A method, device and equipment for unmanned aerial vehicle laser cleaning of solar photovoltaic panels

Abstract

The application discloses a kind of unmanned aerial vehicle laser cleaning method, device, electronic equipment and storage medium of solar photovoltaic panel, method includes: obtaining dirt image on solar photovoltaic panel and determining the position of the dirt image;Based on channel state information, obtain the initial position of unmanned aerial vehicle;According to the dirt image position to be cleaned and the initial position of unmanned aerial vehicle, adopt path planning model, determine the desired path of unmanned aerial vehicle from initial position to the position of dirt image;Based on the desired path, so that the unmanned aerial vehicle reaches the position of the dirt image and hovers, and using a preset unified accurate attitude control method, determine the working distance of unmanned aerial vehicle relative to solar photovoltaic panel;Based on the working distance, using the laser cleaning device to clean the dirt on solar photovoltaic panel.The application solves the technical problems of low cleaning efficiency and large pollution of solar photovoltaic panel in the prior art.

Description

Technical Field

[0001] This invention relates to the field of laser cleaning technology, and specifically to a method, apparatus, electronic device, and storage medium for unmanned aerial vehicle (UAV) laser cleaning of solar photovoltaic panels. Background Technology

[0002] In today's world, the use of renewable energy is increasing to reduce the use of coal and water for power generation. Effectively utilizing solar energy is an important way to address the power crisis, but some problems exist. For example, the accumulation of dirt and dust on the surface of solar photovoltaic panels reduces the amount of sunlight penetrating and reaching the solar cells, thus reducing the efficiency of the photovoltaic panels.

[0003] Currently, there are various technologies available, such as manual cleaning, vacuum cleaning, and electrostatic precipitator cleaning. Through these cleaning mechanisms, the efficiency of the panels can be improved by about 15-20%, but each technology has some drawbacks. Manual cleaning requires more manpower, continuous use of water or liquids may damage the photovoltaic panels, and vacuum cleaning mechanisms can cause scratches on the panels.

[0004] Therefore, a new solar photovoltaic panel cleaning system is needed to achieve the goals of low pollution, high precision, and high efficiency. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned technical deficiencies and provide a method, apparatus, equipment and storage medium for UAV laser cleaning of solar photovoltaic panels, thereby solving the technical problems of low cleaning efficiency, high pollution and low precision of solar photovoltaic panels in the prior art.

[0006] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a method for UAV laser cleaning of solar photovoltaic panels, comprising:

[0008] Acquire images of dirt on a solar photovoltaic panel and determine the location of the dirt images;

[0009] The initial position of the UAV is obtained based on channel state information;

[0010] Based on the location of the dirt image to be cleaned and the initial position of the drone, a path planning model is used to determine the expected path of the drone from the initial position to the location of the dirt image.

[0011] Based on the desired path, the drone reaches the location of the dirt image and hovers, and a preset unified and precise attitude control method is used to determine the working distance of the drone relative to the solar photovoltaic panel.

[0012] Based on the aforementioned working distance, the laser cleaning device is used to clean the dirt on the solar photovoltaic panel.

[0013] In some embodiments, acquiring an image of dirt on a solar photovoltaic panel and determining the location of the dirt image includes:

[0014] Using a pre-defined improved YOLOv5 model, image information of dirt on solar photovoltaic panels is extracted;

[0015] A rectangular coordinate system is established with the center point of the solar photovoltaic panel as the origin. The camera coordinate information is determined based on the dirt image information. The dirt image position coordinates corresponding to the camera coordinate information are determined based on the rectangular coordinate system.

[0016] In some embodiments, the preset improved YOLOv5 model includes a first deep convolutional layer, a convolutional layer, a batch normalization layer, an activation function layer, and a second deep convolutional layer connected in sequence.

[0017] In some embodiments, obtaining the initial position of the UAV based on channel state information includes:

[0018] Acquire drone signals;

[0019] Based on the UAV signal, a preset six-antenna circular array model is used to obtain six-channel synchronized channel state information;

[0020] Based on the channel state information, a preset multi-signal classification and synthesis algorithm is used to obtain the spatial characteristic azimuth and elevation angles.

