An unmanned aerial vehicle path obstacle avoidance optimization control method and system for a photovoltaic power station

Abstract

The application relates to the technical field of unmanned aerial vehicle intelligent inspection, and specifically discloses an unmanned aerial vehicle path obstacle avoidance optimization control method and system for a photovoltaic power station, a three-dimensional digital twin model of the photovoltaic power station is acquired, an initial inspection path is determined, environment three-dimensional point cloud data is matched with the three-dimensional digital twin model, the degree of environment change is calculated, dynamic obstacles are identified, and the state parameters of the dynamic obstacles are acquired to determine the risk level of the dynamic obstacles; local obstacle avoidance optimization is carried out based on an MPC optimization model, and it is judged in real time whether the local obstacle avoidance optimization result meets the requirements; image data collected after the unmanned aerial vehicle completes an inspection task is acquired, an inspection quality evaluation index is generated, it is judged whether a retake task needs to be carried out according to the inspection quality evaluation index, and the protection level of the inspection process is determined. The application improves the environmental adaptability of the unmanned aerial vehicle inspection and the fault identification accuracy of photovoltaic components, and reduces the flight energy consumption and the frequency of manual intervention.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection technology for unmanned aerial vehicles (UAVs), and more specifically, to a method and system for optimizing the path avoidance and obstacle avoidance control of UAVs for photovoltaic power plants. Background Technology

[0002] As photovoltaic power plant construction expands into complex terrain areas such as mountains and hills, traditional drone inspection methods face numerous challenges. Existing drone inspection methods typically rely on offline path planning based on pre-established map models, generating fixed inspection routes according to the location of photovoltaic modules. However, in actual inspections, the environment surrounding photovoltaic power plants is highly uncertain. Dynamic factors such as temporary construction equipment, moving vehicles, bird activity, and vegetation changes can all alter the actual flight environment of the drone. Because traditional path planning methods primarily optimize for static environments and lack the ability to dynamically perceive and assess real-time environmental changes, significant discrepancies between the actual environment and the pre-set model can easily lead to drones failing to follow the optimal path, or even posing a collision risk.

[0003] Existing drone obstacle avoidance methods typically prioritize collision avoidance. Upon detecting an obstacle, they adjust flight direction or replan the path to avoid it. However, photovoltaic (PV) inspection tasks not only require safe drone flight but also demand that the acquired images meet the requirements for subsequent fault identification. For example, during obstacle avoidance, the drone may deviate from the optimal shooting position for PV modules, resulting in changes in shooting angle, missing module edges, blurred images, or increased light reflection, thereby reducing the accuracy of PV module defect detection. Furthermore, existing dynamic obstacle avoidance technologies typically resume the inspection task by returning to the original flight path or continuing flight from the current position after successfully avoiding an obstacle, lacking an optimization mechanism for the post-obstacle path recovery process. When the drone bypasses an obstacle, it may experience significant path deviation, increasing additional flight distance and energy consumption, and even causing some PV modules to be missed or repeatedly inspected, reducing overall inspection efficiency.

[0004] Therefore, it is necessary to provide a method and system for optimizing the path avoidance of unmanned aerial vehicles (UAVs) for photovoltaic power plants to solve the above-mentioned technical problems. In order to solve the above problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the aforementioned deficiencies in the prior art, this invention provides a method and system for optimizing the path avoidance control of unmanned aerial vehicles (UAVs) for photovoltaic power plants. This addresses the problems existing in the current UAV inspection process for photovoltaic power plants in complex terrain, such as difficulty in responding to environmental changes in a timely manner, inability to classify and handle obstacle risks, impact on inspection quality during obstacle avoidance, and low path recovery efficiency after obstacle avoidance.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for optimizing obstacle avoidance path control for unmanned aerial vehicles (UAVs) used in photovoltaic power plants includes the following steps: A three-dimensional digital twin model of a photovoltaic power station is obtained, and a global path optimization is performed using a multi-objective genetic algorithm to determine the initial inspection path. The system uses lidar to collect 3D point cloud data of the environment during flight, downsamples the 3D point cloud data and extracts the region of interest, and uses an improved NDT algorithm to register the 3D point cloud data of the environment with a 3D digital twin model to calculate the degree of environmental change. Identify dynamic obstacles based on the degree of environmental change, obtain the state parameters of the dynamic obstacles, construct an environmental risk assessment model, and determine the risk level of the dynamic obstacles; By acquiring the flight status parameters of the UAV and the status parameters of dynamic obstacles, and combining the target input of the photovoltaic module's shooting angle, local obstacle avoidance optimization is performed. The MPC controller is used to perform rolling optimization solution, and the local obstacle avoidance optimization results are judged in real time to see if they meet the requirements. After acquiring image data collected by the drone after completing the inspection task, image evaluation parameters are extracted, input into a lightweight convolutional neural network image quality evaluation model, and inspection quality evaluation indicators are generated. Based on the inspection quality evaluation indicators, it is determined whether a reshoot task needs to be carried out. By setting up a three-level safety protection mechanism, the protection level of the inspection process is determined based on the local obstacle avoidance optimization results.

[0007] As a further aspect of the present invention, the calculation of the degree of environmental change involves the following steps: spatially registering the real-time acquired 3D point cloud data of the environment with the 3D digital twin model, calculating the distance error between the registered spatial position of the real-time 3D point cloud data and the corresponding position of the 3D digital twin model, and using this distance as the degree of environmental change.

