A multi-mode intelligent cleaning drone system suitable for photovoltaic panels

Abstract

The application belongs to the technical field of cleaning unmanned aerial vehicles, and discloses a multi-mode intelligent cleaning unmanned aerial vehicle system suitable for photovoltaic panels; the system comprises an unmanned aerial vehicle and a cleaning control unit; the unmanned aerial vehicle is provided with an infrared vision sensor, a laser radar and a cleaning module; the cleaning module is arranged at the bottom of the unmanned aerial vehicle; the cleaning control unit comprises an AI control navigation module; the AI control navigation module is arranged at the central position of the unmanned aerial vehicle body and connected with a flight control bus; the infrared vision sensor and the laser radar are fused to accurately identify the contaminated area; in combination with the parameterized control of the cleaning execution mechanism, the system can maintain good coverage and adhesion under the wind field, reduce the consumption of cleaning agents and water, and improve the cleaning quality and system reliability.

Description

Technical Field

[0001] This invention belongs to the field of cleaning drone technology, and in particular relates to a multi-mode intelligent cleaning drone system suitable for photovoltaic panels. Background Technology

[0002] With the widespread application of photovoltaic power generation, the number and installed capacity of photovoltaic panels have grown rapidly, and their operating efficiency is significantly affected by the cleanliness of their surfaces. During long-term operation, photovoltaic panels are susceptible to the adhesion of pollutants such as windblown sand, dust, bird droppings, and water stains, leading to decreased light transmittance, localized heating, and power output attenuation, thus impacting power generation efficiency and economic benefits. Traditional photovoltaic panel cleaning methods mainly rely on manual rinsing or fixed spray devices, which suffer from high labor intensity, water waste, low efficiency, and high safety risks. Furthermore, some ground-mounted or tracked cleaning equipment is limited by terrain and wind conditions, making it difficult to achieve large-scale, uniform, and efficient cleaning operations.

[0003] In recent years, drones have been widely used in photovoltaic (PV) operation and maintenance due to their high mobility and flexible operation. However, existing cleaning drones mostly use fixed spraying parameters and preset routes, making it difficult to dynamically adjust according to real-time wind fields and pollution distribution, which can easily lead to uneven cleaning or wasted dosage. Therefore, developing a PV cleaning drone system with intelligent pollution identification and adaptive spraying control capabilities has become a key technological direction for improving the efficiency and intelligence level of PV operation and maintenance. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-mode intelligent cleaning drone system suitable for photovoltaic panels, which solves the problems of high manual labor intensity, low cleaning efficiency, fixed spraying parameters, and inaccurate pollution identification in existing photovoltaic module cleaning methods. It enables intelligent detection and adaptive spraying of the photovoltaic panel surface, thereby improving the cleaning effect and the level of system intelligence.

[0005] To achieve the above objectives, the present invention provides a multi-mode intelligent cleaning drone system suitable for photovoltaic panels, including a drone and a cleaning control unit. The drone is equipped with an infrared vision sensor, a lidar and a cleaning module, and the cleaning module is located at the bottom of the drone.

[0006] The cleaning control unit includes an AI control and navigation module, which is located in the center of the drone fuselage and connected to the flight control bus. It is used to plan the cleaning route and control the operation of the cleaning module to achieve non-contact spray cleaning and accelerated drying of the photovoltaic panel surface.

[0007] Furthermore, the cleaning module consists of a liquid path, a gas path, a mixing spray chamber, a water outlet, a liquid storage filter, a cooling fan, and an electromagnetic vaporization device. The liquid path is equipped with a pressure stabilizing pump, a liquid path solenoid valve, a flow sensor, a liquid pressure sensor, and a liquid path proportional valve. The gas path is equipped with a gas storage chamber, a gas path solenoid valve, a gas pressure sensor, and a gas path proportional valve. The liquid and gas paths are mixed in the mixing spray chamber by a venturi structure and then sprayed out by an electronically controlled nozzle.

[0008] Furthermore, a liquid transmission interface is provided on the liquid line, and a gas transmission interface is provided on the gas line. Both the liquid transmission interface and the gas transmission interface are equipped with a self-sealing structure.

[0009] Furthermore, the drone is equipped with a powered rotor assembly, which consists of a brushless motor, blades, and an electronic speed controller. The brushless motor is fixed to the drone, and the output shaft of the brushless motor is connected to the blades via the electronic speed controller to drive the blades to rotate.