[0021] Based on the azimuth and elevation angles, the initial position of the UAV is determined using a preset least squares method.

[0022] In some embodiments, employing a path planning model to determine the desired path of the UAV from its initial position to the location of the dirt image includes:

[0023] Based on the image location of the dirt to be cleaned and the initial position of the drone, a preset discrete A is used. * The search method determines the initial path of the drone;

[0024] The initial path of the UAV is smoothed using a pre-defined hybrid path planning method with turning constraints to obtain the desired path.

[0025] In some embodiments, the use of a preset discrete A * The search method determines the initial path of the drone, including:

[0026] Obtain the actual distance between the initial position of the drone and the position of the image of the dirt to be cleaned, as well as the travel distance between the starting node and the position of the image of the dirt;

[0027] Based on the actual distance and travel distance, a preset discrete A* search method is used to determine the initial path between the UAV's initial position and the location of the dirt image.

[0028] In some embodiments, determining the working distance of the UAV relative to the solar photovoltaic panel using a preset unified and precise attitude control method includes:

[0029] Based on the interference dynamics of the three attitude axes, a multi-source interference system model for attitude control of the UAV in the pitch, roll, and yaw channels is established respectively.

[0030] Based on the attitude control multi-source disturbance system model, the inner loop nominal control quantity and disturbance estimate are obtained;

[0031] The nominal control quantity of the inner loop and the interference estimate are combined to determine the composite anti-interference attitude control law, so as to determine the working distance of the UAV relative to the solar photovoltaic panel.

[0032] Secondly, the present invention also provides a drone laser cleaning device for solar photovoltaic panels, comprising:

[0033] The acquisition module is used to acquire images of dirt on a solar photovoltaic panel and determine the location of the dirt images;

[0034] The UAV positioning module is used to obtain the initial position of the UAV based on channel state information.

[0035] The desired path determination module is used to determine the desired path of the drone from its initial position to the position of the dirt image based on the location of the dirt image to be cleaned and the initial position of the drone, using a path planning model.

[0036] The working distance determination module is used to determine the working distance of the UAV relative to the solar photovoltaic panel based on the desired path, so that the UAV reaches the position of the dirt image and hovers, and adopts a preset unified and precise attitude control method.

[0037] The cleaning module is used to clean dirt on the solar photovoltaic panel using the laser cleaning device based on the working distance.

[0038] Thirdly, the present invention also provides an electronic device, comprising: a processor and a memory;

[0039] The memory stores a computer-readable program that can be executed by the processor;

[0040] When the processor executes the computer-readable program, it implements the steps in the UAV laser cleaning method for solar photovoltaic panels as described above.

[0041] Fourthly, the present invention also provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the UAV laser cleaning method for solar photovoltaic panels as described above.

[0042] Compared with existing technologies, the UAV laser cleaning method, apparatus, electronic device, and storage medium for solar photovoltaic panels provided by this invention first acquires images of dirt on the solar photovoltaic panel and determines the locations of the dirt images, which may be in multiple locations. Then, based on channel state information, the initial position of the UAV is obtained. Based on the locations of the dirt images to be cleaned and the initial position of the UAV, the UAV is selected to clean one location of dirt. A path planning model is used to determine the desired path for the UAV from its initial position to the location of the dirt image. Then, based on the desired path, the UAV reaches the location of the dirt image and hovers. A preset unified and precise attitude control method is used to determine the working distance of the UAV relative to the solar photovoltaic panel. Finally, based on the working distance, the UAV approaches the solar photovoltaic panel at a set angle, thereby enabling the laser cleaning device to efficiently clean the dirt on the solar photovoltaic panel, achieving a high-efficiency, low-pollution cleaning purpose. Attached Figure Description

[0043] Figure 1 This is a flowchart of an embodiment of the UAV laser cleaning method for solar photovoltaic panels provided by the present invention;