[0008] As a further aspect of the present invention, dynamic obstacles are identified based on the degree of environmental change, and the state parameters of the dynamic obstacles are obtained. An environmental risk assessment model is then constructed to determine the risk level of the dynamic obstacles. The specific steps are as follows: When the degree of environmental change exceeds the preset threshold, new obstacles are identified. The new obstacles are compared with the obstacles in the 3D point cloud data of the environment during the monitoring period. If a new obstacle is identified that appears repeatedly, it is marked as a dynamic obstacle. Extract the state parameters of dynamic obstacles, including category, spatial location, speed, direction of movement, and size information; Based on the state parameters of dynamic obstacles, the distance to the drone, relative speed, direction of motion, type of dynamic obstacle, and size information of dynamic obstacles are obtained to construct an environmental risk assessment model and calculate the risk level corresponding to each dynamic obstacle. Dynamic obstacles are classified into low-risk, medium-risk, and high-risk obstacles based on their risk level.

[0009] As a further aspect of the present invention, by acquiring the flight state parameters of the UAV and the state parameters of dynamic obstacles, and combining them with the target input of the photovoltaic module's shooting angle, local obstacle avoidance optimization is performed in the MPC optimization model. The MPC controller is used to perform rolling optimization and solve the problem, and the local obstacle avoidance optimization result is judged in real time to see if it meets the requirements. The specific steps are as follows: By acquiring the flight status parameters of the drone and the status parameters of dynamic obstacles, and combining them with the target shooting angle of the photovoltaic module, a comprehensive state vector is constructed. An MPC optimization model is established based on the comprehensive state vector. A multi-objective optimization function is constructed, which includes safety target optimization, quality target optimization, path following target optimization and energy consumption target optimization. The optimal control command for the current control cycle is output, and the weights of each optimization target are dynamically adjusted according to the risk level through an adaptive weight mechanism. The environmental level is updated and control commands are adjusted by rolling optimization. When the MPC optimization model cannot solve the problem in time or an obstacle enters the emergency distance range, the emergency obstacle avoidance mechanism is triggered. Once obstacle avoidance is complete, the path recovery and optimization mechanism is activated to reconnect to the initial inspection path.

[0010] As a further aspect of the present invention, the UAV flight status parameters include the UAV's current three-dimensional spatial position, the UAV's flight attitude angle, the UAV's current speed, and the UAV's direction of motion; The state parameters of dynamic obstacles include the obstacle's current position, obstacle's speed, obstacle's direction of movement, and the distance between the obstacle and the drone.

[0011] As a further aspect of the present invention, a path recovery optimization mechanism is initiated to reconnect to the initial inspection path. Specifically, the initial inspection path, the current location of the UAV, and the uncompleted inspection area are used as constraints to construct a path recovery optimization model. The model comprehensively considers the recovery distance, recovery energy consumption, inspection integrity, and recovery time cost, evaluates several candidate recovery points, and automatically selects the candidate recovery point with the lowest comprehensive cost as the recovery location to reconnect to the initial inspection path.

[0012] As a further aspect of the present invention, the determination of whether a reshoot task needs to be performed is based on the inspection quality evaluation index. Specifically, the inspection quality evaluation index is compared with a preset evaluation index. If the inspection quality evaluation index is less than the preset evaluation index, the image data collected by the UAV after completing the inspection task is unqualified, and a reshoot task needs to be triggered. If the inspection quality evaluation index is greater than or equal to the preset evaluation index, the image data collected by the UAV after completing the inspection task is qualified, and a reshoot task does not need to be triggered.

[0013] As a further aspect of the present invention, the three-level security protection mechanism includes first-level security protection, second-level security protection, and third-level security protection.

[0014] A path avoidance optimization and control system for unmanned aerial vehicles (UAVs) used in photovoltaic power plants includes an inspection path determination module, a three-dimensional model registration module, an environmental risk assessment module, a local obstacle avoidance optimization module, an inspection quality assessment module, and a protection level classification module. The inspection path determination module is used to obtain a three-dimensional digital twin model of the photovoltaic power station, and combine it with a multi-objective genetic algorithm to perform global path optimization and determine the initial inspection path. The 3D model registration module is used to collect 3D point cloud data of the environment during flight using lidar, downsample the 3D point cloud data of the environment and extract the region of interest, and use an improved NDT algorithm to register the 3D point cloud data of the environment with the 3D digital twin model to calculate the degree of environmental change. The environmental risk assessment module is used to identify dynamic obstacles based on the degree of environmental change, obtain the state parameters of the dynamic obstacles, construct an environmental risk assessment model, and determine the risk level of the dynamic obstacles. The local obstacle avoidance optimization module is used to obtain the flight state parameters of the UAV and the state parameters of dynamic obstacles, and combine them with the target input of the photovoltaic module's shooting angle to the MPC optimization model to perform local obstacle avoidance optimization. The MPC controller is used to perform rolling optimization solution, and the local obstacle avoidance optimization results are judged in real time to determine whether they meet the requirements. The inspection quality assessment module is used to acquire image data collected by the UAV after completing the inspection task, extract image assessment parameters, input them into a lightweight convolutional neural network image quality assessment model, generate inspection quality evaluation indicators, and determine whether a reshoot task is needed based on the inspection quality evaluation indicators. The protection level classification module is used to determine the protection level of the inspection process based on the local obstacle avoidance optimization results by setting a three-level safety protection mechanism.

[0015] A readable storage medium storing a computer program, which, when executed by a processor, is used to implement the steps of the above-described method for optimizing the path avoidance control of a drone for a photovoltaic power plant.

