turbine blades and uses the structural data and the temporal motion trajectory to perceive wind field disturbances caused by blade rotation, establishing a disturbance perception factor. It then uses the disturbance perception factor and the temporal target detection position to execute follow-up mode and static suspension mode decisions, configures the inspection route based on the decision results, and performs wind turbine blade inspection management. This invention solves the technical problems of unintelligent route planning, low detection efficiency, and significant susceptibility to wind field disturbances in the prior art during wind turbine blade inspection. By introducing prior defect probability and disturbance perception factor for intelligent route planning and dynamic decision-making, it achieves the technical effect of improving inspection accuracy and stability.

[0085] Example 2, based on the same inventive concept as the wind turbine inspection route planning method using a drone equipped with intelligent sensors in the foregoing examples, such as... Figure 2 As shown, this application provides a wind turbine inspection route planning system for UAVs equipped with intelligent sensors. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0086] The gridding module 11 is used to perform gridding processing of the wind turbine blades and calculate the prior defect probability of each blade grid. The prior defect probability is constructed using historical data, thermal image data, and long-range visual data. The decision module 12 is used to make a detection balance decision using the prior defect probability and downtime cost. The detection position configuration module 13 is used to acquire the temporal motion trajectory of the wind turbine blades if the detection balance decision is to make a start-up detection decision, and configure the temporal target detection position according to the temporal motion trajectory and the prior defect probability. The disturbance perception module 14 is used to acquire the structural data of the wind turbine blades, perform wind field disturbance perception based on the structural data and the temporal motion trajectory of the blades, and establish a disturbance perception factor. The inspection management module 15 is used to execute follow mode and static suspension mode decisions using the disturbance perception factor and the temporal target detection position, configure the inspection route using the decision results, and perform inspection management of the wind turbine blades.

[0087] Furthermore, the system is also used to implement the following functions:

[0088] Configure a spatial trajectory sequence with temporal identifiers based on the time-series target detection location; if the decision result is a follow-up mode decision, then use the spatial trajectory sequence as a reference to construct a safe following curve for the UAV; perform position correction of the safe following curve based on the disturbance perception factor to establish a dynamically corrected flight path; perform path planning optimization of the dynamically corrected flight path, and use the path planning optimization result to establish a calibrated following path; configure an inspection flight path using the calibrated following path to perform inspection management of the wind turbine blades.

[0089] Furthermore, the system is also used to implement the following functions:

[0090] The sampling density of the wind turbine blade's position trajectory is adaptively configured using the prior defect probability to establish non-uniform inspection trajectory points; path trajectory compensation is performed on the calibration and following path based on the non-uniform inspection trajectory points to establish a first compensation result; an attitude stabilization control model is constructed based on the first compensation result, and the attitude stabilization control model dynamically adjusts the UAV's attitude angle and heading speed according to the transient airflow distribution and disturbance perception factor of the wind turbine blade rotation to establish a dynamic attitude correction result; the dynamic attitude correction result is fused with the first compensation result to establish an inspection route.

[0091] Furthermore, the system is also used to implement the following functions:

[0092] Acquire time-series wind data and weather forecast data; perform transient wind fluctuation analysis based on the time-series wind data and weather forecast data, and establish a time-series wind transient response node; after configuring the tolerance space using the time-series wind transient response node, correct the first compensation result and establish a second compensation result; fuse the second compensation result and the dynamic attitude correction result to establish an inspection route.

[0093] Furthermore, the system is also used to implement the following functions:

[0094] If the decision result is a static suspension mode decision, then the dynamic passing angle interval of the blade tip and middle is extracted according to the blade's temporal motion trajectory, and a temporal channel model of the blade is established. The temporal channel model models the blade's rotational motion as a time-space channel to predict the blade's temporal spatial position. Based on the temporal channel model and the disturbance sensing factor, a static suspension safety interval is determined, and a set of UAV static suspension points is constructed. The prior defect probability is used to perform confidence-weighted selection on the UAV static suspension point set to determine the target static suspension detection point. The inspection route is configured using the target static suspension detection point to perform the inspection management of the wind turbine blades.

[0095] Furthermore, the system is also used to implement the following functions:

[0096] The detection time window sequence is configured according to the rotation period of the wind turbine blade and the target static suspension detection point, and dynamic trigger acquisition is performed when the blade passes through each detection time window; self-balancing control of static suspension attitude is performed according to real-time wind field disturbance signal, and time-series stable observation attitude compensation is established; the time-series stable observation attitude compensation is used to control the UAV to perform synchronous sampling in multiple rotation periods, establish dynamic defect dataset of blade, and perform inspection management according to the dynamic defect dataset.