[0044] Figure 2 This is a flowchart of an embodiment of step S101 in the UAV laser cleaning method for solar photovoltaic panels provided by the present invention;

[0045] Figure 3 This is a flowchart of an embodiment of step S102 in the UAV laser cleaning method for solar photovoltaic panels provided by the present invention;

[0046] Figure 4 This is a flowchart of an embodiment of step S103 in the UAV laser cleaning method for solar photovoltaic panels provided by the present invention;

[0047] Figure 5 This is a flowchart of an embodiment of step S104 in the UAV laser cleaning method for solar photovoltaic panels provided by the present invention;

[0048] Figure 6This is a schematic diagram of an embodiment of the UAV laser cleaning device for solar photovoltaic panels provided by the present invention;

[0049] Figure 7 This is a schematic diagram of the operating environment of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0051] The present invention relates to a drone-mounted laser cleaning method, apparatus, electronic device, and storage medium for solar photovoltaic panels. This method utilizes a drone equipped with a laser cleaning device to clean solar photovoltaic panels, reducing the inefficiency of manual cleaning and alleviating physical labor. Furthermore, by employing laser cleaning, the laser causes a series of complex physical changes on the surface of the object to be cleaned, including vibration, melting, evaporation, and combustion, thereby removing the contaminants from the object's surface and recovering the contaminant powder, significantly reducing pollution. The method, apparatus, device, or computer-readable storage medium involved in this invention can be integrated with the aforementioned system or operate independently.

[0052] This embodiment provides a drone-based laser cleaning method for solar photovoltaic panels. Figure 1 This is a flowchart of the UAV laser cleaning of solar photovoltaic panels provided in an embodiment of the present invention. Please refer to [link / reference]. Figure 1 ,include:

[0053] S101. Obtain an image of dirt on the solar photovoltaic panel and determine the location of the dirt image;

[0054] S102. Obtain the initial position of the UAV based on channel state information;

[0055] S103. Based on the location of the dirt image to be cleaned and the initial position of the drone, a path planning model is used to determine the desired path of the drone from the initial position to the location of the dirt image.

[0056] S104. Based on the desired path, the drone reaches the location of the dirt image and hovers, and a preset unified and precise attitude control method is used to determine the working distance of the drone relative to the solar photovoltaic panel.

[0057] S105. Based on the working distance, the laser cleaning device is used to clean the dirt on the solar photovoltaic panel.

[0058] In this embodiment, firstly, images of dirt on the solar photovoltaic panel are acquired and the locations of these dirt images are determined. There may be multiple dirt images. Then, based on channel state information, the initial position of the drone is obtained. Based on the locations of the dirt images to be cleaned and the drone's initial position, the drone is selected to clean one specific dirt image. A path planning model is used to determine the desired path for the drone from its initial position to the location of the dirt image. Then, based on this desired path, the drone reaches the location of the dirt image and hovers. A preset unified and precise attitude control method is used to determine the working distance of the drone relative to the solar photovoltaic panel. Finally, based on this working distance, the drone approaches the solar photovoltaic panel at a set angle, enabling the laser cleaning device to efficiently clean the dirt on the solar photovoltaic panel, achieving a high-efficiency, low-pollution cleaning purpose.

[0059] In some embodiments, please refer to Figure 2 The step of acquiring an image of dirt on a solar photovoltaic panel and determining the location of the dirt image includes:

[0060] S201. Using the improved YOLOV5 model, extract image information of dirt on solar photovoltaic panels;

[0061] S202. Establish a rectangular coordinate system with the center point of the solar photovoltaic panel as the origin, determine the camera coordinate information based on the dirt image information, and determine the dirt image position coordinates corresponding to the camera coordinate information based on the rectangular coordinate system.

[0062] In this embodiment, the improved YOLOv5 model adds centering and scaling calibration at the beginning and end of the original batch normalization module, enhancing effective features and forming a more stable feature distribution, thereby improving the feature extraction capability of the network model. Simultaneously, the original cross-entropy loss function based on confidence is improved to a loss function based on smooth Kullback-Leibler divergence. Furthermore, to reduce information loss, a CSandGlass module is designed on the backbone feature extraction network of YOLOv5 to replace the remaining modules. Using the improved YOLOv5 model can improve the accuracy and speed of object detection, while also making convergence easier.