[0016] The technical effects and advantages of this invention, a method and system for optimizing UAV path avoidance for photovoltaic power plants, are as follows: This invention establishes a three-dimensional digital twin model of the photovoltaic power plant and uses a multi-objective optimization algorithm to generate a global inspection path, improving the rationality of the initial inspection path while considering inspection efficiency, flight energy consumption, optimal shooting angle, and fixed obstacle avoidance. It utilizes LiDAR to collect real-time three-dimensional point cloud data of the environment and combines it with an improved NDT algorithm to achieve dynamic registration between the real-time point cloud and the digital twin model, enabling timely identification of areas of environmental change and dynamic adjustment of the local path, thus improving the UAV's adaptability to complex environmental changes. Furthermore, by constructing a local MPC that integrates the UAV's flight status, dynamic obstacle status, and the photovoltaic module's shooting target, this invention further enhances the UAV's ability to adapt to complex environmental changes. The obstacle avoidance optimization model integrates safety, inspection quality, path following, and energy consumption targets into a rolling optimization control process. While ensuring flight safety, it maintains the optimal shooting posture as much as possible, achieving synergistic optimization of flight safety and inspection quality. Through inspection quality prediction and active control mechanisms, it can predict the impact of obstacle avoidance on image acquisition quality in advance and proactively adjust flight attitude and control parameters to reduce image blurring, shooting angle deviation, and component omissions caused by obstacle avoidance. Through a post-obstacle avoidance path recovery optimization mechanism, the UAV can automatically calculate the optimal recovery path based on its current position, original inspection path, and remaining inspection tasks, quickly returning to the global inspection route, reducing detour distance and repeated inspections, and improving overall inspection efficiency.

[0017] This invention can significantly improve the environmental adaptability, flight safety, inspection continuity, and photovoltaic module fault identification accuracy of drone inspections of photovoltaic power plants in complex terrain, while reducing flight energy consumption and the frequency of manual intervention. Attached Figure Description

[0018] Figure 1 A flowchart of an unmanned aerial vehicle (UAV) path obstacle avoidance optimization control method for photovoltaic power plants is provided as an embodiment of the present invention; Figure 2 This is a system block diagram of a drone path avoidance optimization control system for photovoltaic power plants, provided as an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of this invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described technical solutions are only a part of this invention, and not all of it. All other technical solutions obtained by those skilled in the art based on the technical solutions of this invention without inventive effort are within the scope of protection of this invention.

[0020] Example 1: As Figure 1 The diagram shown is a flowchart of a drone path obstacle avoidance optimization control method for photovoltaic power plants provided by an embodiment of the present invention. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not limit this. Steps S1 to S6 are detailed as follows: Step S1: Obtain a three-dimensional digital twin model of the photovoltaic power station, and use a multi-objective genetic algorithm to perform global path optimization to determine the initial inspection path; The optimization objectives of the multi-objective genetic algorithm include the inspection path length objective, the UAV flight energy consumption objective, the photovoltaic module shooting angle objective, and the fixed obstacle avoidance objective. A three-dimensional digital twin model of the photovoltaic power station is obtained, and a global inspection environment model is established based on the spatial location of the photovoltaic modules, the distribution of fixed obstacles, and the inspection task requirements. Based on the three-dimensional digital twin model, a multi-objective optimization algorithm is used, with the inspection path length objective, the UAV flight energy consumption objective, the photovoltaic module shooting angle objective, and the fixed obstacle avoidance objective as joint optimization objectives, to generate the initial inspection waypoints and initial flight path of the UAV. The initial flight path is uploaded to the UAV flight control system as the initial inspection path, and the optimal shooting pose corresponding to each inspection waypoint is recorded simultaneously to provide a reference benchmark for subsequent local obstacle avoidance control and inspection quality prediction.

[0021] Step S2: Use lidar to collect 3D point cloud data of the environment during flight, downsample the 3D point cloud data and extract the region of interest, and use the improved normal distribution transform (NDT) algorithm to register the 3D point cloud data of the environment with the 3D digital twin model to calculate the degree of environmental change. Step S3: Identify dynamic obstacles based on the degree of environmental change, obtain the state parameters of the dynamic obstacles, construct an environmental risk assessment model, and determine the risk level of the dynamic obstacles; Step S4: By acquiring the flight status parameters of the UAV and the status parameters of the dynamic obstacles, and combining them with the target input of the photovoltaic module's shooting angle, local obstacle avoidance optimization is performed on the MPC optimization model. The MPC controller is used to perform rolling optimization and solve the problem, and the local obstacle avoidance optimization results are judged in real time to see if they meet the requirements. Local obstacle avoidance optimization includes safety target optimization, quality target optimization, path following target optimization, and energy consumption target optimization. Step S5: Acquire image data collected by the UAV after completing the inspection task, extract image evaluation parameters, input them into a lightweight convolutional neural network image quality evaluation model, generate inspection quality evaluation indicators, and determine whether a reshoot task is needed based on the inspection quality evaluation indicators; the image evaluation parameters include image sharpness, exposure uniformity, photovoltaic module integrity, and fault visibility. Step S6: By setting up a three-level safety protection mechanism, the protection level of the inspection process is determined based on the local obstacle avoidance optimization results; the three-level safety protection mechanism includes level one safety protection, level two safety protection and level three safety protection.

[0022] Preferably, the environmental 3D point cloud data during flight is collected using lidar, downsampled and region of interest is extracted from the environmental 3D point cloud data, and the improved Normal Distribution Transform (NDT) algorithm is used to register the environmental 3D point cloud data with the 3D digital twin model to calculate the degree of environmental change. The specific steps are as follows: The region of interest (ROI) is extracted by downsampling 3D point cloud data of the environment collected during flight using lidar. Based on the current flight status of the UAV and a preset safety range, the ROI is established as follows: ; In the formula: This is the current location of the drone. To determine the location of the 3D point cloud data of the environment within a preset safe range, The radius of environmental perception; The real-time acquired 3D point cloud data of the environment is spatially registered with the 3D digital twin model. The distance error between the registered spatial position of the real-time 3D point cloud data and the corresponding position in the 3D digital twin model is calculated as the degree of environmental change.

