Vision-enhanced low-visibility fan auxiliary hoisting system and method

Abstract

The invention relates to the technical field of fan hoisting, and discloses a vision-enhanced low-visibility fan auxiliary hoisting system and method, and the system comprises an environment acquisition module, an intelligent processing and decision module and a hoisting installation module. The method comprises the following steps: acquiring image data of a working site, and transmitting the image data to the intelligent processing and decision module through wireless communication; constructing a hoisting equipment fault prediction model, carrying out early warning on abnormal conditions of the image data, and identifying sudden weather changes affecting hoisting operation through meteorological data analysis; according to a prediction result of the hoisting equipment fault prediction model, a hoisting scheme is adjusted, and an emergency plan is formulated; and carrying out real-time monitoring by using an unmanned aerial vehicle, comparing real-time monitored image data with a prediction result, and adjusting the hoisting equipment fault prediction model based on a comparison result. Through a data comparison and feedback mechanism, the hoisting equipment fault prediction model is continuously optimized, and the prediction precision and reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fan hoisting, and particularly to a vision - enhanced low - visibility fan auxiliary hoisting system and method. Background Art

[0002] Wind power generation, as a clean and renewable energy source, has occupied an important position in the global energy structure. It converts wind energy into electrical energy, reducing greenhouse gas emissions and air pollution. Wind energy resources are rich and widely distributed. The global wind energy is about 2.74×10 9 MW, and the exploitable wind energy is about 2×10 7 MW, which is 10 times larger than the total exploitable water energy on the earth. With the progress of technology and the reduction of costs, the economy and competitiveness of wind power generation have been improved.

[0003] Currently, the hoisting of fans requires large - scale cranes and professional hoisting teams, which not only increases the time and labor costs but also prolongs the construction period. Moreover, fan hoisting is a high - altitude operation, and workers need to operate at high altitudes, posing safety hazards such as high - altitude falls and object drops. In addition, during the hoisting process, mechanical failures or operational errors of the equipment may occur, resulting in damage to the fan or lifting equipment and even causing casualties. Furthermore, currently, fan hoisting can only be carried out during the day. At night, on cloudy days or in foggy weather, due to low visibility and great difficulty in supplementary lighting, there are problems such as difficult positioning and high risk coefficients during hoisting.

[0004] Prior Art One, a Chinese patent with the patent number: 202411582515.7 discloses a monitoring device and method for the blade sweeping of a wind power generation unit. The device includes a fan main control, a pitch driver, and a metal - coated net arranged on the blade; a reference resistor is arranged at both ends of the measurement port of the pitch driver; the fan main control is connected to the pitch driver; the wire resistance and the metal - coated net are connected in series and then connected in parallel with the reference resistor; the metal - coated net is composed of sequentially arranged metal - coated strips, and both ends of the metal - coated net are electrically connected to the embedded wiring posts in the internal cavity of the blade; the blade body embedded wiring posts are used to connect the metal - coated net to the pitch driver through a wire during the fan hoisting and commissioning stage; the wire resistance is the resistance of the wire between the blade embedded wiring post and the pitch driver. Although it can accurately identify dangerous working conditions of tower sweeping and control the unit to complete safe pitch - down, improving the accuracy of fault identification and the safety of the unit; however, at night, on cloudy days or in foggy weather, due to low visibility and great difficulty in supplementary lighting, there are problems such as difficult positioning and high risk coefficients during hoisting.

[0005] Prior Art Two, a Chinese patent with the patent number 202410657868.2, relates to a method, device, electronic device, and storage medium for safety control in wind farm operations, belonging to the field of operation safety management. The method includes: obtaining a hoisting task of installing a hoisting component to a target position; obtaining image information captured by multiple cameras and weather information of the wind farm operation site; determining a first hoisting movement area and a first safe operation area in a virtual model according to the hoisting task and the weather information; determining a second hoisting movement area and a second safe operation area corresponding to the second hoisting movement area in each piece of image information according to the first hoisting movement area and the first safe operation area; determining the pose information of the hoisting component and the position information of the worker according to the image; if it is determined according to the pose information of the hoisting component that the hoisting component is not in the second hoisting movement area and / or it is determined according to the position information of the worker that the worker is not in the second safe operation area, an alarm message is generated. Although it improves the operation safety during the hoisting of the wind turbine, it does not consider that in the night, cloudy days or foggy days, due to low visibility, it is difficult to supplement light, and there are difficulties in positioning and high risk coefficients during hoisting.

[0006] Prior Art Three, a Chinese patent with the patent number 202311533337.4, relates to the technical field of hoisting tools, and specifically provides an adaptable wind turbine hoisting tool and a hoisting system. The adaptable wind turbine hoisting tool includes: a tower barrel hoisting device and a blade hoisting device. Both the tower barrel hoisting device and the blade hoisting device include: a support device and a suspension rope, and the suspension rope is connected to the support device; the support device includes: a first support plate, and both ends of the support frame are connected to a fixed seat, and the fixed seat is fixed on the first support plate. A number of transverse support columns are respectively arranged on the opposite sides of the support frame; a tower barrel connection mechanism of the support device of the tower barrel hoisting device and a blade connection mechanism connected to the support device of the blade hoisting device. Although the suspension bearing force is adjusted by setting a number of transverse support columns, and an adjustable tower barrel connection mechanism and a blade connection mechanism are set to adapt to different sizes of wind turbine tower barrels and wind turbine blade sizes, it does not consider that in the night, cloudy days or foggy days, due to low visibility, it is difficult to supplement light, and there are difficulties in positioning and high risk coefficients during hoisting.

[0007] Currently, Prior Art One, Prior Art Two, and Prior Art Three have problems of difficult light supplement due to low visibility in the night, cloudy days or foggy days, and difficulties in positioning and high risk coefficients during hoisting. To solve the above problems, the present invention provides a vision-enhanced low-visibility wind turbine auxiliary hoisting system. Summary of the Invention

[0008] The main object of the present invention is to provide a vision-enhanced low-visibility wind turbine auxiliary hoisting system and method to solve the problems in the prior art that in the night, cloudy days or foggy days, due to low visibility, it is difficult to supplement light, and there are difficulties in positioning and high risk coefficients during hoisting.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A visual enhancement low visibility fan auxiliary hoisting system, said visual enhancement low visibility fan auxiliary hoisting system, comprising:

[0011] An environment acquisition module, configured to sense the position of obstacles and the relative distance of fan components by using lidar to scan the environment, and construct a three-dimensional point cloud map of the fan hoisting area; collect image data of the operation site and transmit it to the intelligent processing and decision-making module through wireless communication;

[0012] An intelligent processing and decision-making module, configured to perform defogging processing on the image data of the operation site and perform logarithmic operation on the foggy image; construct a hoisting equipment failure prediction model to give early warning of abnormal conditions in the image data, and identify sudden weather changes affecting hoisting operations through meteorological data analysis;

[0013] A hoisting and installation module, configured to adjust the hoisting plan according to the prediction result of the hoisting equipment failure prediction model and formulate an emergency plan; and use an unmanned aerial vehicle for real-time monitoring, compare the real-time monitored image data with the prediction result, and adjust the hoisting equipment failure prediction model based on the comparison result.

[0014] As a further improvement of the present invention, the environment acquisition module includes:

[0015] An unmanned aerial vehicle acquisition sub-module, configured to construct a three-dimensional point cloud map of the fan hoisting area through lidar, sense obstacles and the relative distance of fan components; detect obstacles at close range through millimeter wave radar;

[0016] A three-dimensional point cloud map construction sub-module, configured to obtain the distance and angle information between the obstacle area and the fan components by measuring the laser reflection time; the data of each sampling point includes the distance, the vertical angle with the horizontal, and the angle with the x-axis, and is converted into three-dimensional coordinates through calculation;

[0017] A three-dimensional map sub-module, configured to remove noise points generated by high-reflection / absorbing materials, dust or sensor errors, add RGB color information to the point cloud, extract the structural features of the hoisting area, segment different categories such as the ground and moving objects, and convert the processed point cloud data into a three-dimensional model and output a three-dimensional map.