[0097] Furthermore, the system is also used to implement the following functions:

[0098] The defect level of the wind turbine blades is identified using the dynamic defect dataset, and a defect early warning signal is established. The wind turbine anomaly is reported based on the defect early warning signal. If the defect early warning signal meets the preset early warning threshold, a shutdown re-acquisition task is established. After compensation acquisition based on the shutdown re-acquisition task, the defect early warning signal is corrected and reported.

[0099] Furthermore, the system is also used to implement the following functions:

[0100] If the detection balance decision is a shutdown detection decision, then a static inspection trajectory under the detection density constraint is configured according to the spatial distribution of the wind turbine blades after shutdown and the prior defect probability; the static inspection trajectory is used as the target trajectory, and the UAV's follow-up control optimization is performed, and the follow-up control optimization result is used to perform inspection management.

[0101] Furthermore, the system is also used to implement the following functions:

[0102] A risk energy distribution model for the blade region is constructed based on the prior defect probability. The risk energy characterizes the potential failure accumulation trend of each region of the blade under continuous wind turbine operation. Real-time operating parameters and downtime cost data of the wind turbine are obtained, and the energy decay function of the blade's current operating state is calculated. The risk energy distribution model and the energy decay function are subjected to dynamic game analysis to establish a detection balance spectrum. The detection balance spectrum characterizes the detection benefit boundary under different blade states. A detection balance decision is made based on the detection balance spectrum.

[0103] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0104] The above description is only a preferred embodiment of this application and is not intended to limit this application. 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.