[0023] Preferably, dynamic obstacles are identified based on the degree of environmental change, and the state parameters of the dynamic obstacles are obtained. An environmental risk assessment model is then constructed to determine the risk level of the dynamic obstacles. The specific steps are as follows: When the degree of environmental change exceeds the preset threshold, new obstacles are identified. The new obstacles are compared with the obstacles in the 3D point cloud data of the environment during the monitoring period. If a new obstacle is identified that appears repeatedly, it is marked as a dynamic obstacle. Extract the state parameters of dynamic obstacles, including category, spatial location, speed, direction of movement, and size information; Based on the state parameters of dynamic obstacles, the distance to the drone, relative speed, direction of motion, type of dynamic obstacle, and size information of dynamic obstacles are obtained to construct an environmental risk assessment model and calculate the risk level corresponding to each dynamic obstacle. Dynamic obstacles are classified into low-risk, medium-risk, and high-risk obstacles based on their risk level.

[0024] Preferably, by acquiring the flight state parameters of the UAV and the state parameters of the dynamic obstacles, and combining them with the target input of the photovoltaic module's shooting angle, local obstacle avoidance optimization is performed in the MPC optimization model. The MPC controller is used to perform rolling optimization and solve the problem, and the local obstacle avoidance optimization results are judged in real time to see if they meet the requirements. The specific steps are as follows: By acquiring the flight status parameters of the drone and the status parameters of dynamic obstacles, and combining them with the target shooting angle of the photovoltaic module, a comprehensive state vector is constructed. An MPC optimization model is established based on the comprehensive state vector. A multi-objective optimization function is constructed, which includes safety target optimization, quality target optimization, path following target optimization and energy consumption target optimization. The optimal control command for the current control cycle is output, and the weights of each optimization target are dynamically adjusted according to the risk level through an adaptive weight mechanism. The environmental level is updated and control commands are adjusted by rolling optimization. When the MPC optimization model cannot solve the problem in time or an obstacle enters the emergency distance range, the emergency obstacle avoidance mechanism is triggered. Once obstacle avoidance is complete, a path recovery optimization mechanism is activated. Using the initial inspection path, the current drone position, and the uninspected area as constraints, a path recovery optimization model is constructed. Taking into account recovery distance, recovery energy consumption, inspection completeness, and recovery time cost, several candidate recovery points are evaluated, and the candidate recovery point with the lowest overall cost is automatically selected as the recovery position to reconnect to the initial inspection path. This allows the drone to quickly resume normal inspection tasks and avoid long-distance detours and inspection omissions caused by local obstacle avoidance.

[0025] The drone's flight status parameters include the drone's current three-dimensional spatial position, the drone's flight attitude angle, the drone's current speed, and the drone's direction of motion.

[0026] The state parameters of dynamic obstacles include the obstacle's current position, obstacle's speed, obstacle's direction of movement, and the distance between the obstacle and the drone.

[0027] In this embodiment of the invention, during the inspection process of the UAV following the initial inspection path, if the risk level of a dynamic obstacle is detected to reach the local obstacle avoidance trigger condition, the local obstacle avoidance control module is activated. The flight control system acquires the UAV's flight status parameters in real time, including the UAV's current three-dimensional spatial position, flight attitude angle, flight speed, and direction of motion. Simultaneously, the environmental perception module acquires the current position, speed, direction of motion of the dynamic obstacle, and the distance between the obstacle and the UAV. The optimal shooting angle target for the photovoltaic module corresponding to the current inspection target is obtained from the global path planning module. Subsequently, the UAV's flight status parameters, the dynamic obstacle's status parameters, and the photovoltaic module's shooting angle target are fused to construct a comprehensive state vector, which is then input into the MPC optimization model.

[0028] The MPC optimization model uses the UAV's kinematics model as the prediction model and constructs a multi-objective optimization function in the prediction time domain, including safety objective optimization, quality objective optimization, path-following objective optimization, and energy consumption objective optimization. Specifically, safety objective optimization ensures the UAV maintains a safe distance from dynamic obstacles; quality objective optimization constrains the UAV to maintain the optimal shooting angle of the photovoltaic modules during obstacle avoidance, reducing image quality degradation; path-following objective optimization constrains the local obstacle avoidance trajectory to be as close as possible to the initial inspection path, reducing detour distance; and energy consumption objective optimization limits the UAV's speed and attitude changes, reducing flight energy consumption. Simultaneously, based on the risk level output by the environmental risk assessment module, an adaptive weighting mechanism dynamically adjusts the weights of each optimization objective. When the risk level is high, the weight of the safety objective is increased; when the risk level is low, the weights of the quality and path-following objectives are increased, enabling the UAV to dynamically adjust its control strategy according to different environmental conditions.

[0029] The MPC controller employs a rolling optimization approach, predicting the UAV's motion state over multiple future control cycles, solving for the control sequence that minimizes the comprehensive cost function, and executing only the optimal control commands corresponding to the current control cycle, including speed, attitude, and heading adjustments. After executing the control commands, the UAV re-acquires flight state parameters and dynamic obstacle state parameters, updates the comprehensive state vector, and re-inputs it into the MPC optimization model to achieve real-time closed-loop optimization of local obstacle avoidance control. Simultaneously, it continuously assesses whether the local obstacle avoidance optimization results meet safety distance constraints, shooting angle constraints, path deviation constraints, and energy consumption constraints. When the MPC optimization model cannot complete the solution within a preset time, or when a dynamic obstacle enters the emergency safety distance range, an emergency obstacle avoidance mechanism is immediately triggered, executing emergency hovering or rapid lateral obstacle avoidance control to ensure UAV flight safety.

[0030] Once a dynamic obstacle is completely avoided, the path recovery optimization mechanism is automatically activated. Using the initial inspection path, the current drone position, and the area that has not yet been inspected as constraints, multiple candidate recovery points are generated. The recovery distance, recovery energy consumption, recovery time, and inspection integrity cost corresponding to each candidate recovery point are calculated. The candidate recovery points are comprehensively evaluated, and the candidate recovery point with the lowest comprehensive cost is automatically selected as the path recovery position. The drone then reconnects to the initial inspection path and continues to perform the inspection task. This reduces the extra flight distance and energy consumption caused by local obstacle avoidance, avoids duplicate inspections or missed inspections, and improves the continuity and overall efficiency of drone inspection tasks for photovoltaic power stations in complex terrain.

