An intelligent route planning system for fan UAV inspection

Abstract

The present invention provides an intelligent route planning system for fan drone inspection. The system includes a data acquisition module, a route planning module, a dynamic adjustment module, and an analysis and optimization module. The data acquisition module is used to obtain the fan operation status information and environmental information in real time. The route planning module is used to generate the inspection route of the drone according to the preset task requirements and fan distribution. The dynamic adjustment module is used to dynamically optimize the flight route of the drone during the drone inspection process. The analysis and optimization module is used to optimize and analyze the subsequent task planning route of the drone. Through real-time data acquisition, dynamic path adjustment, and intelligent optimization, the present invention realizes the high efficiency, self-adaptability, and accuracy of the drone inspection route, significantly improving the inspection efficiency and task quality.

Description

Technical Field

[0001] The present invention relates to the technical field of UAV route planning systems, and in particular to an intelligent route planning system for UAV inspection of wind turbines. Background Art

[0002] As a clean energy source, wind power generation has been widely applied globally. The operation and maintenance of wind turbines directly affect power generation efficiency and equipment safety. Among them, the condition monitoring of wind turbine blades is particularly important. Damage to the blades such as cracks, corrosion, or icing problems may lead to serious performance degradation or equipment failures. Therefore, regular inspection of wind turbine blades has become an important part of wind farm operation. Traditional methods for inspecting wind turbines mainly rely on manual labor or ground equipment. These methods have problems such as long time consumption, low efficiency, and high safety hazards. Especially when facing wind farms at high altitudes or with complex terrains, they are difficult to meet the actual needs. The application of UAV technology provides an efficient and safe new way for wind turbine inspection. UAVs have the advantages of flexible flight, diverse perspectives, and convenient operation. They can quickly approach wind turbine blades, collect high-definition images and videos, and achieve efficient monitoring of the wind turbine status.

[0003] Referring to relevant publicly disclosed technical solutions, the technology with the publication number CN115145314B proposes a method for planning the inspection path of wind turbine blades based on UAVs, including: calibrating a template wind turbine, substituting the basic parameters of the template wind turbine into the calculation to obtain the inspection path of the template wind turbine. When the UAV flies to each planned route point, it can adjust its pose to capture a more accurate picture, and then save it as an inspection path template; calibrating the wind turbine to be inspected by the UAV; substituting the calibration result into the inspection path template for calculation to obtain the inspection path of the wind turbine to be inspected. This solution can be quickly applied to wind turbines of the same model at different locations, improving the inspection efficiency of the UAV. However, this solution mainly relies on a preset inspection path template for calculation, and cannot respond in real time to changes in the operating state of the wind turbine or dynamic environmental impacts; and lacks the ability to continuously improve the path planning strategy. Summary of the Invention

[0004] The purpose of the present invention is to propose an intelligent route planning system for UAV inspection of wind turbines to address the current deficiencies.

[0005] The present invention adopts the following technical solutions:

[0006] An intelligent route planning system for fan UAV inspection, characterized in that the system includes a data acquisition module, a route planning module, a dynamic adjustment module, and an analysis and optimization module; the data acquisition module is used to obtain the fan operation status information and environmental information in real time; the route planning module is used to generate the inspection route of the UAV according to the preset task requirements and fan distribution; the dynamic adjustment module is used to dynamically optimize the flight route of the UAV during the UAV inspection process; the analysis and optimization module is used to optimize and analyze the planned route of the UAV's subsequent tasks.

[0007] The data acquisition module includes an environmental perception unit and a fan status acquisition unit; the environmental perception unit is used to obtain the surrounding environmental information of the UAV in real time during the UAV inspection process; the fan status acquisition unit is used to obtain the fan operation status information in real time during the UAV inspection process.

[0008] The dynamic adjustment module includes a route optimization unit and a route recording unit; the route optimization unit is used to dynamically optimize the flight route of the UAV according to environmental changes; the route recording unit is used to record the navigation trajectory during each UAV inspection process.