[0063] Specifically, in the CSandGlass model, assuming the input is a feature map of height × width × number of channels, the input feature map is first subjected to a 3×3 depthwise convolution, which yields the first feature map. Then, the first feature map is convolved using cbl, followed by batch normalization, and then an activation function layer (Leaky ReLU) is applied to obtain the second feature map. The second feature map is then subjected to another 3×3 depthwise convolution, which yields the third feature map. Finally, the original feature map and the third feature map are added together to obtain target feature maps with different channel importance.

[0064] Among them, cbl is the smallest component in the YOLOv5 network structure, which consists of three parts: convolution, batch normalization, and Leaky ReLU activation function.

[0065] Furthermore, by using a depth camera mounted on the drone to identify dirt, the target object is identified using the improved YOLOv5 model. Then, the camera coordinates of all detected target objects are converted into real-world coordinates using the depth camera to determine the location coordinates of the dirt image and transmit it back to the ground control console to provide coordinate information for the drone's trajectory planning.

[0066] In some embodiments, the preset improved YOLOv5 model includes a first deep convolutional layer, a convolutional layer, a batch normalization layer, an activation function layer, and a second deep convolutional layer connected in sequence.

[0067] In this embodiment, the improved YOLOv5 model can reduce information loss and obtain more accurate dirt information.

[0068] In some embodiments, please refer to Figure 3 The step of obtaining the initial position of the UAV based on channel state information includes:

[0069] S301. Acquire drone signals;

[0070] S302. Based on the UAV signal, a preset six-antenna circular array model is used to obtain the channel state information of six channels synchronized.

[0071] S303. Based on the channel state information, a preset multi-signal classification and synthesis algorithm is used to obtain the spatial characteristic azimuth and elevation angles.

[0072] S304. Based on the azimuth and elevation angles, the initial position of the UAV is determined using a preset least squares method.

[0073] In this embodiment, the multi-signal classification and synthesis algorithm specifically includes a multi-signal classification method and a recursive application and projection-multi-signal classification algorithm. Specifically, the signal transmitted by the UAV is received, and the signal is decomposed into multiple sub-signals using Empirical Mode Decomposition (EMD). Then, Fourier Transform (FT) is applied to the sub-signals to solve for the spectrum and extract SFS (signal spectrum) features. Wavelet transform is used to decompose the signal and solve for the entropy to extract WEE (wavelet energy entropy) features. At the same time, Short Time Fourier Transform (STFT) is used to obtain the power spectral density (PSD) and solve for the entropy to extract the power spectral density features. Then, the signal features transmitted by the UAV are used as training data, and the collected sample data is input into the classifier for UAV detection. The UAV detection uses multiple machine learning algorithms such as Support Vector Machine (SVM), Random Forest (RF), Naive Bayes (NB), Ensemble Learning (EL), and k Nearest Neighbors (KNN) to distinguish UAV signals from non-UAV signals by detecting UAV signals.

[0074] Furthermore, a six-antenna circular array model is used to obtain six-channel synchronous CSI, and the spatial information AOA (azimuth angle) and AOE (elevation angle) are obtained through the MUSIC (multi-signal classification) and RAP (recursive application and projection)-MUISC algorithms. Finally, the least squares algorithm is used to combine multiple receivers to locate the UAV in space.

[0075] It should be noted that the CSI model, which reflects the channel impulse response (CIR) of signal propagation characteristics, can be expressed by the following formula: Where N is the total number of paths, α_n is the path coefficient of the Nth path, and τ_n is the propagation delay of the Nth path.