[0010] Furthermore, a water tank is fixed to the middle of the fuselage of the UAV by a vibration damping bracket. The water tank is sealed to one end of a pressure-resistant hose, and the other end of the pressure-resistant hose is connected to a liquid storage and filtration device. The outlet of the liquid storage and filtration device is connected to the liquid path, and the other end of the liquid path is connected to the outlet, together forming a closed fluid transmission circuit.

[0011] Furthermore, the AI ​​control navigation module includes a GPS positioning unit, an IMU inertial measurement unit, a central processing unit, and a data fusion unit;

[0012] The GPS positioning unit is used to obtain the drone's position, relative altitude, and flight speed.

[0013] The IMU (Inertial Measurement Unit) is used to acquire the tilt angle of the aircraft.

[0014] The data fusion unit is used to fuse the visible light image of the photovoltaic panel surface collected by the infrared vision sensor and the infrared temperature matrix, as well as the three-dimensional point cloud data collected by the lidar.

[0015] The central processing unit, based on the processing results of the data fusion unit, combined with the UAV's flight speed, relative altitude, and body tilt angle, and according to the pollution intensity on the photovoltaic panel surface, performs closed-loop adjustment of the liquid flow rate, injection pressure, and proportional valve opening of the cleaning module, and corrects the liquid flow rate and injection pressure when the lateral wind speed is greater than the preset wind speed threshold.

[0016] Furthermore, the method for closed-loop regulation of the liquid flow rate, injection pressure, and proportional valve opening of the cleaning module includes:

[0017] When the drone flies over the area to be cleaned, it collects visible light images and corresponding infrared temperature matrices of the photovoltaic panel surface, denoises and synchronizes the images and temperature data to form a fused image, performs region segmentation to obtain sub-regions and calculates the pollution intensity of the sub-regions.

[0018] A local grid map is generated based on 3D point cloud data collected by lidar to obtain relative altitude, aircraft tilt angle, and flight speed;

[0019] Based on pollution intensity, flight speed and relative altitude, the target liquid flow rate, target injection pressure and injection pattern are determined by consulting the injection parameter mapping table.

[0020] Obtain the current actual liquid flow rate and actual injection pressure, and adjust the opening of the liquid circuit proportional valve, the speed of the pressure stabilizing pump, and the opening of the gas circuit proportional valve to gradually bring the actual injection flow rate and pressure to the target values.

[0021] Furthermore, the method for performing region segmentation to obtain sub-regions includes:

[0022] Real-time acquisition of visible light grayscale images of the photovoltaic panel surface and infrared temperature distribution matrix corresponding to the grayscale images, where each pixel corresponds to both a grayscale value and a temperature value.

[0023] Median filtering and grayscale normalization are performed on the grayscale image and the infrared temperature matrix respectively. The two types of images are synchronized in time based on the timestamp, so that grayscale pixels and temperature pixels acquired at the same time form fused pixels to merge the image.

[0024] For each pixel, a feature vector containing temperature and grayscale values ​​is constructed. When the temperature difference between adjacent pixels is less than a preset temperature similarity threshold and the grayscale difference is less than a preset grayscale similarity threshold, the adjacent pixels are merged into the same sub-region. The above merging process is repeated until there are no more adjacent pixels that can be merged, thereby obtaining multiple sub-regions.

[0025] Furthermore, the method for correcting the liquid flow rate and injection pressure when the lateral wind speed is greater than a preset wind speed threshold includes:

[0026] The heading angle and ground speed vector of the UAV are obtained based on the GPS positioning unit and the IMU inertial measurement unit. The trajectory lateral deviation caused by the wind field is calculated based on the deviation angle between the heading angle and the ground speed vector, and the lateral wind speed component is obtained.

[0027] When the lateral wind speed component is greater than the preset wind speed threshold, it is determined that there is significant lateral wind disturbance in the current flight segment, and the wind disturbance compensation control state is entered.

[0028] The flow compensation and pressure compensation are calculated based on the crosswind speed component. The flow compensation is determined by multiplying the crosswind speed by a coefficient kw1, and the pressure compensation is determined by multiplying the crosswind speed by a coefficient kw2.

[0029] The flow compensation is added to the original target fluid flow rate, and the pressure compensation is added to the original target injection pressure to obtain the corrected target fluid flow rate and the corrected target injection pressure.