[0031] Preferably, the image data collected by the UAV after completing the inspection task is acquired, image evaluation parameters are extracted, and the data are input into a lightweight convolutional neural network image quality evaluation model to generate inspection quality evaluation indicators. Based on the inspection quality evaluation indicators, it is determined whether a reshoot task needs to be performed. The specific steps are as follows: After acquiring image data collected by the drone after completing its inspection mission, image evaluation parameters are extracted and input into a lightweight convolutional neural network image quality evaluation model to generate inspection quality evaluation indicators. The calculation formula is as follows: ; In the formula: As an indicator for evaluating the quality of inspections. For image clarity, For exposure uniformity, For the integrity of photovoltaic modules, For fault visibility, This represents the weighting coefficient for image sharpness. This is a weighting factor for exposure uniformity. This is a weighting coefficient for the integrity of photovoltaic modules. This is the weighting coefficient for fault visibility; The system determines whether a reshoot task is needed based on the inspection quality evaluation indicators. Specifically, the inspection quality evaluation indicators are compared with preset evaluation indicators. If the inspection quality evaluation indicators are less than the preset evaluation indicators, the image data collected by the UAV after completing the inspection task is unqualified, and a reshoot task needs to be triggered. If the inspection quality evaluation indicators are greater than or equal to the preset evaluation indicators, the image data collected by the UAV after completing the inspection task is qualified, and a reshoot task does not need to be triggered.

[0032] Preferably, a three-level safety protection mechanism is set up to determine the protection level of the inspection process based on the local obstacle avoidance optimization results. The specific steps are as follows: When the model predictive control solution times out or an obstacle is detected entering the emergency safety distance, the first-level safety protection is immediately activated, and the emergency hovering or rapid lateral obstacle avoidance is completed through preset actions.

[0033] When an abnormal flight condition such as insufficient battery power, abnormal positioning, or sensor failure is detected, the inspection mission is automatically terminated, a level 2 safety protection is activated, and the drone returns to the take-off and landing point along a safe path.

[0034] When ground monitoring personnel detect special circumstances, they can take over the flight control of the drone in real time through the communication link, activate the three-level safety protection, and perform hovering, emergency landing or manual operation on the drone to achieve a safety guarantee that combines autonomous control and manual control.

[0035] Example 2: In this embodiment of the invention, a 100MW photovoltaic power station in a mountainous area is used as an example. The existing photovoltaic power station is built in a hilly area with a site area of ​​approximately 2.8 square kilometers. The photovoltaic modules are arranged in layers along the mountainside, with significant terrain undulations and an elevation change of approximately 180m. In addition to the fixed photovoltaic modules, combiner boxes, inverters, and transmission lines, the inspection area also contains inspection vehicles, maintenance personnel, construction machinery, and dynamic obstacles such as birds. To improve the efficiency and flight safety of UAV inspections in complex terrain environments, this embodiment uses the aforementioned UAV path avoidance optimization control method for photovoltaic power stations in complex terrain to complete the entire inspection task.

[0036] Before the inspection begins, the backend management platform retrieves a 3D digital twin model of the photovoltaic power station. This model includes information such as the spatial location of photovoltaic modules, terrain elevation data, the location of fixed obstacles, road distribution, and no-fly zones for drones. Then, an inspection task is generated based on the module areas to be inspected that day, and the locations of the modules to be inspected are imported into the path planning module as the set of inspection targets. The path planning module combines the drone's current remaining battery power, maximum flight time, flight altitude limits, and the spatial relationships between the inspection targets, employing a multi-objective genetic algorithm for global path optimization. The joint optimization objectives are the shortest inspection path length, lowest flight energy consumption, minimum deviation of the optimal shooting angle for photovoltaic modules, and maximum avoidance distance for fixed obstacles. The inspection path is iteratively calculated continuously, ultimately outputting an initial inspection path that meets the multi-objective optimization requirements. Simultaneously, a corresponding sequence of waypoints is generated, and the waypoints and flight parameters are sent to the drone's flight control system. The drone then begins executing the inspection task according to the generated initial inspection path.

[0037] During the drone's flight, the onboard LiDAR continuously scans the inspection area, collecting real-time 3D point cloud data of the environment in front of and around the drone, and uploading it to the environmental perception module at a set frequency. The environmental perception module first performs voxel downsampling on the raw point cloud data to reduce the number of points and lower the computational complexity. Then, based on the drone's current position, it establishes a region of interest, retaining only the point cloud data within the drone's flight direction and safe distance range, removing irrelevant point cloud data and noise data from distant locations to improve environmental analysis efficiency. Next, an improved NDT algorithm is used to spatially register the processed real-time environmental point cloud with a pre-established 3D digital twin model. By calculating the matching error between the real-time point cloud and the digital twin model, the degree of environmental change is obtained. When the matching error is within the normal range, it indicates that the current environment is basically consistent with the digital twin model, and no adjustment to the flight path is needed. When the matching error exceeds a preset threshold, it is determined that there are new environmental changes in the current area.

[0038] Further analysis of the changed areas identifies the sources of environmental change. If identified as dynamic targets such as temporarily parked vehicles, inspection personnel, construction equipment, or birds, the location coordinates, speed, direction of movement, and size information of the corresponding dynamic obstacles are extracted. An environmental risk assessment model is then established based on the distance relationship between the obstacle and the UAV. This model comprehensively considers factors such as obstacle distance, movement trend, size, and category to calculate a risk assessment value, classifying dynamic obstacles into low-risk, medium-risk, and high-risk levels based on the risk assessment value. For low-risk obstacles, only the local environmental model is updated, without changing the original flight path; for medium-risk obstacles, local path optimization is initiated; and for high-risk obstacles, a high-priority obstacle avoidance control mode is immediately entered, and a risk warning is sent to the flight control system.