[0009] Further, the route optimization unit includes a UAV status monitoring subunit, a dynamic perception subunit, and a path adjustment execution subunit; the UAV status monitoring subunit is used to obtain the flight status information of the UAV itself; the dynamic perception subunit is used to dynamically adjust and generate the optimized flight route of the UAV according to the flight status information and environmental information; the path adjustment execution subunit is used to control the UAV to navigate along the optimized flight route after dynamic adjustment.

[0010] Further, the dynamic perception subunit dynamically generates the optimized flight route in the following manner during the UAV inspection process:

[0011] S11: Obtain the flight status information and environmental information during the UAV inspection process;

[0012] S12: Analyze the obstacle distribution in the UAV inspection route according to the environmental information. If there is an obstacle distribution in the current UAV inspection route and the distance between the obstacle and the inspection route is less than the preset safety distance threshold, then generate multiple candidate adjustment paths according to the following conditions:

[0013]

[0014] Among them, is the direction vector of the i-th obstacle avoidance path, P obstacle is the current position of the UAV, P current,i is the position of the i-th obstacle avoidance point; satisfying:

[0015]

[0016] Among them, P current is the position of the obstacle center point, and r is the obstacle avoidance radius, which is set according to a preset safety distance threshold; R(θ i ) is the rotation matrix for rotating the angle θ around the obstacle center position i , and this rotation matrix is used to generate obstacle avoidance paths with different offset angles; is the unit direction vector pointing to the obstacle center;

[0017] After obtaining the direction vectors of multiple obstacle avoidance paths, the paths are extended point by point based on each direction vector to generate a candidate adjustment path point sequence as multiple candidate adjustment paths;

[0018] S13: Evaluate and select the optimal candidate adjustment path and use it as the optimized flight route; the specific implementation process is as follows:

[0019] Path optimal = argmin(C total,i );

[0020] Among them, Path optimal is the selected optimal candidate adjustment path, and argmin() is the minimum value function; C total,i is the comprehensive cost index of the i-th candidate adjustment path;

[0021] C total,i = α1·C distance,i + α2·C deviation,i + α3·C time,i ;

[0022] Among them, C distance,i is the path length cost of the i-th candidate adjustment path, which is obtained according to the total length of this candidate adjustment path; C deviation,i is the data acquisition cost of the i-th candidate adjustment path, which is obtained according to the amount of wind turbine operation status information obtained on this candidate adjustment path; C time,i is the time cost of the i-th candidate adjustment path, which is obtained according to the length of this candidate adjustment path and the UAV flight status information; α1, α2, and α3 are the weight coefficients of the path length cost, data acquisition cost, and time cost, respectively, which are preset by the user according to historical experience;

[0023] S14: Convert the optimized flight route into an executable flight instruction and transmit this flight instruction to the path adjustment execution subunit.

[0024] Further, the analysis and optimization module includes a data storage unit, a model evaluation unit, and an optimized route output unit; the data storage unit is used to store the flight information during each inspection of the UAV, and the flight information includes the navigation trajectory of the UAV, the task completion status, and the UAV status information; the model evaluation unit is used to train and learn through a machine learning model combined with the flight information during each inspection of the UAV, so as to optimize and generate a path planning strategy adapted to different inspection task requirements and fan distributions; the optimized route output unit is used to output a path planning strategy that matches the subsequent task planning according to the output result of the model evaluation unit.

[0025] Further, the model evaluation unit includes a feature extraction subunit, a machine learning model, and an update unit; the feature extraction unit is used to extract key features for model input from the flight information; the machine learning model is used to optimize the path planning strategy through training and learning based on the data information provided by the feature extraction subunit, and generate an optimal route adapted to different task requirements and fan distributions; the update unit is used to, after obtaining a certain amount of flight information each time, organize this part of the flight information into a new training sample and input it into the machine learning model for incremental learning, so as to dynamically update the weights of the machine learning model.