[0105] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for planning a fan inspection route of an unmanned aerial vehicle carrying an intelligent sensor, characterized in that, The method includes: The wind turbine blades are meshed, and the prior defect probability of each blade mesh is calculated. The prior defect probability is constructed using historical data, thermal image data, and long-range visual data. The prior defect probability and downtime cost are used to make a detection balance decision; If the detection balance decision is the start-up detection decision, then the blade temporal motion trajectory of the wind turbine blade is obtained, and the temporal target detection position is configured according to the blade temporal motion trajectory and the prior defect probability. Obtain structural data of wind turbine blades, and based on the structural data and the time-series motion trajectory of the blades, sense wind field disturbances during blade rotation and establish disturbance sensing factors. The following mode and static suspension mode decisions are made using the disturbance sensing factor and the time-series target detection position. The inspection route is configured using the decision results, and the inspection management of the wind turbine blades is performed. The detection balance decision is made using the prior defect probability and downtime cost, including: Based on the prior defect probability, a risk energy distribution model for the blade region is constructed. The risk energy characterizes the potential failure accumulation trend of each region of the blade under continuous operation of the wind turbine. Obtain real-time operating parameters and downtime cost data of the wind turbine, and calculate the energy decay function of the blades under current operating conditions; The risk energy distribution model and the energy decay function are subjected to dynamic game analysis to establish a detection equilibrium spectrum, which represents the detection benefit boundary under different blade states. The detection balance decision is made based on the detection balance spectrum. 2.The unmanned aerial vehicle-mounted intelligent sensor fan inspection route planning method of claim 1, wherein, Configure inspection routes based on decision results and implement inspection management for wind turbine blades, including: Configure a spatial trajectory sequence with temporal identifier based on the temporal target detection location; If the decision result is a follower mode decision, then the spatial trajectory sequence is used as a reference to construct a safe following curve for the UAV. Based on the disturbance perception factor, the position of the safety following curve is corrected, and a dynamically corrected flight path is established. Perform path planning and optimization for the dynamically corrected route, and use the path planning and optimization results to establish a calibration and following path; The inspection route is configured using the calibrated following path to perform inspection management of the wind turbine blades. 3.The unmanned aerial vehicle-mounted intelligent sensor fan inspection route planning method of claim 2, wherein, Configuring inspection routes using the calibrated following path includes: The sampling density of the wind turbine blade's position trajectory is adaptively configured using the prior defect probability to establish a non-uniform inspection trajectory point. Based on the non-uniform inspection trajectory points, the path trajectory compensation of the calibrated following path is performed to establish the first compensation result; Based on the first compensation result, an attitude stabilization control model is constructed. The attitude stabilization control model dynamically adjusts the attitude angle and heading speed of the UAV according to the transient airflow distribution of the wind turbine blades and the disturbance perception factor, and establishes dynamic attitude correction results. The dynamic attitude correction result is fused with the first compensation result to establish an inspection route. 4.The unmanned aerial vehicle (UAV) fan inspection route planning method of claim 3, wherein, The dynamic attitude correction result is fused with the first compensation result to establish an inspection route, including: Acquire time-series wind data sets and weather forecast data; Based on the time-series wind data set and the weather forecast data, transient wind fluctuation analysis is performed to establish a time-series wind transient response node. After configuring the tolerance space using the time-series wind transient response node, the first compensation result is corrected to establish a second compensation result; The second compensation result and the dynamic attitude correction result are fused to establish an inspection route. 5.The unmanned aerial vehicle (UAV) fan inspection route planning method of claim 2, wherein, The method further includes: If the decision result is a static suspension mode decision, then the dynamic passing angle range of the blade tip and middle is extracted according to the blade temporal motion trajectory, and a temporal channel model of the blade is established. The temporal channel model models the blade rotational motion as a time-space channel, which is used to predict the temporal spatial position of the blade. Based on the time-series channel model and the disturbance perception factor, the static hovering safety zone is determined, and a set of UAV static hovering points is constructed. The target static suspension detection point is determined by using the prior defect probability to perform confidence-weighted selection on the set of UAV static suspension points; The inspection route is configured using the target static suspension detection point to perform inspection management of the wind turbine blades. 6.The UAV-mounted intelligent sensor fan inspection route planning method of claim 5, wherein, Using the target static suspension detection points to configure inspection routes, the inspection management of wind turbine blades is performed, including: Configure the detection time window sequence according to the rotation period of the wind turbine blades and the target static suspension detection point, and perform dynamic trigger acquisition of the blade passage in each detection time window; Self-balancing control of static suspension attitude is performed based on real-time wind field disturbance signals, and time-series stable observation attitude compensation is established. The timing-stable observation attitude compensation control UAV is used to perform synchronous sampling in multiple rotation cycles to establish a dynamic defect dataset of the blades, and inspection management is carried out based on the dynamic defect dataset. 7.The UAV-mounted intelligent sensor fan inspection route planning method of claim 6, wherein, Inspection management is performed based on the dynamic defect dataset, including: The dynamic defect dataset is used to identify the defect level of wind turbine blades and establish a defect early warning signal. The wind turbine is reported as an anomaly based on the defect warning signal. If the defect warning signal meets the preset warning threshold, a shutdown re-acquisition task is established. After compensation acquisition based on the shutdown re-acquisition task, the defect warning signal is corrected and reported. 8.The unmanned aerial vehicle (UAV) fan inspection route planning method of claim 1, wherein, The detection balance decision is made using the prior defect probability and downtime cost, including: If the detection balance decision is a shutdown detection decision, then the static inspection trajectory under the detection density constraint is configured according to the spatial distribution of the wind turbine blades after shutdown and the prior defect probability. Using the static inspection trajectory as the target trajectory, the drone performs follow-up control optimization, and the follow-up control optimization results are used to perform inspection management.

9. A UAV-mounted intelligent sensor fan inspection route planning system, characterized in that, The system is used to execute the wind turbine inspection route planning method for a UAV equipped with intelligent sensors as described in any one of claims 1-8, and the system includes: The meshing module is used to perform meshing processing of wind turbine blades and calculate the prior defect probability of each blade mesh. The prior defect probability is constructed using historical data, thermal image data, and long-distance visual data. The decision module is used to make a detection balance decision based on the prior defect probability and downtime cost; The detection location configuration module is used to acquire the blade temporal motion trajectory of the wind turbine blades if the detection balance decision is the power-on detection decision, and configure the temporal target detection location according to the blade temporal motion trajectory and the prior defect probability. The disturbance sensing module is used to acquire the structural data of the wind turbine blades, and to perform wind field disturbance sensing of the blade rotation based on the structural data and the blade time-series motion trajectory, and to establish a disturbance sensing factor. The inspection management module is used to perform follow-up mode and static suspension mode decisions using the disturbance sensing factor and the time-series target detection position, configure the inspection route using the decision results, and perform inspection management of the wind turbine blades.

Images