[0076] Furthermore, using Orthogonal Frequency Division Multiplexing (OFDM) modulation with K subcarriers, the CFR matrix of all subcarriers across all antennas on each receiver can be expressed by the following formula: H = (H 1，1 H 1，K H M，1 H M，K ], where H can be represented as the joint estimate of AOA and AOE, expressed by the following formula: H = [H 1，1 H 1，K H M，1 H M，K ], where E is the noise vector and S is the attenuation vector. It is the turning matrix; with the center of the circle as the reference point, the distance between the signal to each array element and the center of the circle is expressed by the following formula: H = (H 1，1 H 1，K H M，1 HM，K Meanwhile, the array flow matrix for constructing a circular array is expressed by the following formula: Where r_i = 2π(i-1) / M, i = 0, 1, 2, ..., M-1, when there are N paths, the direction vector of the signal of the k-th path can be expressed as: Therefore, the M×N-dimensional direction matrix can be represented as:

[0077] Furthermore, by arranging the eigenvalues ​​and eigenvectors sequentially, the first L eigenvectors form the signal space U_S, and the last ML eigenvectors form the noise space U_N. A spatial spectrum function of a uniform circular array can be constructed to search for the signal's direction angle, thus yielding the following model: Among them, angle Corresponding to These are the AOE and AOA of the incident wave. Since MUSIC has greater complexity for 2D angle searches, it is applied to verify the feasibility of the UAV's AOA. The estimated AOA is compared with the actual AOA to verify the accuracy of the UAV's positioning.

[0078] Specifically, this patent locates the drone by combining the service levels of multiple receivers. Assuming there are R receivers used for positioning in the experiment, the drone can be projected onto the ground using a least squares algorithm, which can be expressed by the following model: in This represents the actual AOA from position to the i-th receiver, while The AOA representing the line-of-sight path is selected by the i-th receiver. The drone altitude hi can be calculated by the pitch angle θi of each receiver and the position of the drone projected onto the ground. Finally, h is obtained by averaging the hi values ​​of multiple receivers, and the drone's spatial position pos_d is obtained by combining h and pos.

[0079] In some embodiments, please refer to Figure 4 The method of using a path planning model to determine the desired path of the UAV from its initial position to the location of the dirt image includes:

[0080] S401. Based on the image location of the dirt to be cleaned and the initial position of the drone, a preset discrete A is used. * The search method determines the initial path of the drone;

[0081] S402. The initial path of the UAV is smoothed using a preset hybrid path planning method with turning constraints to obtain the desired path.

[0082] In this embodiment, a novel integrated path planning and trajectory tracking control framework for online autonomous flight of unmanned aerial vehicles is adopted to simultaneously avoid static and dynamic obstacles and compensate for the adverse effects of model uncertainties, thereby bridging the gap between planning and control in integrated guidance and control.

[0083] In one specific embodiment, the path planner's input consists of the target configuration X_G and the occupancy grid provided by the camera. After the input is imported, it is checked to ensure that the starting and target configurations are within the occupancy grid. The hybrid path planning algorithm includes A... * The method and potential field are used to perform path search tasks and generate flyable paths for UAVs. The generated paths are usually not smooth. The path is then passed to a pre-set hybrid path planning method with turning constraints to smooth it. The final smooth path ensures that the UAV can fly and is very close to the optimal path.

[0084] In some embodiments, the use of a preset discrete A * The search method determines the initial path of the drone, including:

[0085] Obtain the actual distance between the initial position of the drone and the position of the image of the dirt to be cleaned, as well as the travel distance between the starting node and the position of the image of the dirt;

[0086] Based on the actual distance and travel distance, a preset discrete A* search method is used to determine the initial path between the UAV's initial position and the location of the dirt image.

[0087] In this embodiment, it should be noted that A * The search is a deterministic heuristic that uses a strategy of distance to the target and travel distance from the starting node to the target to explore fewer nodes in a given number of nodes X. i The search agent finds the data structure and selects the node with the lowest heuristic value. The heuristic calculation formula is F(X). i )=G(X i )+H(X i ), where G(X) i ) is from the starting node X S The cost to reach the current node Xi, H(X) i ) can be considered as originating from the current node X i To target set X G Search heuristics for cost estimation of the shortest path.