[0026] The beneficial effects achieved by the present invention:

[0027] By combining the real-time collection, dynamic adjustment of UAV flight information, and the optimization of the machine learning model, the present invention constructs a full-process inspection path planning system; during the UAV inspection process, it analyzes the distribution of obstacles in real time, generates multiple candidate paths, and selects the optimal path based on the comprehensive cost evaluation as the optimized flight route, and finally converts it into an executable flight instruction to guide the UAV's navigation during the inspection process; by storing the UAV historical flight information and inputting it into the machine learning model, continuously optimizing the path planning strategy through a deep reinforcement learning architecture, making the path planning ability adaptive and learning-capable, and outputting the optimal path planning scheme according to user needs, not only realizes the efficient execution of inspection tasks, but also improves the intelligent level of UAV inspection path planning, providing reliable technical support for fan inspection in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The present invention can be further understood from the following description in conjunction with the drawings. The components in the drawings are not necessarily drawn to scale, but the emphasis is placed on showing the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0029] Figure 1 It is a schematic diagram of the overall module of the present invention.

[0030] Figure 2Schematic diagram of the working process of the dynamic perception sub-unit of the present invention.

[0031] Figure 3 Schematic diagram of the training and learning process of the machine learning model of the present invention. Detailed implementation manners

[0032] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with its embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention; for those skilled in the art, after referring to the following detailed description, other systems, methods and / or features of this embodiment will become obvious; it is intended that all such additional systems, methods, features and advantages are included in this specification; included within the scope of the present invention and protected by the appended claims; additional features of the disclosed embodiments are described in the following detailed description, and these features will be obvious according to the following detailed description.

[0033] In the drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", etc. indicating the orientation or positional relationship, they are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the drawings are only for illustrative purposes and cannot be understood as a limitation of this patent. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0034] Embodiment 1:

[0035] As Figure 1 shown, this embodiment provides an intelligent route planning system for fan drone inspection. The system includes a data acquisition module, a route planning module, a dynamic adjustment module, and an analysis and optimization module; the data acquisition module is used to obtain the fan operation status information and environmental information in real time; the route planning module is used to generate the inspection route of the drone according to the preset task requirements and fan distribution; the dynamic adjustment module is used to dynamically optimize the flight route of the drone during the drone inspection process; the analysis and optimization module is used to optimize and analyze the subsequent task planning route of the drone.

[0036] The data acquisition module includes an environmental perception unit and a fan status acquisition unit; the environmental perception unit is used to obtain the environmental information around the drone in real time during the drone inspection process; the fan status acquisition unit is used to obtain the fan operation status information in real time during the drone inspection process.

[0037] Specifically, the environmental perception unit obtains environmental information through lidar, high-definition cameras, meteorological sensors, and ultrasonic sensors installed on the UAV. The environmental information includes wind speed, wind direction, and obstacle distribution information. The fan status acquisition unit obtains the fan operation status information through high-definition cameras and infrared thermal imagers installed on the UAV. The fan operation status information includes fan images and fan temperature information.

[0038] The dynamic adjustment module includes a flight route optimization unit and a flight route recording unit. The flight route optimization unit is used to dynamically optimize the flight route of the UAV according to environmental changes. The flight route recording unit is used to record the navigation trajectory during each UAV inspection.

[0039] Further, the flight route optimization unit includes a UAV status monitoring sub-unit, a dynamic perception sub-unit, and a path adjustment execution sub-unit. The UAV status monitoring sub-unit is used to obtain the flight status information of the UAV itself. The dynamic perception sub-unit is used to dynamically adjust and generate an optimized flight route of the UAV according to the flight status information and environmental information. The path adjustment execution sub-unit is used to control the UAV to navigate along the optimized flight route after dynamic adjustment.

[0040] Further, as Figure 2 shown, the dynamic perception sub-unit dynamically generates an optimized flight route in the following way during the UAV inspection:

[0041] S11: Obtain the flight status information and environmental information during the UAV inspection.