[0018] As a further improvement of the present invention, the three-dimensional map sub-module includes:

[0019] A preprocessing unit, configured to remove noise points generated by high-reflection materials, dust or sensors through a reflection intensity threshold; align point clouds collected at multiple times or at different times to the same coordinate system; reduce the data volume through uniform sampling;

[0020] Surface reconstruction unit, which is used to extract the structural features of the hoisting area, segment the ground and the dynamic mapless type; segment the point cloud plane to obtain classified point clouds; attach RGB color information to the point clouds, color the point clouds using orthophotos, and fuse the color information with the color images obtained by the camera.

[0021] 3D model generation unit, which is used to generate a mesh model by fitting point clouds, parameterize the segmented point clouds into a 3D model through calculation; generate a triangular mesh model from the classified point clouds; convert the 3D model from the relative coordinate system to the absolute geographic coordinate system and register it with the remote sensing image; output a 3D map.

[0022] As a further improvement of the present invention, the intelligent processing and decision-making module includes:

[0023] Fog removal processing sub-module, which is used to confirm the obstacles on the hoisting path through the 3D model, collect the image data of the operation site in real time through a high-resolution camera, perform fog removal calculation on the real-time data to generate clear images; transmit the clear images to the ground operation station to construct an obstacle recognition model for recognition.

[0024] Fault parameter combination sub-module, which is used to form a structural data set based on the real-time operation data of the hoisting equipment, the historical hoisting equipment fault data, and the real-time 3D image; combine the historical hoisting equipment fault data to locate the key data affecting equipment anomalies and determine the parameter combinations related to the faults.

[0025] Fault prediction model construction sub-module, which is used to construct a hoisting equipment fault prediction model and train it based on historical fault data samples; input the real-time 3D image into the model to predict the fault risk; automatically trigger an alarm according to the prediction result and extract the maintenance plan.

[0026] As a further improvement of the present invention, the fog removal processing sub-module includes:

[0027] Color image unit, which is used to form a Retinex theory imaging model based on two parts: the incident component and the reflection component.

[0028] Logarithmization unit, which is used to perform logarithmization operations on the foggy images.

[0029] Obstacle recognition unit, which is used to obtain the reflection component in the logarithmic domain; construct an obstacle recognition model to automatically identify and dynamically track the fan components, hoisting equipment, and obstacles.

[0030] As a further improvement of the present invention, the obstacle recognition unit includes:

[0031] The acquisition and processing subunit is used to label 3D images, extract key features from sensor data, and convert them into a unified format; extract target feature representations through intermediate layers or global pooling layers; perform temporal modeling on the target motion trajectory to predict future positions;

[0032] The model adjustment subunit is used to combine the trajectory prediction error and the trajectory safety reward for tracking and path planning; dynamically adjust the learning rate to balance the convergence speed and stability, initialize the experience pool in reinforcement learning to store state-action pairs, and add noise; adjust the obstacle recognition model according to real-time environmental data;

[0033] The path planning subunit is used to regenerate the optimal path based on the perception data when it detects that the device deviates or the obstacles change dynamically, and adjust the actions of the lifting device through control instructions; use historical data to test the monitoring accuracy, tracking stability, and path planning rules of the obstacle recognition model.

[0034] As a further improvement of the present invention, the fault parameter combination sub-module includes:

[0035] The integrated data unit is used to centrally store and integrate real-time data, historical fault data, and 3D image data through the cloud platform into a unified structured data set; extract key parameters from the structured data set, consult historical fault records, and screen out parameters highly relevant to the known fault time;

[0036] The combination analysis unit is used to analyze the interactions between parameters, identify the cases where most faults are triggered by single-parameter or double-parameter combinations, and test the coverage of key parameter pairs; and calculate the contribution degrees of different parameter combinations to faults, and screen out the coefficient combinations higher than the preset value;

[0037] The key parameter combination is used to dynamically update the parameter combination by combining real-time monitoring data, determine the parameter combination related to the solid phase, and convert it into an actionable early warning indicator.

[0038] As a further improvement of the present invention, the fault prediction model construction sub-module includes:

[0039] The fault prediction model architecture design unit is used to divide the data set into a training set, a validation set, and a test set; use the training set to train the lifting device fault prediction model; input the 3D image as height, width, depth, and number of channels; input the time series as the time step and the feature dimension;

[0040] The fault prediction model architecture construction unit is used to extract spatial features, reduce the dimension by combining pooling layers; capture the time dependence of time series data, fuse the processed image and the processed time series to construct a multimodal network; use the fully connected layer to combine the classification task to output the fault probability;

[0041] A fault prediction model prediction unit is used to learn fault parameter combinations during the training process, preprocess real-time three-dimensional images, input the preprocessed real-time three-dimensional images into the hoisting equipment fault prediction model for evaluation, and output the fault probability and risk level.

[0042] As a further improvement of the present invention, the hoisting and installation module includes:

[0043] A pre-plan making sub-module is used to combine the results of the fault prediction model and the obstacle recognition model, automatically optimize the hoisting trajectory or increase temporary reinforcement measures according to the prediction results, deploy spare lifting tools, emergency power supplies and other resources in advance, and formulate targeted emergency plans;

[0044] An early warning sub-module is used to compare the real-time data with the predicted values of the fault prediction model. If the deviation value exceeds the preset threshold, a level-three early warning is triggered and pushed to the ground control station;

[0045] An adjustment model sub-module is used to mark abnormal data in the early warning time and incorporate it into the fault prediction model training library; if abnormal efficiency is found through multiple monitors, the environmental factor parameters are corrected accordingly; the optimized fault prediction model is reapplied to the new operation scenario.

[0046] To achieve the above object, the present invention also provides the following technical solutions:

[0047] A visual enhancement low visibility fan assisted hoisting method, which is applied to the visual enhancement low visibility fan assisted hoisting system. The visual enhancement low visibility fan assisted hoisting method includes:

[0048] By using lidar to scan the environment, sense the position of obstacles and the relative distance of fan components, and construct a three-dimensional point cloud map of the fan hoisting area; collect the image data of the operation site and transmit it to the intelligent processing and decision-making module through wireless communication;

[0049] Perform defogging processing on the image data of the operation site, and perform logarithmic operation on the foggy image; construct a hoisting equipment fault prediction model, give early warning of abnormal conditions of the image data, and identify sudden weather changes affecting hoisting operations through meteorological data analysis;

[0050] According to the prediction results of the hoisting equipment fault prediction model, adjust the hoisting plan and formulate an emergency plan; use drones for real-time monitoring, compare the real-time monitored image data with the prediction results, and adjust the hoisting equipment fault prediction model based on the comparison results.

[0051] According to the prediction results of the fault prediction model, the present invention dynamically adjusts the hoisting plan and formulates emergency measures; compares the real-time data collected by the drone with the prediction results, and adjusts the fault prediction model of the hoisting equipment based on the comparison results; by dynamically adjusting the hoisting plan, the operation delay caused by equipment failure or environmental change is reduced; the real-time monitoring and dynamic adjustment mechanism effectively reduces the potential risks in the hoisting operation and ensures the construction safety; through the data comparison and feedback mechanism, the fault prediction model of the hoisting equipment is continuously optimized to improve the prediction accuracy and reliability. Description of the Drawings

[0052] Figure 1 It is a schematic diagram of the function modules of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0053] Figure 2 It is a schematic diagram of the function modules of the environment acquisition module of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0054] Figure 3 It is a schematic diagram of the function modules of the 3D map sub-module of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0055] Figure 4 It is a schematic diagram of the function modules of the intelligent processing and decision-making module of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0056] Figure 5 It is a schematic diagram of the function modules of the dehazing processing sub-module of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0057] Figure 6 It is a schematic diagram of the function modules of the obstacle recognition unit of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0058] Figure 7 It is a schematic diagram of the function modules of the fault parameter combination sub-module of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0059] Figure 8 It is a schematic diagram of the function modules of the sub-module for constructing the fault prediction model of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0060] Figure 9 It is a schematic diagram of the function modules of the hoisting and installation module of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0061] Figure 10 It is a flow chart of the fan hoisting of an embodiment of the visual enhancement low visibility fan auxiliary hoisting system of the present invention;

[0062] Figure 11 This is a flowchart of the steps of an embodiment of the visual enhancement low visibility fan-assisted hoisting method of the present invention;

[0063] Figure 12 This is a schematic structural diagram of an embodiment of an electronic device of the present invention;

[0064] Figure 13 This is a schematic structural diagram of an embodiment of a storage medium of the present invention. Detailed implementation manners

[0065] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0066] The terms "first", "second", and "third" in the present invention are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first", "second", and "third" may explicitly or implicitly include at least one of such features. In the description of the present invention, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically and clearly defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the drawings). If the specific posture changes, the directional indications will also change accordingly. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes unlisted steps or units, or optionally further includes other steps or units inherent to these processes, methods, products, or devices.