[0088] Furthermore, a sufficiently smooth path is crucial for practical applications. To achieve this, we first consider that the final path consists of a series of waypoints, which can be represented as X. i =(xei y ei ), i∈[1,N], where N represents the total number of waypoints in the generated path; then, O i ΔX represents the position of the obstacle closest to the i-th waypoint. i =X i -X i-1 This represents the displacement vector of the i-th waypoint; based on the distance between the waypoint and the obstacle, the first term of the cost function is defined by the formula... It means that d max This is the maximum distance that an obstacle can affect the path cost; this distance is constant for all obstacles and waypoints; the weighting coefficient w obs This determines the importance of the cost function; in order to penalize heavier obstacles when approaching obstacles at waypoints, δ obs It is a quadratic penalty function. Furthermore, if |X i -O i |-d max >0 allows us to ignore the term POB because the drone is now too far from the obstacle. At each waypoint, we must check that we haven't created a path the drone cannot follow, meaning the curvature needs to be greater than the maximum flyable curvature. Therefore, the curvature term is defined as... in Let f represent the change in the tangent angle of the i-th waypoint relative to the other two waypoints, i-1 and i+1. cmax w represents the maximum flyable curvature. cur The impact of the control cost function on path variation, σ cur It is a quadratic function. The curvature of a path is defined as... The third and final term of the cost function evaluates the displacement vector between two waypoints. Waypoints farther from their neighbors receive a higher cost. Furthermore, waypoints that change path direction are also assigned a higher cost. Finally, the cost function is minimized by running gradient descent, and the algorithm's output is a geometrically smooth list of waypoints suitable for UAV flight.

[0089] In some embodiments, please refer to Figure 5 The method of using a preset unified and precise attitude control to determine the working distance of the UAV relative to the solar photovoltaic panel includes:

[0090] S501. Based on the interference dynamics of three attitude axes, establish a multi-source interference system model for attitude control of the UAV in the pitch, roll, and yaw channels respectively.

[0091] S502. Based on the attitude control multi-source interference system model, obtain the inner loop nominal control quantity and interference estimate;

[0092] S503. Combine the nominal control quantity of the inner loop and the interference estimate to determine the composite anti-interference attitude control law, so as to determine the working distance of the UAV relative to the solar photovoltaic panel.

[0093] In this embodiment, a unified precise attitude control method based on improved linear active disturbance rejection control (LADRC) is adopted, which can achieve precise attitude control. Compared with model-based control algorithms, this control algorithm has higher robustness to model mismatch. At the same time, flight results show that the controller ensures high control accuracy and uniform control quality in different flight modes, which can ensure that the UAV maintains a fixed distance from the solar photovoltaic panel, so that the energy of the laser beam is concentrated and the purpose of efficiently cleaning the solar panel is achieved.

[0094] It should be noted that the attitude controller in this embodiment has been optimized to achieve precise attitude control and high robustness. In a specific embodiment, the controller can be divided into two loops. One is an outer loop that generates the required angular acceleration vector Ω, which uses a cascaded proportional controller based on attitude error and angular rate error. The other is an inner loop controller consisting of an improved LADRC controller and a control distributor. The improved LADRC compensates for the relatively slow rotor slapping dynamics to avoid oscillations. Despite disturbances and model uncertainties, the LADRC controller forces the device to track the required angular acceleration. Then, based on the angular acceleration and thrust command, the control distributor uses a daisy-chain-based control distribution algorithm to calculate the control command vector u. cmd .

[0095] Based on the above-described method for UAV laser cleaning of solar photovoltaic panels, this invention also provides a corresponding UAV laser cleaning device 600 for solar photovoltaic panels. Please refer to [link to relevant documentation]. Figure 6 The drone laser cleaning device 600 for solar photovoltaic panels includes an acquisition module 610, a drone positioning module 620, a desired path determination module 630, a working distance determination module 640, and a cleaning module 650.