[0039] Upon entering the local obstacle avoidance control phase, the control module acquires real-time flight status parameters of the UAV, including its current position, flight attitude, flight speed, and heading angle. Simultaneously, it acquires dynamic obstacle status parameters, including obstacle position, speed, direction of motion, and predicted trajectory. Combined with the optimal shooting angle target for the photovoltaic modules determined in the global path planning phase, a comprehensive state vector is constructed and input into the MPC optimization model. The MPC optimization model is based on the UAV's kinematic model, establishing a prediction time domain. Within this domain, it simultaneously constructs optimizations for safety, quality, path following, and energy consumption. The safety target ensures the UAV maintains a safe distance from dynamic obstacles; the quality target ensures the UAV maintains the optimal shooting angle during obstacle avoidance, allowing the photovoltaic modules to fully enter the camera's field of view; the path following target constrains the local obstacle avoidance trajectory to closely approximate the global inspection path, reducing detour distances; and the energy consumption target limits the amplitude of changes in UAV control variables, avoiding additional energy consumption caused by frequent acceleration / deceleration and large-angle attitude adjustments.

[0040] The MPC controller predicts the UAV's motion state over multiple future control cycles based on the integrated state vector and uses a rolling optimization method to find the control sequence that minimizes the integrated cost function in the current prediction time domain. Only the first set of control commands for the current control cycle is executed, including UAV speed, attitude, and heading adjustments. Simultaneously, the latest flight status and environmental information are reacquired before entering the next control cycle to continue optimization, achieving real-time closed-loop local obstacle avoidance control. After completing the current cycle optimization, the controller checks in real-time whether the local obstacle avoidance optimization result meets safety distance constraints, path deviation constraints, camera angle constraints, and energy consumption constraints. If all constraints are met, the local obstacle avoidance strategy continues to be executed. If optimization fails or obstacles enter the emergency safety distance, the normal MPC control process is immediately exited, and an emergency safety protection mechanism is activated.

[0041] After the UAV completes local obstacle avoidance and continues its inspection mission, the onboard visible light camera continuously acquires inspection images of the photovoltaic modules and transmits these images to the image quality assessment module. The image quality assessment module first extracts image evaluation parameters, including features such as image sharpness, exposure level, illumination uniformity, module integrity, and image noise. These extracted parameters are then input into a lightweight convolutional neural network image quality assessment model. The lightweight convolutional neural network scores the inspection images and outputs corresponding inspection quality evaluation indicators. When the inspection quality evaluation indicator is higher than a preset quality threshold, the current inspection image is considered to meet the requirements for subsequent fault identification. When the inspection quality evaluation indicator is lower than the preset quality threshold, the corresponding photovoltaic module location, the cause of the image defect, and the inspection time are recorded. A re-shooting task is automatically generated and added to the next round of inspection plans to ensure the integrity and reliability of the inspection data.

[0042] Meanwhile, throughout the inspection process, the drone's operational status, local obstacle avoidance optimization results, and environmental risk level are continuously monitored, and the protection level of the inspection process is dynamically determined based on the inspection status. When the local obstacle avoidance optimization results are normal and the flight status is stable, the first-level protection mode is maintained, and the drone autonomously completes the inspection task. When an increase in environmental risk is detected, local obstacle avoidance is frequently triggered, or the MPC optimization result approaches the safety boundary, the system automatically switches to the second-level protection mode, raising the flight safety threshold and limiting the drone's flight speed and attitude change range. When situations such as MPC solution failure, obstacles entering the emergency safety distance, insufficient drone power, abnormal positioning, or communication interruption occur, the system immediately enters the third-level protection mode, executing emergency control strategies such as emergency hovering, safe return, or manual takeover to ensure the safety of the drone and inspection equipment.

[0043] Through the above implementation methods, this embodiment realizes a complete technical closed loop from global path planning, dynamic environmental perception, risk level assessment, local intelligent obstacle avoidance, image quality evaluation to safety protection and control. This enables the UAV to adjust its flight strategy in real time according to complex terrain and dynamic obstacle changes, ensuring flight safety while maintaining the best inspection quality of photovoltaic modules. It also effectively reduces the inspection omission rate, duplicate inspection rate, and flight energy consumption, improving the stability, reliability, and intelligence level of autonomous inspection of photovoltaic power stations in complex terrain.

[0044] Example 3: A drone path obstacle avoidance optimization control system for photovoltaic power plants, comprising an inspection path determination module, a 3D model registration module, an environmental risk assessment module, a local obstacle avoidance optimization module, an inspection quality assessment module, and a protection level classification module; the inspection path determination module is connected to the 3D model registration module, the 3D model registration module is connected to the environmental risk assessment module, the environmental risk assessment module is connected to the local obstacle avoidance optimization module, the local obstacle avoidance optimization module is connected to the inspection quality assessment module, and the inspection quality assessment module is connected to the protection level classification module.

[0045] The inspection path determination module is used to obtain a three-dimensional digital twin model of the photovoltaic power station, and combine it with a multi-objective genetic algorithm to perform global path optimization and determine the initial inspection path.

[0046] The 3D model registration module is used to collect 3D point cloud data of the environment during flight using lidar, downsample the 3D point cloud data and extract the region of interest, and use an improved NDT algorithm to register the 3D point cloud data of the environment with the 3D digital twin model to calculate the degree of environmental change.

[0047] The environmental risk assessment module is used to identify dynamic obstacles based on the degree of environmental change, obtain the state parameters of dynamic obstacles, construct an environmental risk assessment model, and determine the risk level of dynamic obstacles.

[0048] The local obstacle avoidance optimization module is used to obtain the flight state parameters of the UAV and the state parameters of dynamic obstacles, and combine them with the target input of the photovoltaic module's shooting angle into the MPC optimization model to perform local obstacle avoidance optimization. The MPC controller is used to perform rolling optimization solution, and the module judges in real time whether the local obstacle avoidance optimization result meets the requirements.