[0067] Referring to "embodiments" herein means that the specific features, structures, or characteristics described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The phrase appears in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

[0068] Such as Figure 1As shown, this embodiment provides an embodiment of a visual enhancement low visibility fan assisted hoisting system. In this embodiment, the visual enhancement low visibility fan assisted hoisting system specifically includes:

[0069] An environment acquisition module 1, which is used to scan the environment using lidar, sense the positions of obstacles and the relative distances of fan components, and construct a three-dimensional point cloud map of the fan hoisting area; collect image data of the operation site and transmit it to the intelligent processing and decision-making module through wireless communication;

[0070] An intelligent processing and decision-making module 2, which is used to perform defogging processing on the image data of the operation site and perform logarithmic operation on the foggy image; construct a hoisting equipment fault prediction model, give early warnings of abnormal situations in the image data, and identify sudden weather changes affecting the hoisting operation through meteorological data analysis;

[0071] A hoisting and installation module 3, which is used to adjust the hoisting plan according to the prediction results of the hoisting equipment fault prediction model and formulate an emergency plan; and use a drone for real-time monitoring, compare the real-time monitored image data with the prediction results, and adjust the hoisting equipment fault prediction model based on the comparison results.

[0072] Preferably, in this embodiment, the environmental acquisition module 1 emits and receives laser beams through lidar, measures the time difference and phase difference of signals to determine the position; transmits the image data of the operation site to the intelligent processing and decision-making module through wireless communication technology; can construct a three-dimensional point cloud map of the wind turbine hoisting area in real time, accurately perceive the position of obstacles and the relative distance of wind turbine components, improve the safety and efficiency of hoisting operations; provide accurate environmental data support for subsequent modules to ensure the precise positioning and operation of hoisting equipment; the intelligent processing and decision-making module 2 performs defogging processing and logarithmic operation on foggy images to improve image clarity and readability; constructs a fault prediction model for hoisting equipment by using big data analysis and machine learning technologies based on historical data and real-time monitoring data; real-time monitors meteorological parameters such as wind speed and temperature through meteorological sensors and data acquisition modules, and identifies sudden weather changes; issues early warnings for abnormal situations to reduce potential safety hazards caused by equipment failures or bad weather; adjusts the hoisting plan according to the prediction results, formulates emergency plans, and improves the flexibility and safety of hoisting operations. The hoisting and installation module 3 uses a drone equipped with a lidar and a multi-gas monitoring system to collect environmental data of the hoisting area in real time; dynamically adjusts the hoisting plan according to the prediction results of the fault prediction model, and formulates emergency measures; compares the real-time data collected by the drone with the prediction results, and adjusts the fault prediction model of the hoisting equipment based on the comparison results; reduces operation delays caused by equipment failures or environmental changes by dynamically adjusting the hoisting plan; the real-time monitoring and dynamic adjustment mechanism effectively reduces potential risks in hoisting operations and ensures construction safety; continuously optimizes the fault prediction model of hoisting equipment through data comparison and feedback mechanisms to improve prediction accuracy and reliability.

[0073] Furthermore, as Figure 2 shown, the environmental acquisition module 1 specifically includes:

[0074] The drone acquisition sub-module 11 is used to construct a three-dimensional point cloud map of the wind turbine hoisting area through lidar, perceive obstacles and the relative distance of wind turbine components; detect obstacles at close range through millimeter-wave radar;

[0075] The three-dimensional point cloud map construction sub-module 12 is used to obtain the distance and angle information between the obstacle area and the wind turbine components by measuring the laser reflection time; the data of each sampling point includes the distance, the vertical angle with the horizontal, and the angle with the x-axis, and is converted into three-dimensional coordinates by calculation and image conversion;

[0076] The three-dimensional map sub-module 13 is used to remove noise points generated by high-reflection / absorption materials, dust or sensor errors, add RGB color information to the point cloud, extract the structural features of the hoisting area, segment different categories such as the ground and dynamic objects, convert the processed point cloud data into a three-dimensional model, and output a three-dimensional map.

[0077] Among them, the expression of the dynamic relative distance between the obstacle and the fan component in the UAV acquisition sub-module 11:

[0078]

[0079] In the formula, Δd represents the dynamic relative distance between the obstacle and the fan component; r2(t) - r1(t) represents the time-varying coordinates of the UAV and the obstacle; Δradar represents the wavelength of the millimeter-wave radar; erf(x) represents the error function, which is used for dynamic filtering; σscan represents the standard deviation of the lidar scanning error; ρradar represents the power density of the lidar signal; Fcalib represents the calibration factor, which is used to correct the error of the lidar measurement; t represents the time variable, which represents the time point in the dynamic process; θ r represents the main lobe width (beam angle) of the lidar antenna; t1, t2 represent the upper and lower limits of time integration, which represent the time interval of signal acquisition;

[0080] The expression of the point cloud data in the three-dimensional space in the three-dimensional point cloud map construction sub-module 12:

[0081]

[0082] In the formula, P 3D represents the point cloud data in the three-dimensional space, which is a set of spatial points collected by the lidar or other three-dimensional scanning devices. Each point (x, y, z) represents a specific spatial position in the three-dimensional coordinate system; F represents the Fourier transform; F -1 represents the inverse Fourier transform; j represents the imaginary unit; Hankel(k, r) represents the Hankel function, which is used to describe the wave motion in cylindrical coordinates; k represents the wave number; r represents the radial distance; Fτ lidar (t) represents the Fourier transform of the lidar signal, which converts the time-domain signal into the frequency domain for spectral analysis; c represents the speed of light; Θ(φ, ψ) represents the angular function, which describes the spatial distribution characteristics of the signal; R(α, β) represents the rotation matrix, which represents the rotation of the coordinate system in the three-dimensional space.

[0083] The above formula: the expression of the dynamic relative distance between the obstacle and the wind turbine component. When the radar wave encounters an object, it will be reflected back, and the distance can be determined by measuring the round-trip time of the wave. The radar wavelength in the formula is used to calculate the reflection characteristics of the radar wave; in signal processing, the error function is used to describe the cumulative distribution of the Gaussian distribution and is used here for dynamic filtering to eliminate noise; the power density is used to describe the intensity attenuation of the radar signal and affects the detection range and accuracy of the signal; the calibration factor is used to correct the measurement error and improve the measurement accuracy. By accurately measuring the distance and angle between the obstacle and the wind turbine component, the UAV can navigate safely and avoid collisions; by integrating over time, the dynamic relative position change between the obstacle and the wind turbine component can be monitored, which is particularly important for hoisting operations;

[0084] The expression of point cloud data in three-dimensional space. The Fourier transform is used to convert the signal from the time domain to the frequency domain for easy analysis and processing; the Hankel function describes the wave motion in cylindrical coordinates and is applicable to the propagation model of radar waves; the wave number and the radial distance describe the propagation characteristics of the wave in space; the angular function describes the directivity of the signal in space; the rotation matrix is used for coordinate transformation to ensure that the point cloud data is in the correct spatial direction; through the Fourier transform and the Hankel function, the position of the spatial point can be accurately obtained to construct an accurate three-dimensional point cloud map; the frequency domain analysis helps to identify and filter out noise and improve the quality of the point cloud data; the angular function and the rotation matrix ensure that the direction of the point cloud data is correct, which is crucial for subsequent data processing and analysis. The theoretical basis of the two formulas is mainly based on the propagation characteristics of radar waves, signal processing techniques, and spatial geometric transformations, which play an active role in accurate measurement and high-quality data generation in the UAV acquisition module and provide a solid foundation for data processing and analysis.