[0042] S12: Analyze the obstacle distribution in the UAV inspection route according to the environmental information. If there is an obstacle distribution in the current UAV inspection route and the distance between the obstacle and the inspection route is less than the preset safety distance threshold, then generate multiple candidate adjustment paths according to the following conditions:

[0043]

[0044] where, is the direction vector of the i-th obstacle avoidance path, P obstacle is the current position of the UAV, P current,i is the position of the i-th obstacle avoidance point; satisfying:

[0045]

[0046] where, P current is the position of the obstacle center point, r is the obstacle avoidance radius, set according to the preset safety distance threshold; R(θ i ) is the rotation angle θ iThe rotation matrix is used to generate obstacle avoidance paths with different offset angles; Is the unit direction vector pointing to the center of the obstacle;

[0047] After obtaining the direction vectors of multiple obstacle avoidance paths, the paths are extended point by point based on each direction vector to generate a candidate adjustment path point sequence as multiple candidate adjustment paths;

[0048] S13: Evaluate and select the optimal candidate adjustment path and use it as the optimized flight route; The specific implementation process is as follows:

[0049] Path optimal = argmin(C total,i );

[0050] Among them, Path optimal Is the selected optimal candidate adjustment path, and argmin() is the minimum value function; C total,i Is the comprehensive cost index of the i-th candidate adjustment path;

[0051] C total,i = α1·C distance,i + α2·C deviation,i + α3·C time,i ;

[0052] Among them, C distance,i Is the path length cost of the i-th candidate adjustment path, obtained according to the total length of this candidate adjustment path; C deviation,i Is the data acquisition cost of the i-th candidate adjustment path, obtained according to the amount of fan operation status information obtained on this candidate adjustment path; C time,i Is the time cost of the i-th candidate adjustment path, obtained according to the length of this candidate adjustment path and the UAV flight status information; α1, α2, and α3 are the weight coefficients of the path length cost, data acquisition cost, and time cost, respectively, and are preset by the user according to historical experience;

[0053] S14: Convert the optimized flight route into an executable flight instruction and transmit this flight instruction to the path adjustment execution subunit;

[0054] Furthermore, some functional implementation codes of the dynamic perception subunit are as follows:

[0055]

[0056]

[0057]

[0058] This solution obtains the flight status information and environmental information of the drone, analyzes the obstacle distribution in real time, generates multiple candidate paths, selects the optimal path based on the comprehensive cost evaluation as the optimized flight route, and finally converts it into an executable flight instruction to guide the drone's navigation during the inspection; the evaluation of the comprehensive cost includes the path length cost, data acquisition cost, and time cost, so as to optimize the flight efficiency, task execution quality, and integrity of inspection data acquisition while meeting the obstacle avoidance requirements.

[0059] Embodiment 2:

[0060] This embodiment should be understood as including at least all the features of any one of the foregoing embodiments and being further improved on this basis;

[0061] This embodiment provides an intelligent drone recognition system for wind turbine blade defects. The system includes a data acquisition module, a flight path planning module, a dynamic adjustment module, and an analysis and optimization module; the data acquisition module is used to obtain the operation status information and environmental information of the wind turbine in real time; the flight path planning module is used to generate the inspection flight path of the drone according to the preset task requirements and the distribution of wind turbines; the dynamic adjustment module is used to dynamically optimize the flight path of the drone during the inspection of the drone; the analysis and optimization module is used to optimize and analyze the subsequent task planning route of the drone;

[0062] Furthermore, the analysis and optimization module includes a data storage unit, a model evaluation unit, and an optimized flight path output unit; the data storage unit is used to store the flight information during each inspection of the drone, and the flight information includes the navigation trajectory of the drone, the task completion situation, and the drone status information; the model evaluation unit is used to train and learn through a machine learning model combined with the flight information during each inspection of the drone, so as to optimize and generate a path planning strategy adapted to different inspection task requirements and wind turbine distributions; the optimized flight path output unit is used to output the path planning strategy that matches the subsequent task planning according to the output result of the model evaluation unit.

[0063] Furthermore, the model evaluation unit includes a feature extraction subunit, a machine learning model, and an update unit; the feature extraction unit is used to extract the key features for model input from the flight information; the machine learning model is used to optimize the path planning strategy through training and learning based on the data information provided by the feature extraction subunit, and generate the optimal flight path adapted to different task requirements and wind turbine distributions; the update unit is used to organize this part of the flight information into a new training sample and input it into the machine learning model for incremental learning after obtaining a certain amount of flight information, so as to dynamically update the weights of the machine learning model.