[0049] The inspection quality assessment module is used to acquire image data collected by the UAV after completing the inspection task, extract image assessment parameters, input them into a lightweight convolutional neural network image quality assessment model, generate inspection quality evaluation indicators, and determine whether a reshoot task is needed based on the inspection quality evaluation indicators.

[0050] The protection level classification module is used to determine the protection level of the inspection process based on the local obstacle avoidance optimization results by setting a three-level safety protection mechanism.

[0051] like Figure 2 The diagram shown is a system block diagram of a drone path avoidance optimization control system for a photovoltaic power station according to an embodiment of the present invention, which can be used to execute... Figure 1 The steps in the method embodiments shown are implemented in a similar manner and have similar technical effects, and will not be repeated here.

[0052] A readable storage medium storing a computer program, which, when executed by a processor, is used to implement the steps of a drone path avoidance optimization control method for photovoltaic power plants as described in any of the above claims.

[0053] The readable storage medium can be a computer storage medium or a communication medium. A communication medium includes any medium that facilitates the transfer of computer programs from one location to another. A computer storage medium can be any available medium accessible to a general-purpose or special-purpose computer. For example, a readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application-Specific Integrated Circuit (ASIC). Alternatively, the ASIC can be located in a user equipment. Of course, the processor and the readable storage medium can also exist as discrete components in a communication device. The readable storage medium can be a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0054] The present invention also provides a program product including executable instructions stored in a readable storage medium. At least one processor of the device can read the executable instructions from the readable storage medium, and the at least one processor executes the executable instructions to cause the device to implement the methods provided in the various embodiments described above.

[0055] In the embodiments of the above-described device, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0056] Through the above embodiments, this invention establishes a three-dimensional digital twin model of a photovoltaic power station and uses a multi-objective optimization algorithm to generate a global inspection path. This improves the rationality of the initial inspection path while considering inspection efficiency, flight energy consumption, optimal shooting angle, and fixed obstacle avoidance. It utilizes LiDAR to collect real-time three-dimensional point cloud data of the environment and combines it with an improved NDT algorithm to achieve dynamic registration between the real-time point cloud and the digital twin model. This enables timely identification of areas of environmental change and dynamic adjustment of local paths, improving the UAV's adaptability to complex environmental changes. Furthermore, by constructing an MPC local obstacle avoidance optimization model that integrates the UAV's flight status, dynamic obstacle status, and the photovoltaic module's shooting target, it addresses safety targets, Inspection quality targets, path following targets, and energy consumption targets are all integrated into the rolling optimization control process. While ensuring flight safety, the optimal shooting posture is maintained as much as possible, achieving synergistic optimization of flight safety and inspection quality. Through inspection quality prediction and active control mechanisms, the impact of obstacle avoidance on image acquisition quality can be predicted in advance, and flight attitude and control parameters can be actively adjusted to reduce image blurring, shooting angle deviation, and component omissions caused by obstacle avoidance. Through the path recovery optimization mechanism after obstacle avoidance, the UAV can automatically calculate the optimal recovery path based on its current position, original inspection path, and remaining inspection tasks, quickly returning to the global inspection route, reducing detour distance and repeated inspections, and improving overall inspection efficiency.

[0057] This invention can significantly improve the environmental adaptability, flight safety, inspection continuity, and photovoltaic module fault identification accuracy of drone inspections of photovoltaic power plants in complex terrain, while reducing flight energy consumption and the frequency of manual intervention.

[0058] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

[0059] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing the path avoidance and obstacle avoidance control of unmanned aerial vehicles (UAVs) for photovoltaic power plants, characterized in that, Includes the following steps: A three-dimensional digital twin model of a photovoltaic power station is obtained, and a global path optimization is performed using a multi-objective genetic algorithm to determine the initial inspection path. The system uses lidar to collect 3D point cloud data of the environment during flight, downsamples the 3D point cloud data and extracts the region of interest, and uses an improved NDT algorithm to register the 3D point cloud data of the environment with a 3D digital twin model to calculate the degree of environmental change. Identify dynamic obstacles based on the degree of environmental change, obtain the state parameters of the dynamic obstacles, construct an environmental risk assessment model, and determine the risk level of the dynamic obstacles; By acquiring the flight status parameters of the UAV and the status parameters of dynamic obstacles, and combining the target input of the photovoltaic module's shooting angle, local obstacle avoidance optimization is performed. The MPC controller is used to perform rolling optimization solution, and the local obstacle avoidance optimization results are judged in real time to see if they meet the requirements. After acquiring image data collected by the drone after completing the inspection task, image evaluation parameters are extracted, input into a lightweight convolutional neural network image quality evaluation model, and inspection quality evaluation indicators are generated. Based on the inspection quality evaluation indicators, it is determined whether a reshoot task needs to be carried out. By setting up a three-level safety protection mechanism, the protection level of the inspection process is determined based on the local obstacle avoidance optimization results.

2. The method for optimizing and controlling the path avoidance of unmanned aerial vehicles (UAVs) in a photovoltaic power plant according to claim 1, characterized in that, The specific steps for calculating the degree of environmental change are as follows: Spatially register the real-time acquired 3D point cloud data of the environment with the 3D digital twin model, calculate the spatial position of the real-time 3D point cloud data after registration, and the distance error between the corresponding position of the 3D digital twin model, which is used as the degree of environmental change.