[0085] Preferably, the laser radar of this embodiment constructs a three-dimensional point cloud map of the wind turbine hoisting area in real time to perceive the position of obstacles and the relative distance of wind turbine components. Millimeter-wave radar close-range obstacle detection ensures the safe flight of the drone in a complex environment; high-resolution cameras capture details of the operating scene and provide high-definition images for the docking of key components. Meteorological sensors monitor environmental parameters such as wind speed, wind direction, and humidity in the hoisting area in real time to assist path planning and risk assessment. The inertial measurement unit monitors the flight attitude and dynamic stability of the drone in real time to ensure accurate positioning in strong wind environments. The drone is equipped with high-precision GNSS (Global Navigation Satellite System) and RTK (Real-time Dynamic Differential Positioning) to achieve centimeter-level positioning in low-visibility environments. Equipped with a 5G communication module to ensure real-time data upload to the ground control center, supporting remote control and team collaboration. Before the wind turbine is hoisted, the environment is scanned by using a laser radar to build a three-dimensional model and confirm obstacles on the hoisting path. During the hoisting of the wind turbine under low-visibility conditions, the drone collects image data of the operation site in real time through its high-resolution camera, and these images are transmitted to the intelligent processing and decision-making module in real time through the communication and data management module. The intelligent processing and decision-making module integrates an advanced defogging algorithm based on a deep learning model and image enhancement technology. It can effectively eliminate image blur caused by haze, smoke or insufficient light, and generate high-quality clear images in real time. The processed clear images are transmitted back to the ground operation station in real time for the operator to monitor and assist in decision-making. At the same time, the intelligent processing and decision-making module, combined with the drone's built-in laser radar or ultrasonic distance measurement sensor, can accurately measure the distance between the lifting equipment and the wind turbine components. By fusing image data and distance measurement data, the system can build a real-time three-dimensional model of the work site, providing dual protection for lifting path planning, equipment docking accuracy calibration, and obstacle avoidance. In addition, the integration of clear images after defogging and sensor measurement data enables the system to dynamically adjust the lifting strategy. For example, when wind speed changes or equipment offsets during the lifting process, the system can analyze the risk through real-time images and distance data, issue an early warning to the operator, and recommend or automatically adjust the lifting path, thereby reducing the risk of accidents. This multi-level guarantee mechanism significantly improves the feasibility and safety of wind turbine installation under low visibility conditions. It not only extends the operation window for installation, but also effectively reduces the reliance on manual visual operations, providing reliable technical support for wind turbine construction under severe weather conditions.

[0086] Furthermore, if Figure 3 As shown, the three-dimensional map submodule 13 specifically includes:

[0087] The preprocessing unit 131 is used to remove the noise points generated by high reflective materials, dust or sensors through the reflection intensity threshold; align the point clouds collected at multiple times or at different times to the same coordinate system; and reduce the amount of data through uniform sampling;

[0088] The surface reconstruction unit 132 is used to extract the structural features of the hoisting area, segment the ground and the dynamic mapless type; segment the point cloud plane to obtain the classified point cloud; attach the RGB color information to the point cloud, color the point cloud using the orthophoto, and fuse the color information with the color image obtained by the camera;

[0089] The 3D model generation unit 133 is used to generate a mesh model by point cloud fitting, parameterize the segmented point cloud into a 3D model through calculation; generate a triangular mesh model from the classified point cloud; convert the 3D model from the relative coordinate system to the absolute geographic coordinate system and register it with the remote sensing image; output a 3D map.

[0090] Among them, the preprocessing unit 131 represents the calculation expression of the filtered point cloud data:

[0091]

[0092] In the formula, Z represents the partition function, introduced from statistical mechanics; Li2(x) represents the dilogarithmic function; β represents the inverse temperature parameter; Laplace(Malign,γ) represents the Laplace transform; γ represents the attenuation coefficient; N filter represents the filtered point cloud data; N p represents the total number of points in the point cloud data; i represents the index variable, representing the i-th point in the point cloud data; I i represents the intensity value or feature value of the i-th point cloud data; μ represents the mean of the point cloud intensity values; represents the variance of the i-th point cloud data;

[0093] The surface reconstruction expression of the surface reconstruction unit 132:

[0094]

[0095] In the formula, represents the Green's function, used to describe the potential field distribution; Ricci(N j ) represents the Ricci curvature tensor; Kronecker(C j ,N j ) represents the Kronecker function, used for spatial point correlation analysis; dA represents the area element; S recon represents the surface reconstruction result; domain represents the integration region, representing the physical range of the surface reconstruction; (C j ,N j ) represents the color C j and the normal vector N j ; represents the color gradient; A project represents the area of the projected area; Represents the exponential decay factor based on the projected area;

[0096] The three-dimensional model generation unit 133 three-dimensional geometric model generation expression:

[0097]

[0098] In the formula, Wigner(V k ,T geo ) represents the Wigner function, which is used for quantum phase space description; Trace(T geo ) represents the trace of the matrix; Fourier-Bessel(F transform ,ρ,θ) represents the Fourier-Bessel transform, which is used for spectral analysis in cylindrical coordinates; ρ,θ represent the polar coordinate parameters; M 3D represents the generated three-dimensional geometric model; V k represents the k-th volume unit or geometric unit; T geo represents the geometric transformation matrix; F transform represents the transformation function.

[0099] The above formula: The partition function of the formula of the preprocessing unit 131 is introduced from statistical mechanics and is used to describe the probability distribution of the macroscopic state of the system; the double logarithmic function is used to describe the distribution characteristics of the intensity value; the inverse temperature parameter is used to control the smoothness of the model; the Laplace transform is used to reduce noise and improve the stability of the model; the noise points generated by high-reflection materials, dust or sensors are removed by the reflection intensity threshold; the point clouds collected at multiple times or different times are aligned to the same coordinate system; the data volume is reduced by uniform sampling, and the computational complexity is reduced;

[0100] In the formula of the surface reconstruction unit 132, the Green's function describes the potential field distribution; the Ricci curvature tensor is used to describe the geometric characteristics of the surface; the Kronecker function is used for spatial point correlation analysis; the color gradient represents the change of color; the structural characteristics of the lifting area are extracted, and the ground and the dynamic mapless type are segmented; the point cloud plane is segmented by the point cloud plane to obtain the classified point cloud; the color information fusion attaches the RGB color information to the point cloud, colors the point cloud using the orthophoto image, and fuses the color information through the color image obtained by the camera;

[0101] The Wigner function of the formula of the three-dimensional model generation unit 133 is used for quantum phase space description to describe the distribution characteristics of geometric units; the geometric transformation matrix is used to describe geometric transformations; the Fourier-Bessel transform is used for spectral analysis in cylindrical coordinates; a mesh model is generated by point cloud fitting, and the segmented point cloud is parameterized into a three-dimensional model through calculation; a triangular mesh model is generated from the classified point cloud; the three-dimensional model is converted from a relative coordinate system to an absolute geographic coordinate system, registered with a remote sensing image, and a three-dimensional map is output; it can be seen that the formulas of each sub-module have clear theoretical bases and play important roles in the construction process of the three-dimensional map. The formulas not only improve the efficiency and accuracy of data processing, but also enhance the visualization effect and practicality of the three-dimensional map.

[0102] Preferably, in the preprocessing unit 131 of this embodiment, by setting a reflection intensity threshold, noise points generated by high-reflection materials, dust or sensors are removed to ensure the purity of the point cloud data; the point cloud data collected at different times is aligned to the same coordinate system to eliminate coordinate deviations caused by time differences; the amount of point cloud data is reduced through a uniform sampling algorithm (such as voxel gridization), while key features are retained to improve the efficiency of subsequent processing; the quality of the point cloud data is improved, redundant information is reduced, and the efficiency and accuracy of subsequent processing are enhanced; it is ensured that the point cloud data collected at different times can be accurately stitched together, providing a reliable basis for three-dimensional modeling; through uniform sampling, the relationship between the amount of data and feature retention is balanced, and the use of computing resources is optimized. In the surface reconstruction unit 132, the structural features of the hoisting area are extracted from the point cloud data, and the ground is separated from other areas through a segmentation algorithm; the point cloud is classified according to geometric and spatial features, such as dividing the ground, buildings, vegetation, etc. into different categories; through the orthophoto or the color image obtained by the camera, the RGB color information is attached to the point cloud to enhance the visualization effect of the point cloud; the point cloud is colored using the orthophoto and the color information is fused with the point cloud data to enhance the realism and readability of the point cloud; the key structural features of the hoisting area are extracted, providing accurate geometric information for subsequent three-dimensional modeling; the ground and other areas are segmented to avoid the influence of ground noise on modeling and improve the accuracy of the model; by attaching color information, the visualization effect of the point cloud is enhanced, facilitating users to intuitively understand the scene; after fusing the color information, the point cloud data is more realistic and suitable for fields such as urban planning and environmental monitoring. In the three-dimensional model generation unit 133, a mesh model is generated through a point cloud fitting algorithm (such as RANSAC), and the point cloud data is parameterized into a three-dimensional model; a triangular mesh model is generated based on the classified point cloud data to ensure the geometric accuracy of the model; the three-dimensional model is converted from a relative coordinate system to an absolute geographic coordinate system and registered with a remote sensing image to ensure the spatial consistency of the model; finally, a three-dimensional map is output to support further applications and analyses.