[0064] Furthermore, the key features extracted by the feature extraction unit include task requirement features, flight trajectory features, task completion features, and UAV state features; the task requirement features include inspection task requirements and fan distribution; the flight trajectory features include flight path length and flight trajectory; the task completion features include the acquisition quality and acquisition rate of fan operation state information; the UAV state features include flight energy consumption and obstacle avoidance times;

[0065] Furthermore, as Figure 3 shown, the training and learning process of the machine learning model is as follows:

[0066] S21: Data preparation and feature input: Obtain the key features extracted by the feature extraction unit corresponding to each flight information as training data;

[0067] S22: Establish a machine learning model using the Actor-Critic architecture in deep reinforcement learning; the Actor network in the architecture is used to generate path planning strategies, and the Critic network is used to evaluate the comprehensive value of the current path planning strategy; among them, the evaluation of the current path planning strategy by the Critic network satisfies:

[0068] U task =β1·U distance +β2·U quality +β3·U energy +β4·U disturbance ;

[0069] Among them, U task is the comprehensive value coefficient of the current path planning strategy, U distance is the path value coefficient of the current path planning strategy, which is proportional to the path length of the current path planning strategy; U quality is the task completion value coefficient of the current path planning strategy, which is inversely proportional to the acquisition quality and acquisition rate of the fan operation state information collected by the current path planning strategy; U energy is the energy consumption value coefficient of the current path planning strategy, which is proportional to the flight energy consumption of the current path planning strategy; β1, β2, β3, and β4 are the influence weights of the path value coefficient, task completion value coefficient, energy consumption value coefficient, and disturbance value coefficient respectively, which can be set by the user according to their own needs; U disturbance is the disturbance value coefficient of the current path planning strategy, which satisfies:

[0070] U disturbance =k·exp(ρ·N obstacle );

[0071] Among them, k is the initial disturbance value coefficient, which is set by the user according to the actual task scenario; ρ is the disturbance influence intensity, which is set by the user according to the UAV obstacle avoidance requirements and task requirements; N obstacle is the number of obstacle avoidance times in the current path planning strategy;

[0072] S23: Input the training data obtained in step S21 into the machine learning model for training to update the model parameters; among them, the Critic network optimizes the loss function through the temporal difference error in combination with the comprehensive value coefficient, so as to improve the path evaluation ability; the Actor network optimizes the path generation strategy through the policy gradient method based on the feedback information of the Critic network, so as to gradually generate a path planning strategy with better comprehensive value; complete the model training through multiple rounds of iteration, so that the model realizes the synchronous improvement of the path planning ability and the policy evaluation ability;

[0073] Furthermore, when the user has requirements for the inspection task subsequently, the optimized route output unit inputs the task requirements and fan distribution provided by the user into the feature extraction unit to complete feature extraction and then inputs them into the machine learning model, so as to output the path planning strategy for the current user task requirements, and input the output path planning strategy into the route planning module to make the UAV execute the specific inspection route, realizing efficient inspection operations;

[0074] This solution combines the storage, dynamic learning and path optimization of UAV flight information, integrates historical flight information and real-time task requirements into the machine learning model, and realizes the intelligence and self-adaptability of UAV inspection route planning; in the deep learning model, multi-dimensional factors such as path length, task completion quality, energy consumption and route obstacle avoidance disturbance are integrated to dynamically generate the optimal path planning strategy, thereby effectively improving the execution efficiency and resource utilization rate of the inspection task, and meeting the diverse inspection requirements in complex environments at the same time.

[0075] The content disclosed above is only the preferred feasible embodiment of the present invention, and does not limit the protection scope of the present invention accordingly. Therefore, all equivalent technical changes made by using the content of the specification and drawings of the present invention are included in the protection scope of the present invention. In addition, the elements therein can be updated with the development of technology.