[0030] The corrected target fluid flow rate and target injection pressure are sent to the control terminals of the liquid circuit proportional valve, the pressure stabilizing pump and the gas circuit proportional valve. By increasing the opening of the liquid circuit proportional valve and increasing the speed of the pressure stabilizing pump, the initial injection velocity at the injection outlet is increased synchronously, thereby shortening the flight time of the droplets in the air and reducing the drift distance under the action of crosswind.

[0031] The lateral trajectory compensation is calculated based on the lateral wind speed component. The lateral trajectory compensation is determined by multiplying the lateral wind speed by a coefficient kw3.

[0032] The lateral trajectory compensation is superimposed on the lateral position command of the current flight segment, causing the UAV to deflect laterally in the opposite direction of the wind within the current flight segment.

[0033] Therefore, the present invention employs the above-mentioned multi-mode intelligent cleaning drone system suitable for photovoltaic panels, which has the following beneficial effects:

[0034] This invention adopts a modular structure design, which makes the maintenance and upgrading of the cleaning system easier. It can quickly switch between water washing, air blowing or water-air combined operation modules according to different pollution characteristics to adapt to various climates and pollution environments.

[0035] The drone has achieved full automation of the photovoltaic panel cleaning process. The system integrates GPS and IMU for high-precision navigation and attitude stabilization; the AI ​​control and navigation module optimizes the flight path and spraying parameters online by integrating multi-source data; when the battery or liquid level is below the threshold, it can automatically return to the drone for resupply and upload data, enabling unattended long-cycle operations.

[0036] This invention achieves accurate identification of contaminated areas by fusing infrared vision sensors and lidar; combined with parameterized control of the cleaning actuator, it can maintain good coverage and adhesion even in windy conditions, reduce cleaning agent and water consumption, and improve cleaning quality and system reliability. Attached Figure Description

[0037] Figure 1 This is a three-dimensional structural diagram of a multi-mode intelligent cleaning drone system suitable for photovoltaic panels according to the present invention;

[0038] Figure 2 This is a top view of a multi-mode intelligent cleaning drone system for photovoltaic panels according to the present invention;

[0039] Figure 3 This is a front view of a multi-mode intelligent cleaning drone system for photovoltaic panels according to the present invention;

[0040] Figure 4 This is a schematic diagram of the autonomous cleaning operation process based on AI control for a multi-mode intelligent cleaning drone system for photovoltaic panels according to the present invention;

[0041] Figure 5 This is a cross-sectional view of the cleaning module of a multi-mode intelligent cleaning drone system for photovoltaic panels according to the present invention.

[0042] Figure label:

[0043] 1. Unmanned Aerial Vehicle (UAV); 2. Powered Rotor Assembly; 3. Infrared Vision Sensor; 4. Water Tank; 7. LiDAR; 8. Water Outlet; 9. Cleaning Module; 91. Liquid Storage and Filtration Device; 92. Cooling Fan; 93. Electromagnetic Vaporization Device; 94. Pressure Stabilizing Pump. Detailed Implementation

[0044] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0045] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0046] Example 1

[0047] like Figures 1 to 5 As shown, the present invention provides a multi-mode intelligent cleaning drone system suitable for photovoltaic panels, comprising:

[0048] Drone 1 and cleaning control unit.

[0049] The drone 1 is equipped with an infrared vision sensor 3, a lidar 7, and a cleaning module 9, which is located at the bottom of the drone 1.

[0050] The cleaning control unit includes an AI control and navigation module, which is located in the center of the drone 1 and connected to the flight control bus. It is used to plan the cleaning route and control the operation of the cleaning module 9 to achieve non-contact spray cleaning and accelerated drying of the photovoltaic panel surface.

[0051] The cleaning module 9 consists of a liquid path, an air path, a mixing spray chamber, a water outlet 8, a liquid storage filter 91, a cooling fan 92, and an electromagnetic vaporization device 93. The liquid path includes a pressure stabilizing pump 94, a liquid path solenoid valve, a flow sensor, a liquid pressure sensor, and a liquid path proportional valve. The air path includes an air storage chamber, an air path solenoid valve, a gas pressure sensor, and an air path proportional valve. The liquid and air paths are mixed in the mixing spray chamber via a venturi structure and then sprayed out through an electronically controlled nozzle. The cleaning module 9 can quickly switch between water washing, air blowing, or a water-air composite mode according to different pollution levels. It can adjust the spray pressure, angle, and flow rate in real time based on control signals to dynamically optimize the spray range and droplet diameter, adapting to cleaning needs under different environmental conditions.