[0103] Furthermore, as Figure 4As shown in the figure, the intelligent processing and decision-making module 2 specifically includes:

[0104] The haze removal processing sub-module 21 is used to confirm the obstacles on the hoisting path through a three-dimensional model, and collect the image data of the operation site in real time through a high-resolution camera, perform haze removal calculation on the real-time data to generate a clear image; transmit the clear image to the ground operation station to construct an obstacle recognition model for recognition;

[0105] The fault parameter combination sub-module 22 is used to form a structured data set based on the real-time operation data of the hoisting equipment, the historical hoisting equipment fault data, and the real-time three-dimensional image; combine the historical hoisting equipment fault data to locate the key data affecting equipment anomalies and determine the parameter combinations related to the faults;

[0106] The fault prediction model construction sub-module 23 is used to construct a hoisting equipment fault prediction model and train it based on historical fault data samples; input the real-time three-dimensional image into the model to predict the fault risk; automatically trigger an early warning according to the prediction result and extract the maintenance plan.

[0107] Preferably, in the haze removal processing sub-module 21 of this embodiment, the image data of the operation site is collected in real time through a high-resolution camera, and the obstacles on the hoisting path are confirmed by using a three-dimensional model; the real-time data is processed by using a haze removal algorithm to generate a clear image. The haze removal algorithm can be based on the atmospheric scattering model or a multi-scale feature fusion method combined with an attention mechanism; the haze removal module can use a multi-layer convolutional network and skip connections to extract and fuse multi-scale features, thereby improving the haze removal effect; improving the image clarity to ensure the accuracy of obstacle recognition; generating clear images in real time and transmitting them to the ground operation station to support the construction and operation of the obstacle recognition model; improving the visual effect of the operation environment and reducing misjudgments or missed judgments caused by haze or insufficient light. In the fault parameter combination sub-module 22, a structured data set is formed based on the real-time operation data of the hoisting equipment, the historical fault data, and the real-time three-dimensional image; combined with the historical fault data, the key data affecting equipment anomalies is located, and the parameter combinations related to the faults are determined; a fuzzy neural network or a multi-layer neural network can be used to analyze and model the data; improving the accuracy and efficiency of fault diagnosis, reducing manual intervention; quickly locating potential fault points through parameter combination analysis, providing a basis for subsequent maintenance; supporting the construction of a real-time monitoring and early warning system to improve the safety and reliability of equipment operation. In the fault prediction model construction sub-module 23, a hoisting equipment fault prediction model is constructed and trained based on historical fault data samples; the real-time three-dimensional image is input into the model to predict the fault risk; an early warning is automatically triggered according to the prediction result and the maintenance plan is extracted; predicting the equipment fault risk in advance, reducing the occurrence probability of sudden faults; the automated early warning mechanism can timely remind the operator to take preventive measures and reduce the equipment downtime; providing a targeted maintenance plan to optimize the equipment maintenance cost and efficiency.

[0108] Further, as Figure 5 shown, the defogging processing sub-module 21 specifically includes:

[0109] A color image unit 211, configured to construct a Retinex theory imaging model based on two parts of an incident component and a reflection component;

[0110] The color image formula is as follows:

[0111] S(x,y) = L(x,y) × R(x,y)

[0112] Where S(x,y) corresponds to the pixel value of an image point, which is obtained by the product of a reflection component R(x,y) and an incident component L(x,y). The dynamic range of pixel values in the image is determined by the incident component, while the reflection component ultimately determines the inherent property of the image color, to a clear defogged image;

[0113] A logarithmizing unit 212, configured to perform a logarithmizing operation on a foggy image, and the calculation formula is as follows:

[0114] logS(x,y) = log(L(x,y) × R(x,y)) = logL(x,y) + logR(x,y)

[0115] Let the logarithmizing operation be simplified to, and the calculation formula is as follows:

[0116] s(x,y) = logS(x,y)

[0117] l(x,y) = logL(x,y)

[0118] r(x,y) = logR(x,y)

[0119] Then the Retinex theory image expression is:

[0120] s(x,y) = l(x,y) + r(x,y)

[0121] Wherein, for the incident light, an approximate fitting estimation is performed, the incident component is approximately fitted by smoothing, and the estimation value l(x,y) is obtained by calculation to improve the accuracy;

[0122] An obstacle recognition unit 213, configured to obtain a reflection component in the logarithmic domain, and the calculation formula is:

[0123] r(x,y) = s(x,y) - l(x,y)

[0124] The obtained r(x,y) is transformed to the spatial domain to obtain the reflection component R(x,y), and finally a defogged image is obtained;

[0125] Build an obstacle recognition model to automatically recognize and dynamically track fan components, hoisting equipment, and obstacles.

[0126] Preferably, in the color image unit 211 of this embodiment, the color image unit decomposes the image into an incident component (L(x,y)) and a reflection component (R(x,y)) based on the Retinex theory; among them, the incident component L(x,y) determines the dynamic range of the image, while the reflection component R(x,y) determines the color and intrinsic attributes of the image; by separating the incident light and the reflection light, the color image unit can remove the influence of illumination on the image, thereby retaining the true color and detail information of the object. This is of great significance for defogging, enhancing image contrast, and restoring image details. The logarithmic unit 212 converts the image from multiplicative operation to additive operation by taking the logarithm to simplify the calculation process; this logarithmic operation changes the complex multiplicative operation into a simple additive operation, expands the dynamic range of the image, and enhances the visibility of dark pixels; the logarithmic unit improves the processability of the image by enhancing the dynamic range of the image and expanding the details in the dark part. In addition, the calculation in the logarithmic domain is more in line with the perception characteristics of the human visual system, which helps subsequent image enhancement and defogging processing. In the obstacle recognition unit 213, the obstacle recognition unit extracts the reflection component r(x,y) in the logarithmic domain and calculates it through the following formula; then converts r(x,y) from the logarithmic domain to the spatial domain to obtain the reflection component R(x,y), and finally generates a defogged image; by separating the reflection component, the obstacle recognition unit can effectively remove the influence of illumination on the image and restore the true color and detail information of the object. This is of great significance for the automatic recognition and dynamic tracking of obstacles such as fan components and hoisting equipment.

[0127] Furthermore, as Figure 6 shown, the obstacle recognition unit 213 specifically includes:

[0128] An acquisition and processing sub-unit 2131, which is used to label the three-dimensional image, extract key features from the sensor data, and convert them into a unified format; extract the target feature representation through an intermediate layer or a global pooling layer; perform temporal modeling on the target movement trajectory and predict the future position;

[0129] An adjustment model sub-unit 2132, which is used to combine the trajectory prediction error and the trajectory safety reward for tracking and path planning; dynamically adjust the learning rate to balance the convergence speed and stability, initialize the experience pool in reinforcement learning to store state-action pairs, and add noise; adjust the obstacle recognition model according to the real-time environmental data;

[0130] A path planning subunit 2133, configured to regenerate an optimal path based on perception data and adjust the actions of the hoisting device through control instructions when it detects that the device deviates or the obstacles change dynamically; and use historical data to test the monitoring accuracy, tracking stability, and path planning rules of the obstacle recognition model.