Claims

1. An intelligent route planning system for fan UAV inspection, characterized in that, The system includes a data acquisition module, a route planning module, a dynamic adjustment module, and an analysis and optimization module; the data acquisition module is used to obtain the operating status information and environmental information of the wind turbines in real time; the route planning module is used to generate the inspection route of the UAV according to the preset task requirements and the distribution of the wind turbines; the dynamic adjustment module is used to dynamically optimize the flight route of the UAV during the UAV inspection process; the analysis and optimization module is used to optimize and analyze the planned route of the subsequent tasks of the UAV; the data acquisition module includes an environmental perception unit and a wind turbine status acquisition unit; the environmental perception unit is used to obtain the environmental information around the UAV in real time during the UAV inspection process; the wind turbine status acquisition unit is used to obtain the operating status information of the wind turbines in real time during the UAV inspection process; the dynamic adjustment module includes a route optimization unit and a route recording unit; the route optimization unit is used to dynamically optimize the flight route of the UAV according to the environmental changes; the route recording unit is used to record the navigation trajectory during each UAV inspection process; the route optimization unit includes a UAV status monitoring sub-unit, a dynamic perception sub-unit, and a path adjustment execution sub-unit; the UAV status monitoring sub-unit is used to obtain the flight status information of the UAV itself; the dynamic perception sub-unit is used to dynamically adjust and generate the optimized flight route of the UAV according to the flight status information and the environmental information; the path adjustment execution sub-unit is used to control the UAV to navigate along the optimized flight route after dynamic adjustment; The dynamic perception sub-unit dynamically generates the optimized flight route in the following way during the UAV inspection process: S11: Obtain the flight status information and environmental information during the UAV inspection process; S12: Analyze the distribution of obstacles in the UAV inspection route according to the environmental information. If there is an obstacle distribution in the current UAV inspection route and the distance between the obstacle and the inspection route is less than the preset safety distance threshold, then generate multiple candidate adjustment paths according to the following conditions: ； Among them, is the direction vector of the th obstacle avoidance path, is the current position of the UAV, is the position of the th obstacle avoidance point; satisfying: ； Among them, is the position of the obstacle center point, is the obstacle avoidance radius, which is set according to the preset safety distance threshold; is the rotation angle around the obstacle center position is the rotation matrix, which is used to generate obstacle avoidance paths with different offset angles; is the unit direction vector pointing to the obstacle center; After obtaining the direction vectors of multiple obstacle avoidance paths, extend the paths point by point based on each direction vector to generate a candidate adjustment path point sequence as multiple candidate adjustment paths; S13: Evaluate and select the optimal candidate adjustment path and use it as the optimized flight route; the specific implementation process is as follows: ； Among them, is the optimal candidate adjustment path selected, is the minimization value function; is the comprehensive cost index of the $i$-th candidate adjustment path. ； Among them, is the path length cost of the th candidate adjustment path, which is obtained according to the total length of this candidate adjustment path; is the data acquisition cost of the th candidate adjustment path, which is obtained according to the amount of wind turbine operating status information acquired on this candidate adjustment path; is the time cost of the th candidate adjustment path, which is obtained according to the length of this candidate adjustment path and the UAV flight status information; , and are the weight coefficients of the path length cost, data acquisition cost, and time cost, respectively, which are preset by the user according to historical experience; S14: Convert the optimized flight route into an executable flight instruction and transmit the flight instruction to the path adjustment execution sub-unit.

2. The intelligent route planning system for fan UAV inspection according to claim 1, wherein, The analysis and optimization module includes a data storage unit, a model evaluation unit, and an optimized route output unit; the data storage unit is used to store the flight information during each UAV inspection process, and the flight information includes the navigation trajectory of the UAV, the task completion situation, and the UAV status information; the model evaluation unit is used to train and learn through a machine learning model combined with the flight information during each UAV inspection process, so as to optimize and generate a path planning strategy suitable for different inspection task requirements and wind turbine distributions; The optimized route output unit is used to output the path planning strategy that matches the subsequent task planning according to the output result of the model evaluation unit.

3. The intelligent route planning system for fan UAV inspection according to claim 1, characterized in that, The model evaluation unit includes a feature extraction subunit, a machine learning model, and an update unit; the feature extraction unit is used to extract key features for model input from flight information; The machine learning model is used to optimize the path planning strategy through training and learning based on the data information provided by the feature extraction subunit, and generate an optimal flight route adapted to different task requirements and wind turbine distributions; the update unit is used to, after obtaining a certain amount of flight information each time, organize this part of the flight information into a new training sample and input it into the machine learning model for incremental learning, so as to dynamically update the weights of the machine learning model.

Images