[0052] Among them, the pressure stabilizing pump 94 is used to provide a stable liquid supply pressure for the nozzle; the electromagnetic vaporization device 93 can partially vaporize or atomize the liquid before spraying, and its main function is to enhance the spraying effect and droplet characteristics.

[0053] The liquid circuit is equipped with a liquid transfer interface, and the gas circuit is equipped with a gas transfer interface. Both the liquid transfer interface and the gas transfer interface are equipped with a self-sealing structure. The self-sealing structure is a quick-change sealing assembly formed by combining an elastic sealing plug and a spring return valve core.

[0054] The UAV 1 is equipped with a powered rotor assembly 2, which consists of a brushless motor, blades, and an electronic speed controller. The brushless motor is fixed to the UAV 1, and its output shaft is connected to the blades via the electronic speed controller to drive the blades to rotate. To ensure overall structural strength and center of gravity stability, the UAV 1 has a quadcopter configuration, and the arms use vibration-damping connections, which ensures thrust redundancy and facilitates maintenance.

[0055] A water tank 4 is fixed in the middle of the fuselage of the UAV 1 by a vibration damping bracket. The water tank 4 is sealed to one end of a pressure-resistant hose, and the other end of the pressure-resistant hose is connected to a liquid storage and filtration device 91. The outlet of the liquid storage and filtration device 91 is connected to the liquid path, and the other end of the liquid path is connected to the outlet 8, together forming a closed fluid transmission circuit.

[0056] Among them, water tank 4 is equipped with a liquid level sensor to monitor the remaining liquid level.

[0057] The AI-controlled navigation module includes a GPS positioning unit, an IMU inertial measurement unit, a central processing unit, and a data fusion unit;

[0058] The GPS positioning unit is used to obtain the position, relative altitude, and flight speed of the UAV 1;

[0059] The IMU (Inertial Measurement Unit) is used to acquire the tilt angle of the drone; the data fusion unit is used to fuse the visible light image of the photovoltaic panel surface collected by the infrared vision sensor 3 and the infrared temperature matrix, as well as the three-dimensional point cloud data collected by the lidar 7; the central processing unit, based on the processing results of the data fusion unit, combines the drone's flight speed, relative altitude, and tilt angle, and according to the pollution intensity of the photovoltaic panel surface, performs closed-loop adjustment of the liquid flow rate, injection pressure, and proportional valve opening of the cleaning module 9, and corrects the liquid flow rate and injection pressure when the lateral wind speed is greater than the preset wind speed threshold.

[0060] The method for closed-loop regulation of the liquid flow rate, injection pressure, and proportional valve opening of the cleaning module 9 includes:

[0061] When the drone 1 flies over the area to be cleaned, the infrared vision sensor 3 collects visible light grayscale images of the photovoltaic panel surface in real time and the infrared temperature distribution matrix corresponding to the grayscale images. Each pixel corresponds to both a grayscale value and a temperature value. The AI ​​control and navigation module first performs median filtering and denoising processing on the grayscale images and the infrared temperature matrix, respectively, and performs time synchronization of the two types of images based on the timestamp, so that the grayscale pixels and temperature pixels collected at the same time form fused pixels to merge the images.

[0062] After obtaining the fused image, the AI-controlled navigation module performs region segmentation processing on the fused image.

[0063] The region segmentation process refers to dividing the fused image into several sub-regions with consistent pixel features based on the similarity of temperature difference and grayscale difference between pixels.

[0064] Specifically, a feature vector containing temperature and grayscale values ​​is constructed for each pixel. When the temperature difference between adjacent pixels is less than a preset temperature similarity threshold and the grayscale difference is less than a preset grayscale similarity threshold, the adjacent pixels are merged into the same sub-region. The above merging process is repeated until there are no more adjacent pixels that can be merged, thereby obtaining several sub-regions.

[0065] After completing the region segmentation, the AI ​​control navigation module counts the number of pixels in each sub-region whose grayscale value is lower than the clean baseline grayscale threshold, and uses the ratio of the number of pixels to the total number of pixels in the sub-region as the pollution intensity of the sub-region, which is used to characterize the pollution coverage ratio in the sub-region.