[0131] Preferably, in the acquisition and processing subunit 2131 of this embodiment, three-dimensional images are acquired through sensor data, and a deep learning model is used to annotate the images to extract key features; the acquired sensor data is converted into a unified format for subsequent processing and analysis; a temporal model is built for the target motion trajectory to predict future positions. This may involve using a deep learning model (such as LSTM) to capture the dynamic changes in time series data; through the feature extraction ability of the deep learning model, the target objects in the three-dimensional images can be more accurately identified and annotated; the conversion to a unified format makes data processing more efficient and reduces the compatibility issues between different data sources; temporal modeling and trajectory prediction can predict the motion trend of the target in advance to support subsequent path planning. The adjustment model subunit 2132 combines the trajectory prediction error and the trajectory safety reward to optimize the tracking and path planning processes; by dynamically adjusting the learning rate, the convergence speed and stability of the model are balanced to ensure that the algorithm can quickly adapt to environmental changes during training; in reinforcement learning, an experience pool is initialized to store state-action pairs and noise is added to enhance the robustness of the model; the obstacle recognition model is adjusted according to real-time environmental data to ensure that the model can adapt to a dynamic environment. By combining the trajectory prediction error and the safety reward, the dynamic changes in a complex environment can be more effectively addressed; the dynamic adjustment of the learning rate and the management of the experience pool enable the model to operate efficiently in a constantly changing environment; the real-time adjustment of the obstacle recognition model improves the detection accuracy and response speed for dynamic obstacles. When the path planning subunit 2133 detects that the device deviates or the obstacles change dynamically, it regenerates an optimal path based on perception data; adjusts the actions of the hoisting device through control instructions to ensure that it executes tasks according to the newly generated path; uses historical data to test the monitoring accuracy, tracking stability, and path planning rules of the obstacle recognition model; generating an optimal path based on real-time perception data can quickly respond to environmental changes and improve the efficiency of path planning; adjusting the device actions through control instructions to ensure that it executes tasks according to the optimal path reduces deviations and errors; testing the monitoring accuracy and path planning rules of the model through historical data can continuously optimize the model performance.

[0132] Further, as Figure 7 shown, the fault parameter combination sub-module 22 specifically includes:

[0133] An integrated data unit 221 is used to centrally store and integrate real-time data, historical fault data, and three-dimensional image data through a cloud platform into a unified structured data set; extract key parameters from the structured data set, consult historical fault records, and screen out parameters highly relevant to known fault times;

[0134] A combined analysis unit 222 is used to analyze the interactions between parameters, identify situations where most faults are triggered by single-parameter or two-parameter combinations, and test the coverage of key parameter pairs; and calculate the contribution degrees of different parameter combinations to faults, and screen out coefficient combinations higher than a preset value;

[0135] A key parameter combination 223 is used to dynamically update parameter combinations in combination with real-time monitoring data, determine parameter combinations related to the solid phase, and convert them into actionable warning indicators.

[0136] Preferably, the integrated data unit 221 in this embodiment integrates real-time data, historical fault data, and three-dimensional image data, and centrally stores and integrates them into a unified structured data set through a cloud platform; it can process real-time data streams to ensure the timeliness and accuracy of data; performs preprocessing operations such as denoising, filtering, and normalization on the original data to improve data quality and provide a reliable basis for subsequent analysis; adopts distributed storage technology to ensure the high availability and scalability of data, while supporting the storage and management of massive data; by integrating data from different sources, a unified data set is formed, improving the utilization rate and analysis value of data; providing a high-quality data basis for subsequent fault analysis and prediction, reducing analysis deviations caused by data quality problems; providing comprehensive data support for fault diagnosis and early warning, and enhancing the scientificity and accuracy of decision-making. In the combined analysis unit 222, by analyzing the interactions between parameters, key parameter combinations triggering faults are identified. This analysis ability relies on advanced data analysis algorithms and models, such as machine learning and statistical modeling; tests the coverage of key parameter pairs to ensure the comprehensiveness and accuracy of analysis results. This requires in-depth mining and verification of data; calculates the contribution degree of different parameter combinations to faults and screens out coefficient combinations higher than the preset value. This calculation is usually based on statistical analysis and model optimization techniques; combines real-time monitoring data to dynamically update parameter combinations to ensure the timeliness and adaptability of analysis results; by analyzing the interactions between parameters, the cause of the fault can be quickly located, improving the accuracy and efficiency of fault diagnosis; by screening out parameter combinations with high contribution degrees, warning indicators are optimized, enhancing the sensitivity and reliability of the warning system; providing a scientific basis for equipment maintenance and optimization, and helping enterprises formulate more effective maintenance strategies and improvement measures. In the key parameter combination 223, the parameter combination is dynamically updated by combining real-time monitoring data to ensure the real-time and accuracy of the analysis result; determines the parameter combination highly related to the fault, and this analysis relies on in-depth mining and pattern recognition of historical data; converts the analysis result into an actionable warning indicator to support real-time monitoring and early warning; by dynamically updating the parameter combination, new fault signals can be captured in real time, enhancing the response speed and accuracy of the warning system; through accurate fault prediction and early warning, measures can be taken in advance to reduce equipment downtime and maintenance costs; by optimizing warning indicators, potential faults can be effectively prevented, enhancing the overall stability and reliability of the system.

[0137] Further, as Figure 8 shown, constructing the fault prediction model sub-module 23 specifically includes:

[0138] The fault prediction model architecture design unit 231 is used to divide the data set into a training set, a validation set, and a test set; train the hoisting equipment fault prediction model using the training set; input the three-dimensional image as height, width, depth, and number of channels; input the time series as time step and feature dimension;

[0139] The fault prediction model architecture construction unit 232 is used to extract spatial features and reduce the dimension by combining with a pooling layer; capture the temporal dependence of time series data, fuse and process images and time series to construct a multi-modal network; use a fully connected layer to combine with a classification task to output a fault probability.

[0140] The fault prediction model prediction unit 233 is used to learn the fault parameter combinations during the training process, preprocess the real-time three-dimensional image, input the preprocessed real-time three-dimensional image into the hoisting equipment fault prediction model for evaluation, and output the fault probability and risk level.

[0141] Preferably, the fault prediction model architecture design unit 231 of this embodiment divides the data set into a training set, a validation set, and a test set to ensure the reasonable distribution of data and the generalization ability of the model; uses the training set to train the hoisting equipment fault prediction model, extracts features through historical data and establishes a model; inputs the height, width, depth, and number of channels of the three-dimensional image, combines with the time step and feature dimension of the time series data to construct a multi-modal input; improves the model's processing ability for complex data, and can better capture multi-dimensional features during equipment operation; ensures the stability and reliability of model training, and avoids the problems of overfitting or underfitting. In the fault prediction model architecture construction unit 232, spatial features are extracted and the dimension is reduced through a pooling layer, reducing the computational amount and retaining important information; the temporal dependence of time series data is captured, and image and time series data are fused through a multi-modal network; a fully connected layer is used to combine with a classification task to output a fault probability, realizing the mapping from features to fault probability. It improves the model's spatial and temporal feature extraction ability, enhances the comprehensive perception of the equipment operation state; through multi-modal fusion, it improves the robustness and accuracy of the model and reduces the limitations of a single modality; the application of the fully connected layer enables the model to efficiently perform classification tasks and output the fault probability and risk level. In the fault prediction model prediction unit 233, the fault parameter combinations are learned during the training process to optimize the parameter settings of the model; the real-time three-dimensional image is preprocessed, including operations such as normalization and denoising, to ensure the quality of the input data; the preprocessed real-time three-dimensional image is input into the hoisting equipment fault prediction model for evaluation, and the fault probability and risk level are output; it realizes the fast response and processing of real-time data, improves the real-time and accuracy of prediction; through the preprocessing of the real-time three-dimensional image, it improves the model's adaptability to complex scenarios and reduces the possibility of misjudgment; the output fault probability and risk level provide a scientific basis for equipment maintenance, helping to take preventive measures in advance and reduce the downtime risk.

[0142] Furthermore, as Figure 9 shown, the hoisting installation module 3 specifically includes:

[0143] The emergency plan formulation submodule 31 is used to compare and combine the results of the fault prediction model with the obstacle recognition model, automatically optimize the lifting trajectory or add temporary reinforcement measures according to the prediction results, deploy spare lifting equipment, emergency power supply and other resources in advance, and formulate targeted emergency plans;

[0144] The early warning submodule 32 is used to compare the real-time data with the predicted value of the fault prediction model. If the deviation value exceeds the preset threshold, a third-level early warning is triggered and pushed to the ground control station;

[0145] The model adjustment submodule 33 is used to mark the abnormal data in the warning time and include it in the fault prediction model training library; if abnormal efficiency is found in multiple monitorings, the environmental factor parameters are modified in a targeted manner; and the optimized fault prediction model is reapplied to the new operation scenario.