[0096] The acquisition module 610 is used to acquire images of dirt on a solar photovoltaic panel and determine the location of the dirt images;

[0097] The UAV positioning module 620 is used to obtain the initial position of the UAV based on channel state information;

[0098] The desired path determination module 630 is used to determine the desired path of the drone from its initial position to the position of the dirt image based on the location of the dirt image to be cleaned and the initial position of the drone, using a path planning model.

[0099] The working distance determination module 640 is used to determine the working distance of the drone relative to the solar photovoltaic panel based on the desired path, so that the drone reaches the position of the dirt image and hovers, and adopts a preset unified and precise attitude control method.

[0100] The cleaning module 650 is used to clean dirt on the solar photovoltaic panel using the laser cleaning device based on the working distance.

[0101] like Figure 7 As shown, based on the aforementioned UAV laser cleaning method for solar photovoltaic panels, this invention also provides an electronic device, which can be a mobile terminal, desktop computer, laptop, handheld computer, server, or other computing device. The electronic device includes a processor 710, a memory 720, and a display 730. Figure 7 Only some components of the electronic device are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0102] In some embodiments, memory 720 may be an internal storage unit of the electronic device, such as a hard drive or memory. In other embodiments, memory 7720 may be an external storage device of the electronic device, such as a plug-in hard drive, smart media card (SMC), secure digital card (SD), flash card, etc. Further, memory 720 may include both internal and external storage units. Memory 720 is used to store application software and various types of data installed on the electronic device, such as program code installed on the electronic device. Memory 7720 may also be used to temporarily store data that has been output or will be output. In one embodiment, memory 720 stores a drone laser cleaning program 740 for solar photovoltaic panels, which can be executed by processor 710 to implement the drone laser cleaning method for solar photovoltaic panels according to the embodiments of this application.

[0103] In some embodiments, processor 710 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 720 or process data, such as performing a drone laser cleaning method for solar photovoltaic panels.

[0104] In some embodiments, display 730 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 730 is used to display information about the drone laser cleaning equipment on the solar photovoltaic panel and to display a visual user interface. Components 710-730 of the electronic device communicate with each other via a system bus.

[0105] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The storage medium can be a memory, magnetic disk, optical disk, etc.

[0106] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method of unmanned aerial vehicle laser cleaning of solar photovoltaic panels, characterized in that, The method comprises the following steps: acquiring a dirt image on a solar photovoltaic panel and determining a position of the dirt image; based on channel state information, acquiring an initial position of a UAV, comprising: acquiring a UAV signal; based on the UAV signal, acquiring six-channel synchronous channel state information by using a preset six-antenna circular array model; according to the channel state information, obtaining a spatial feature azimuth angle and an elevation angle by using a preset multiple signal classification comprehensive algorithm; and according to the azimuth angle and the elevation angle, determining the initial position of the UAV by using a preset least square method; according to the position of the dirt image to be cleaned and the initial position of the UAV, determining a desired path of the UAV from the initial position to the position of the dirt image by using a path planning model; based on the desired path, making the UAV reach the position of the dirt image and hover, and determining a working distance of the UAV relative to the solar photovoltaic panel by using a preset unified precise attitude control method, comprising: based on the interference dynamics of three attitude axes, establishing an attitude control multi-source interference system model of the UAV pitch, roll and yaw three channels respectively; based on the attitude control multi-source interference system model, acquiring an inner loop nominal control quantity and an interference estimation value; and by compounding the inner loop nominal control quantity and the interference estimation value, determining a composite anti-interference attitude control law to determine the working distance of the UAV relative to the solar photovoltaic panel; based on the working distance, cleaning the dirt on the solar photovoltaic panel by using a laser cleaning device; the path planning model is used to determine the desired path of the UAV from the initial position to the position of the dirt image, comprising: According to the image position of the dirt to be cleaned and the initial position of the unmanned aerial vehicle, an initial path for the unmanned aerial vehicle to travel is determined by using a preset discrete search method. search method, determine the initial path of the unmanned aerial vehicle to travel; the initial path of the UAV is smoothed by using a preset hybrid path planning method with turning constraint to obtain the desired path.