[0066] The temperature similarity threshold, grayscale similarity threshold, and cleaning baseline grayscale threshold are set by those skilled in the art based on their own experience.

[0067] The LiDAR 7 outputs real-time 3D point cloud data of the environment surrounding the UAV. The AI ​​control and navigation module performs time synchronization and extrinsic parameter calibration on the point cloud data and fused image data, and projects the point cloud data onto the working plane where the photovoltaic panel is located to generate a local 2D occupied grid map. In the occupied grid map, the edges of the photovoltaic array, the support structure and surrounding obstacles are identified through height abrupt change and continuous boundary extraction algorithms, thereby extracting the effective boundary of the area to be cleaned. On this basis, the LiDAR 7 further measures the relative height between the UAV and the photovoltaic panel and the tilt angle of the UAV relative to the photovoltaic panel plane, and the flight control system outputs the flight speed.

[0068] The AI ​​control and navigation module looks up the corresponding injection parameters in the internally stored injection parameter mapping table based on the pollution intensity, flight speed and relative altitude. These parameters include the target liquid flow rate, target injection pressure and injection mode. The target injection pressure includes liquid injection pressure and gas injection pressure, and the injection mode includes water washing mode, air blowing mode and water-air composite mode.

[0069] The spray direction is corrected based on the tilt angle of the drone relative to the photovoltaic panel plane, so that the spray center axis is aligned as much as possible with the normal of the photovoltaic panel surface. This ensures that the sprayed fluid falls vertically onto the photovoltaic panel surface, avoiding insufficient coverage or liquid film displacement caused by tilting.

[0070] By considering the boundary of the area to be cleaned, the spraying start and stop positions are constrained, and spraying is only activated within the effective area of ​​the photovoltaic panels.

[0071] The flow sensor and pressure sensor inside the cleaning module 9 detect the current actual liquid flow rate and actual injection pressure in real time, and feed the detected values ​​back to the AI ​​control and navigation module. The AI ​​control and navigation module adjusts the opening of the liquid circuit proportional valve, the speed of the pressure stabilizing pump 94, and the opening of the air circuit proportional valve according to the deviation between the target liquid flow rate and the actual liquid flow rate, and the deviation between the target injection pressure and the actual injection pressure, so that the actual flow rate and actual pressure gradually converge to the target values, thereby forming a closed-loop control circuit with flow rate and pressure as feedback quantities.

[0072] In this way, the AI ​​control navigation module uses pollution intensity parameters, relative altitude and flight speed as inputs, target liquid flow rate, target injection pressure and nozzle opening as control outputs, and the detection values ​​of flow sensor and pressure sensor as feedback, forming a closed-loop regulation mechanism. This enables adaptive and stable control of the injection parameters of cleaning module 9 under different pollution levels, different thermal states and different flight conditions.

[0073] After cleaning is completed, the pollution intensity is reassessed. If the pollution intensity is higher than the preset judgment threshold, the route is returned to the previous route, triggering a second cleaning.

[0074] It should be noted that the jet parameter mapping table is a pre-established parameter library used to achieve adaptive jet control under different pollution intensities and flight altitudes. The jet parameter mapping table was obtained through experimental calibration. In a controlled experimental environment, contamination samples covering photovoltaic panel surfaces of various pollution intensities were artificially arranged, and the drone hovered or cruised at different altitudes and flight speeds. Subsequently, the cleaning module 9 was operated to try different combinations of flow rates, pressures, and jet modes, and the cleaning effect was quantitatively evaluated. Based on the experimental results, the optimal jet parameters for each pollution intensity, flight altitude, and cleaning effect were mapped, forming a mapping record, which was then stored in the jet parameter mapping table of the AI ​​control and navigation module.

[0075] The method for correcting liquid flow rate and injection pressure when the lateral wind speed is greater than a preset wind speed threshold includes:

[0076] The heading angle and ground speed vector of the UAV are obtained based on the GPS positioning unit and the IMU inertial measurement unit. The trajectory lateral deviation caused by the wind field is calculated based on the deviation angle between the heading angle and the ground speed vector, and the lateral wind speed component is obtained.

[0077] When the lateral wind speed component exceeds the preset wind speed threshold, it is determined that there is significant lateral wind disturbance in the current flight segment, and the system enters the wind disturbance compensation control state.