[0146] Preferably, the plan-making submodule 21 of this embodiment dynamically adjusts the lifting trajectory or takes temporary reinforcement measures by comparing the results of the fault prediction model with the obstacle recognition model, which embodies the characteristics of multi-model fusion and real-time optimization; according to the prediction results, it deploys spare hoisting equipment, emergency power supplies and other resources in advance, and formulates targeted emergency plans, demonstrating intelligent resource scheduling and emergency response capabilities; automatically optimizes the lifting trajectory or takes measures through algorithms, reduces human intervention, and improves work efficiency and safety; dynamically adjusts the lifting trajectory or takes temporary measures, avoids work interruptions caused by faults or obstacles, and improves overall work efficiency; deploys resources and formulates emergency plans in advance to effectively respond to emergencies, reducing the possibility and losses of accidents; realizes full-process automated management from fault prediction to resource scheduling, reduces human errors, and improves the intelligence level of the system. In the early warning submodule 32, the real-time data is compared with the predicted value of the fault prediction model. When the deviation value exceeds the preset threshold, the third-level early warning is triggered and pushed to the ground control station; the three-level early warning is divided into three levels according to the degree of deviation to ensure that abnormalities of different levels can be responded to in a timely manner; by real-time monitoring of the system status and timely pushing of early warning information, a rapid response to abnormal situations is achieved; through real-time data comparison and threshold triggering mechanism, potential problems can be quickly discovered and early warnings can be issued to avoid accidents; the multi-level early warning mechanism ensures that abnormalities of different severity can be properly handled, improving the flexibility and reliability of the system; through automated early warning push, the need for manual monitoring is reduced, and the intelligence level of the system is improved. In the adjustment model submodule 33, the abnormal data in the early warning time is marked and included in the training library of the fault prediction model, reflecting the data-driven model optimization capability; if abnormal efficiency is found in multiple monitorings, the environmental factor parameters are modified in a targeted manner, showing the ability of dynamic adjustment and optimization; the optimized fault prediction model is reapplied to the new operation scene, reflecting the scalability and adaptability of the model.

[0147] The site investigation and cleaning are carried out in combination with drones. The survey is to ensure that the foundation of the hoisting area is solid and there are no obvious obstacles. The cleaning is used to remove obstacles on the hoisting path, such as trees, rocks or sundries. The equipment transportation and inspection use flatbed trucks and other vehicles to transport the fan components such as tower sections, nacelles, and blades to the site. Check whether the hoisting equipment (such as cranes, slings) is in good condition, and carry out necessary repairs and debugging. Check whether the fan components meet the installation requirements and whether there is any damage. Wind condition monitoring and safety assessment According to the terrain, wind conditions and equipment parameters measured by the drone, formulate a detailed hoisting plan and emergency plan. Tower installation The tower installation is carried out in sections from the bottom to the top in sequence, and during the installation process, it is monitored in real time by the drone. Nacelle installation Transport the nacelle to the hoisting position, and check the stability of components such as generators and gearboxes. Use the main hoist to lift the nacelle and slowly raise it to the top of the tower. With the assistance of the auxiliary hoist, dock the nacelle with the top flange of the tower and complete the bolt fixation. And during the installation process, it is monitored in real time by the drone. Blade installation Use a special blade sling to lift a single blade, rotate it to the appropriate angle, and then dock it with the hub. Use hydraulic or guiding equipment to align the bolt holes and complete the fastening. And during the installation process, it is monitored in real time by the drone (for the specific principle, refer to Appendix Figure 10 ).

[0148] As Figure 11 shown, this embodiment also provides an embodiment of the visual enhancement low visibility fan assisted hoisting method. In this embodiment, the visual enhancement low visibility fan assisted hoisting method is applied to the visual enhancement low visibility fan assisted hoisting system as described in the above embodiment, and specifically includes the following steps:

[0149] Step S1: By using lidar to scan the environment, perceive the position of obstacles and the relative distance of fan components, and construct a three-dimensional point cloud map of the fan hoisting area; collect the image data of the operation site and transmit it to the intelligent processing and decision-making module through wireless communication;

[0150] Step S2: Perform defogging processing on the image data of the operation site, and perform logarithmic operation on the foggy image; construct a hoisting equipment failure prediction model, give early warnings of abnormal conditions in the image data, and identify sudden weather changes affecting hoisting operations through meteorological data analysis;

[0151] Step S3: According to the prediction results of the hoisting equipment failure prediction model, adjust the hoisting plan and formulate an emergency plan; and use the drone for real-time monitoring, compare the real-time monitored image data with the prediction results, and adjust the hoisting equipment failure prediction model based on the comparison results.

[0152] Preferably, in step S1 of this embodiment, a lidar is used to scan the environment, sense the positions of obstacles and the relative distances of the fan components, and generate a three-dimensional point cloud map; the image data of the operation site is collected and transmitted to the intelligent processing and decision-making module through wireless communication; the lidar determines the distance by measuring the time difference and phase difference of the laser signal, and uses Doppler imaging technology to draw a clear 3D image; the environmental perception ability of the lifting area is improved, the obstacles and equipment positions can be accurately identified, and collisions and safety hazards can be avoided; through wireless communication technology, the data is transmitted to the intelligent processing module in real time, providing support for subsequent analysis and decision-making. In step S2, the image data of the operation site is dehazed, and logarithmic operation is used to optimize the image quality; a fault prediction model for the lifting equipment is constructed, and historical data and real-time monitoring data are analyzed; through meteorological data analysis, sudden weather changes affecting the lifting operation are identified; the dehazing process and logarithmic operation improve the clarity and readability of the image, providing a high-quality data basis for subsequent analysis; the fault prediction model can early warn of abnormal situations and reduce the impact of equipment failures on the lifting operation; meteorological data analysis helps to identify potential risks, optimize the lifting operation plan, and ensure safety. In step S3, according to the prediction results of the fault prediction model, the lifting plan is adjusted and an emergency plan is formulated; a drone is used for real-time monitoring, and the monitoring data is compared with the prediction results; based on the comparison results, the fault prediction model of the lifting equipment is adjusted to form a closed-loop optimization; the lifting plan is dynamically adjusted to improve the lifting efficiency and safety; the real-time monitoring by the drone provides first-hand data to ensure the timely and effective dynamic adjustment during the lifting process; through comparative analysis, the fault prediction model is continuously optimized to improve the prediction accuracy and reliability.

[0153] As Figure 12 shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 4 includes a processor 41 and a memory 42 coupled to the processor 41.

[0154] The memory 42 stores program instructions for implementing the visual enhancement low visibility fan-assisted lifting system of any of the above embodiments.

[0155] The processor 41 is configured to execute the program instructions stored in the memory 42 to layout the visual enhancement low visibility fan-assisted lifting system.

[0156] Among them, the processor 41 can also be referred to as a CPU (Central Processing Unit). The processor 41 may be an integrated circuit chip with the ability to process signals. The processor 41 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0157] Furthermore, Figure 13 FIG. is a schematic structural diagram of a storage medium according to an embodiment of the present application. The storage medium 5 of the embodiment of the present application stores program instructions 51 that can implement all the above methods. Among them, the program instructions 51 can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media that can store program codes such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0158] In several embodiments provided by the present invention, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces, and the indirect coupling or communication connection of devices or units can be in electrical, mechanical, or other forms.

[0159] In addition, each functional unit in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. The above is only the embodiment of the present invention, and does not limit the patent scope of the present invention accordingly. Any equivalent structural or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

[0160] The specific embodiments of the invention have been described in detail above, but they are only examples, and the invention is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of the invention. Therefore, all equivalent transformations, modifications, improvements, etc. made without departing from the spirit and principles of the invention should be covered by the scope of the invention.