2. The drone laser cleaning method of solar photovoltaic panels according to claim 1, characterized in that, the dirt image on the solar photovoltaic panel is acquired and the position of the dirt image is determined, comprising: an improved YOLOV5 model is used to extract dirt image information on the solar photovoltaic panel; a rectangular coordinate system is established with the center point of the solar photovoltaic panel as the origin, camera coordinate information is determined according to the dirt image information, and the position coordinates of the dirt image corresponding to the camera coordinate information are determined based on the rectangular coordinate system.

3. The drone laser cleaning method of solar photovoltaic panels according to claim 2, characterized in that, The preset improved YOLOV5 model comprises a first deep convolution layer, a convolution layer, a batch normalization layer, an activation function layer and a second deep convolution layer connected in sequence.

4. The drone laser cleaning method of solar photovoltaic panels as claimed in claim 1, wherein, The preset discrete The search method determines the initial path of the UAV, comprising: an actual distance between the initial position of the UAV and the position of the dirt image to be cleaned, and a travel distance between a starting node and the position of the dirt image are acquired; based on the actual distance and the travel distance, a preset discrete A* search method is used to determine an initial path of the UAV between the initial position and the position of the dirt image.

5. An unmanned aerial vehicle laser cleaning device for solar photovoltaic panels, characterized in that, The method comprises the following steps: an acquisition module is configured to acquire a dirt image on a solar photovoltaic panel and determine a position of the dirt image; The unmanned aerial vehicle positioning module is configured to obtain an initial position of the unmanned aerial vehicle based on channel state information, and includes: obtaining an unmanned aerial vehicle signal; obtaining six-channel synchronous channel state information based on the unmanned aerial vehicle signal and using a preset six-antenna circular array model; obtaining a spatial feature azimuth angle and an elevation angle based on the channel state information and using a preset multiple signal classification synthesis algorithm; and determining the initial position of the unmanned aerial vehicle based on the azimuth angle and the elevation angle and using a preset least square method. The expected path determination module is configured to determine an expected path of the unmanned aerial vehicle from the initial position to the position of the dirt image based on the position of the dirt image to be cleaned and the initial position of the unmanned aerial vehicle and using a path planning model. The working distance determination module is configured to determine a working distance of the unmanned aerial vehicle relative to the solar photovoltaic panel based on the expected path so that the unmanned aerial vehicle reaches the position of the dirt image and hovers, and using a preset unified accurate attitude control method, and includes: establishing attitude control multi-source interference system model of three channels of the unmanned aerial vehicle, i.e., pitch, roll, and yaw, based on interference dynamics of three attitude axes; obtaining an inner loop nominal control quantity and an interference estimation value based on the attitude control multi-source interference system model; and determining a composite anti-interference attitude control law by compounding the inner loop nominal control quantity and the interference estimation value, so as to determine the working distance of the unmanned aerial vehicle relative to the solar photovoltaic panel. The cleaning module is configured to clean dirt on the solar photovoltaic panel based on the working distance and using a laser cleaning device. The expected path determination module includes: According to the image position of the dirt to be cleaned and the initial position of the unmanned aerial vehicle, an initial path for the unmanned aerial vehicle to travel is determined by using a preset discrete search method. search method, determine the initial path of the unmanned aerial vehicle to travel; The initial path of the unmanned aerial vehicle is smoothed by using a preset hybrid path planning method with turning constraint, so as to obtain the expected path.

6. An electronic device, comprising: The device includes a memory and a processor, wherein The memory is configured to store a program. The processor is coupled to the memory and is configured to execute the program stored in the memory, so as to implement the steps of the unmanned aerial vehicle laser cleaning method of the solar photovoltaic panel according to any one of claims 1 to 4.

7. A computer readable storage medium characterized by The device is configured to store a computer-readable program or instruction, which, when executed by a processor, can implement the steps of the unmanned aerial vehicle laser cleaning method of the solar photovoltaic panel according to any one of claims 1 to 4.

Images