[0078] The flow rate compensation and pressure compensation are calculated based on the lateral wind speed component. The flow rate compensation is determined by multiplying the lateral wind speed by a coefficient kw1, and the pressure compensation is determined by multiplying the lateral wind speed by a coefficient kw2, where kw1 and kw2 are pre-calibrated wind compensation coefficients. Then, the flow rate compensation is added to the original target fluid flow rate, and the pressure compensation is added to the original target injection pressure to obtain the corrected target fluid flow rate and the corrected target injection pressure.

[0079] The corrected target fluid flow rate and target injection pressure are sent to the control terminals of the liquid circuit proportional valve, the pressure stabilizing pump 94 and the air circuit proportional valve. By increasing the opening of the liquid circuit proportional valve and increasing the speed of the pressure stabilizing pump 94, the initial injection velocity at the injection outlet is increased synchronously, thereby shortening the flight time of the droplets in the air and reducing the drift distance under the action of crosswind.

[0080] The lateral trajectory compensation is calculated based on the lateral wind speed component. The lateral trajectory compensation is determined by multiplying the lateral wind speed by a coefficient kw3, where kw3 is a preset trajectory compensation coefficient. This lateral trajectory compensation is then superimposed on the lateral position command for the current flight segment, causing the UAV to deviate laterally in the opposite direction of the wind within the current flight segment. By synchronously controlling the corrected target fluid flow rate, target jet pressure, and lateral trajectory compensation, the jet center remains within the area to be cleaned despite lateral wind disturbance, thus achieving real-time compensation for jet trajectory deviation caused by lateral wind disturbance.

[0081] The wind speed threshold is set by those skilled in the art based on their own experience.

[0082] kw1, kw2, and kw3 are wind compensation coefficients, obtained as follows: In a controlled experimental environment, photovoltaic panel samples are arranged, and while flying along a planned route, cross airflow at different wind speeds is generated by a fan; subsequently, the droplet landing point offset and spray dosage changes of the UAV sprayed at each wind speed are recorded; then, by adjusting the target fluid flow rate and target spray pressure, and by slightly offsetting the flight path, the flow rate, pressure, and flight path offset increments that make the spray landing point return to the planned area as much as possible and the dosage uniform are found; finally, the flow rate increment, pressure increment, and flight path offset obtained at each wind speed are linearly fitted with the corresponding wind speed to obtain mapping coefficients kw1, kw2, and kw3, which are used to map the actual cross wind speed to the adjustment range of flow rate, pressure, and flight path offset, thereby achieving cross wind disturbance compensation.

[0083] The UAV 1 is equipped with an automatic return and resupply mechanism. When the battery power or the liquid level of the water tank 4 is lower than the preset threshold, it will automatically execute the return command and complete the battery replacement, liquid replenishment and data upload operations at the resupply station to achieve unattended continuous operation.

[0084] Figure 4 The diagram illustrates the AI-controlled autonomous operation process: First, the system powers on and enters a self-test state, checking battery power, water tank 4 level, sensor online status, and communication links. Then, the LiDAR 7 performs tight-coupled positioning and constructs an environmental map; the infrared vision sensor 3 simultaneously acquires images of the target surface and segments contaminated areas. When the drone 1 reaches the area to be cleaned, the water outlet 8 opens, and the cleaning module 9 starts; the system adaptively adjusts the spray flow and scrubbing parameters based on the drone's speed and relative altitude to achieve uniform cleaning. During operation, the infrared vision sensor 3 evaluates the cleaning effect in real time; if the residual confidence level is higher than a threshold, a local backscan is triggered. When the battery power or water tank 4 level falls below a set threshold, the system automatically plans the shortest path to return for refueling or battery replacement. After the task is completed, the AI ​​control navigation module generates an operation report, including trajectory records, a cleaning heatmap, and energy consumption statistics, facilitating operation and maintenance management.