Claims

1. A visual enhancement low visibility fan-assisted hoisting system, characterized in that, The described vision-enhanced low-visibility fan-assisted hoisting system includes: An environmental acquisition module, which is used to sense the position of obstacles and the relative distance of fan components by using lidar to scan the environment, and construct a three-dimensional point cloud map of the fan hoisting area; collect image data of the operation site and transmit it to the intelligent processing and decision-making module through wireless communication; An intelligent processing and decision-making module, which is used to perform defogging processing on the image data of the operation site and perform logarithmic operation on the foggy image; construct a hoisting equipment failure prediction model, give early warnings of abnormal conditions in the image data, and identify sudden weather changes affecting hoisting operations through meteorological data analysis; A hoisting and installation module, which is used to adjust the hoisting plan and formulate an emergency plan according to the prediction results of the hoisting equipment failure prediction model; and use drones for real-time monitoring, compare the real-time monitored image data with the prediction results, and adjust the hoisting equipment failure prediction model based on the comparison results.

2. The visual enhancement low visibility fan-assisted lifting system according to claim 1, wherein The environmental acquisition module includes: A drone acquisition sub-module, which is used to construct a three-dimensional point cloud map of the fan hoisting area through lidar, sense obstacles and the relative distance of fan components; detect obstacles at close range through millimeter-wave radar; A three-dimensional point cloud map construction sub-module, which is used to obtain the distance and angle information between the obstacle area and the fan components by measuring the laser reflection time; the data of each sampling point includes the distance, the vertical angle with the horizontal, and the angle with the x-axis, and is converted into three-dimensional coordinates through calculation; A three-dimensional map sub-module, which is used to remove noise points generated by high-reflection / absorbing materials, dust or sensor errors, add RGB color information to the point cloud, extract the structural features of the hoisting area, segment different categories of the ground and dynamic objects, convert the processed point cloud data into a three-dimensional model, and output a three-dimensional map.

3. The visual enhancement low visibility fan-assisted lifting system according to claim 2, characterized in that, The three-dimensional map sub-module includes: A preprocessing unit, which is used to remove noise points generated by high-reflection materials, dust or sensors through the reflection intensity threshold; align the point clouds collected at multiple times or different times to the same coordinate system; reduce the data volume through uniform sampling; A surface reconstruction unit, which is used to extract the structural features of the hoisting area, segment the ground and dynamic object types; segment the point cloud plane to obtain classified point clouds; attach RGB color information to the point cloud, color the point cloud with an orthophoto image, and fuse the color information with the color image obtained by the camera; A three-dimensional model generation unit, which is used to generate a mesh model by point cloud fitting, parameterize the segmented point cloud into a three-dimensional model through calculation; generate a triangular mesh model through the classified point cloud; convert the three-dimensional model from the relative coordinate system to the absolute geographic coordinate system and register it with the remote sensing image; output a three-dimensional map.

4. The visual enhancement low visibility fan-assisted hoisting system according to claim 1, characterized in that, The intelligent processing and decision-making module includes: A defogging processing sub-module, which is used to confirm the obstacles on the hoisting path through the three-dimensional model, collect the image data of the operation site in real time through a high-resolution camera, perform defogging calculation on the real-time data to generate a clear image; transmit the clear image to the ground operation station and construct an obstacle recognition model for recognition; The fault parameter combination sub-module is used to form a structural data set based on the real-time operation data of the hoisting equipment, historical hoisting equipment fault data, and real-time three-dimensional images; combine the historical hoisting equipment fault data, locate the key data affecting equipment anomalies, and determine the parameter combinations related to faults. The fault prediction model construction sub-module is used to construct a hoisting equipment fault prediction model and train it based on historical fault data samples; input the real-time three-dimensional image into the model to predict the fault risk; automatically trigger an early warning according to the prediction result, and extract the maintenance plan.

5. The visual enhancement low visibility fan-assisted lifting system according to claim 4, characterized in that, The defogging processing sub-module includes: The color image unit is used to form a Retinex theory imaging model based on two parts, the incident component and the reflection component. The logarithmization unit is used to perform a logarithmization operation on the foggy image. The obstacle recognition unit is used to obtain the reflection component in the logarithmic domain; construct an obstacle recognition model to automatically identify and dynamically track the fan components, hoisting equipment, and obstacles.

6. The visual enhancement low visibility fan-assisted hoisting system according to claim 5, wherein, The obstacle recognition unit includes: The acquisition and processing sub-unit is used to annotate the three-dimensional image, extract key features from the sensor data, and convert them into a unified format; extract the target feature representation through the intermediate layer or global pooling layer; perform temporal modeling on the target motion trajectory to predict the future position. The model adjustment sub-unit is used to combine the trajectory prediction error and the trajectory safety reward for tracking and path planning; dynamically adjust the learning rate to balance the convergence speed and stability, initialize the experience pool in reinforcement learning to store state-action pairs, and add noise; adjust the obstacle recognition model according to the real-time environmental data. The path planning sub-unit is used to regenerate the optimal path based on the perception data and adjust the actions of the hoisting equipment through control instructions when detecting equipment deviation or dynamic changes of obstacles; use historical data to test the monitoring accuracy, tracking stability, and path planning rules of the obstacle recognition model.

7. The visual enhancement low visibility fan-assisted hoisting system according to claim 6, wherein, The fault parameter combination sub-module includes: The data integration unit is used to centrally store and integrate the real-time data, historical fault data, and three-dimensional image data through the cloud platform into a unified structured data set; extract key parameters from the structured data set, consult historical fault records, and screen out the parameters highly relevant to the known fault time. The combination analysis unit is used to analyze the interaction between parameters, identify the cases where most faults are triggered by single-parameter or double-parameter combinations, and test the coverage of key parameter pairs; calculate the contribution degree of different parameter combinations to faults, and screen out the coefficient combinations higher than the preset value. The key parameter combination is used to dynamically update the parameter combination by combining real-time monitoring data, determine the parameter combination related to the solid phase, and convert it into an operable early warning index.

8. The visual enhancement low visibility fan-assisted hoisting system according to claim 4, characterized in that The fault prediction model construction sub-module includes: The fault prediction model architecture design unit is used to divide the data set into a training set, a validation set, and a test set; use the training set to train the hoisting equipment fault prediction model; input the three-dimensional image as the height, width, depth, and number of channels; input the time series as the time step and the feature dimension. A fault prediction model architecture construction unit, which is used to extract spatial features and reduce the dimension by combining with a pooling layer; capture the time dependence of time series data, fuse and process images and time series to construct a multimodal network; use a fully connected layer to combine with a classification task to output a fault probability; A fault prediction model prediction unit, which is used to learn the fault parameter combination during the training process, preprocess the real-time three-dimensional image, input the preprocessed real-time three-dimensional image into the hoisting equipment fault prediction model for evaluation, and output the fault probability and risk level.

9. The visual enhancement low visibility fan-assisted hoisting system according to claim 1, characterized in that, The hoisting installation module includes: A pre-plan formulation sub-module, which is used to combine the results of the fault prediction model and the obstacle recognition model, automatically optimize the hoisting trajectory or increase temporary reinforcement measures according to the prediction results, deploy standby lifting tools and emergency power resources in advance, and formulate a targeted emergency plan; An early warning sub-module, which is used to compare the real-time data with the predicted value of the fault prediction model. If the deviation value exceeds the preset threshold, a level-three early warning is triggered and pushed to the ground control station; An adjusted model sub-module, which is used to mark the abnormal data in the early warning time and incorporate it into the fault prediction model training library; if abnormal efficiency is found through multiple monitors, the environmental factor parameters are corrected accordingly; the optimized fault prediction model is reapplied to the new operation scenario.

10. A visual enhancement method for low visibility fan assisted hoisting, which is applied to the visual enhancement low visibility fan assisted hoisting system according to any one of claims 1 to 9, characterized in that, The visual enhancement low visibility fan-assisted hoisting method includes: Scan the environment using lidar to sense the position of obstacles and the relative distance of fan components, and construct a three-dimensional point cloud map of the fan hoisting area; collect the image data of the operation site and transmit it to the intelligent processing and decision-making module through wireless communication; Perform defogging processing on the image data of the operation site and perform logarithmic operation on the foggy image; construct a hoisting equipment fault prediction model to give an early warning of the abnormal situation of the image data, and identify the sudden weather changes affecting the hoisting operation through meteorological data analysis; Adjust the hoisting plan according to the prediction result of the hoisting equipment fault prediction model and formulate an emergency plan; use a drone for real-time monitoring, compare the real-time monitored image data with the prediction result, and adjust the hoisting equipment fault prediction model based on the comparison result.

Images