3. The method for optimizing and controlling the path avoidance of unmanned aerial vehicles (UAVs) in a photovoltaic power station according to claim 2, characterized in that, Identify dynamic obstacles based on the degree of environmental change, obtain the state parameters of the dynamic obstacles, construct an environmental risk assessment model, and determine the risk level of the dynamic obstacles. The specific steps are as follows: When the degree of environmental change exceeds the preset threshold, new obstacles are identified. The new obstacles are compared with the obstacles in the 3D point cloud data of the environment during the monitoring period. If a new obstacle is identified that appears repeatedly, it is marked as a dynamic obstacle. Extract the state parameters of dynamic obstacles, including category, spatial location, speed, direction of movement, and size information; Based on the state parameters of dynamic obstacles, the distance to the drone, relative speed, direction of motion, type of dynamic obstacle, and size information of dynamic obstacles are obtained to construct an environmental risk assessment model and calculate the risk level corresponding to each dynamic obstacle. Dynamic obstacles are classified into low-risk, medium-risk, and high-risk obstacles based on their risk level.

4. The method for optimizing and controlling the path avoidance of unmanned aerial vehicles (UAVs) in a photovoltaic power station according to claim 1, characterized in that, By acquiring the flight state parameters of the UAV and the state parameters of dynamic obstacles, and combining them with the target input from the photovoltaic module's shooting angle, local obstacle avoidance optimization is performed in the MPC optimization model. The MPC controller is used for rolling optimization solution, and the local obstacle avoidance optimization results are judged in real time to determine whether they meet the requirements. The specific steps are as follows: By acquiring the flight status parameters of the drone and the status parameters of dynamic obstacles, and combining them with the target shooting angle of the photovoltaic module, a comprehensive state vector is constructed. An MPC optimization model is established based on the comprehensive state vector. A multi-objective optimization function is constructed, which includes safety target optimization, quality target optimization, path following target optimization and energy consumption target optimization. The optimal control command for the current control cycle is output, and the weights of each optimization target are dynamically adjusted according to the risk level through an adaptive weight mechanism. The environmental level is updated and control commands are adjusted by rolling optimization. When the MPC optimization model cannot solve the problem in time or an obstacle enters the emergency distance range, the emergency obstacle avoidance mechanism is triggered. Once obstacle avoidance is complete, the path recovery and optimization mechanism is activated to reconnect to the initial inspection path.

5. The method for optimizing the path avoidance control of unmanned aerial vehicles (UAVs) for photovoltaic power plants according to claim 4, characterized in that, The drone flight status parameters include the drone's current three-dimensional spatial position, drone flight attitude angle, drone current speed, and drone direction of motion; The state parameters of dynamic obstacles include the obstacle's current position, obstacle's speed, obstacle's direction of movement, and the distance between the obstacle and the drone.

6. The method for optimizing the path avoidance control of unmanned aerial vehicles (UAVs) for photovoltaic power plants according to claim 4, characterized in that, The path recovery optimization mechanism is activated to reconnect to the initial inspection path. Specifically, the initial inspection path, the current location of the UAV, and the uncompleted inspection area are used as constraints to construct a path recovery optimization model. The model comprehensively considers the recovery distance, recovery energy consumption, inspection completeness, and recovery time cost, evaluates several candidate recovery points, and automatically selects the candidate recovery point with the lowest comprehensive cost as the recovery location to reconnect to the initial inspection path.

7. The method for optimizing the path avoidance control of unmanned aerial vehicles (UAVs) for photovoltaic power plants according to claim 1, characterized in that, The system determines whether a reshoot task is needed based on the inspection quality evaluation indicators. Specifically, the inspection quality evaluation indicators are compared with preset evaluation indicators. If the inspection quality evaluation indicators are less than the preset evaluation indicators, the image data collected by the UAV after completing the inspection task is unqualified, and a reshoot task needs to be triggered. If the inspection quality evaluation indicators are greater than or equal to the preset evaluation indicators, the image data collected by the UAV after completing the inspection task is qualified, and a reshoot task does not need to be triggered.

8. The method for optimizing the path avoidance control of unmanned aerial vehicles (UAVs) for photovoltaic power plants according to claim 1, characterized in that, The three-tiered security protection mechanism includes Level 1 security protection, Level 2 security protection, and Level 3 security protection.

9. A drone path obstacle avoidance optimization control system for photovoltaic power plants, applied to the drone path obstacle avoidance optimization control method for photovoltaic power plants as described in any one of claims 1-8, characterized in that, It includes a patrol path determination module, a 3D model registration module, an environmental risk assessment module, a local obstacle avoidance optimization module, a patrol quality assessment module, and a protection level classification module; The inspection path determination module is used to obtain a three-dimensional digital twin model of the photovoltaic power station, and combine it with a multi-objective genetic algorithm to perform global path optimization and determine the initial inspection path. The 3D model registration module is used to collect 3D point cloud data of the environment during flight using lidar, downsample the 3D point cloud data and extract the region of interest, and use an improved NDT algorithm to register the 3D point cloud data of the environment with the 3D digital twin model to calculate the degree of environmental change. The environmental risk assessment module is used to identify dynamic obstacles based on the degree of environmental change, obtain the state parameters of the dynamic obstacles, construct an environmental risk assessment model, and determine the risk level of the dynamic obstacles. The local obstacle avoidance optimization module is used to obtain the flight state parameters of the UAV and the state parameters of dynamic obstacles, and combine them with the target input of the photovoltaic module's shooting angle to the MPC optimization model to perform local obstacle avoidance optimization. The MPC controller is used to perform rolling optimization solution, and the local obstacle avoidance optimization results are judged in real time to determine whether they meet the requirements. The inspection quality assessment module is used to acquire image data collected by the UAV after completing the inspection task, extract image assessment parameters, input them into a lightweight convolutional neural network image quality assessment model, generate inspection quality evaluation indicators, and determine whether to carry out a reshoot task based on the inspection quality evaluation indicators. The protection level classification module is used to determine the protection level of the inspection process based on the local obstacle avoidance optimization results by setting a three-level safety protection mechanism.

10. A readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it is used to implement the steps of the unmanned aerial vehicle path avoidance optimization control method for photovoltaic power plants as described in any one of claims 1-8.

Images