[0085] With the development of big data and artificial intelligence technologies, the drone cleaning system of this invention can form a self-improving closed-loop system through multi-source data accumulation and online / offline joint optimization of models throughout the entire operation process. Specifically, the AI ​​control and navigation module learns from historical samples such as the surface contamination intensity, wind speed, and wind direction of photovoltaic panels acquired by the infrared vision sensor 3 and the lidar 7, dynamically updating the identification threshold, coverage path, and spray parameter library to achieve personalized cleaning solutions for different environments and photovoltaic bases. This not only improves photovoltaic power generation efficiency, reduces cleaning costs, and enhances operational safety, but also provides strong technical support for the lean operation and maintenance of large-scale photovoltaic power plants, possessing continuously growing market demand and broad application prospects.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multi-mode intelligent cleaning drone system for photovoltaic panels, comprising a drone (1) and a cleaning control unit, wherein the drone (1) is equipped with an infrared vision sensor (3), a lidar (7), and a cleaning module (9), characterized in that, The cleaning module (9) is located at the bottom of the drone (1); The cleaning control unit includes an AI control and navigation module, which is located in the center of the fuselage of the UAV (1) and connected to the flight control bus. It is used to plan the cleaning route and control the operation of the cleaning module (9) to achieve non-contact spray cleaning and accelerated drying of the photovoltaic panel surface.

2. The multi-mode intelligent cleaning drone system for photovoltaic panels according to claim 1, characterized in that, The cleaning module (9) consists of a liquid path, an air path and a mixing spray chamber, a water outlet (8), a liquid storage filter (91), a cooling fan (92), and an electromagnetic vaporization device (93). The liquid path is equipped with a pressure stabilizing pump (94), a liquid path solenoid valve, a flow sensor, a liquid pressure sensor, and a liquid path proportional valve. The air path is equipped with a gas storage chamber, a gas path solenoid valve, a gas pressure sensor, and a gas path proportional valve. The liquid path and the air path are mixed in the mixing spray chamber by a venturi structure and then sprayed out by an electrically controlled nozzle.

3. A multi-mode intelligent cleaning drone system suitable for photovoltaic panels according to claim 2, characterized in that, The liquid line is provided with a liquid transmission interface, and the gas line is provided with a gas transmission interface. Both the liquid transmission interface and the gas transmission interface are provided with a self-sealing structure.

4. The multi-mode intelligent cleaning drone system for photovoltaic panels according to claim 1, characterized in that: The UAV (1) is equipped with a power rotor assembly (2), which consists of a brushless motor, blades and an electronic speed controller. The brushless motor is fixed on the UAV (1), and the output shaft of the brushless motor is connected to the blade drive through the electronic speed controller to drive the blade to rotate.

5. A multi-mode intelligent cleaning drone system suitable for photovoltaic panels according to claim 1, characterized in that, The UAV (1) has a water tank (4) fixed in the middle of its belly by a vibration damping bracket. The water tank (4) is sealed to one end of a pressure-resistant hose, and the other end of the pressure-resistant hose is connected to a liquid storage filter (91). The outlet of the liquid storage filter (91) is connected to the liquid path, and the other end of the liquid path is connected to the outlet (8), together forming a closed fluid transmission circuit.

6. A multi-mode intelligent cleaning drone system suitable for photovoltaic panels according to claim 1, characterized in that, The AI ​​control and navigation module includes a GPS positioning unit, an IMU inertial measurement unit, a central processing unit, and a data fusion unit; The GPS positioning unit is used to obtain the position, relative altitude and flight speed of the UAV (1); The IMU (Inertial Measurement Unit) is used to acquire the tilt angle of the aircraft. The data fusion unit is used to fuse the visible light image of the photovoltaic panel surface collected by the infrared vision sensor (3) and the infrared temperature matrix, as well as the three-dimensional point cloud data collected by the lidar (7). The central processing unit, based on the processing results of the data fusion unit, combined with the UAV's flight speed, relative altitude, and body tilt angle, and according to the pollution intensity on the photovoltaic panel surface, performs closed-loop adjustment of the liquid flow rate, injection pressure, and proportional valve opening of the cleaning module (9), and corrects the liquid flow rate and injection pressure when the lateral wind speed is greater than the preset wind speed threshold.

7. A multi-mode intelligent cleaning drone system suitable for photovoltaic panels according to claim 6, characterized in that, The method for closed-loop regulation of the liquid flow rate, injection pressure, and proportional valve opening of the cleaning module (9) includes: When the drone (1) flies over the area to be cleaned, it collects visible light images and corresponding infrared temperature matrices of the photovoltaic panel surface, denoises and synchronizes the images and temperature data and forms a fused image, performs region segmentation to obtain sub-regions and calculates the pollution intensity of the sub-regions; A local grid map is generated based on the three-dimensional point cloud data collected by lidar (7), and the relative height, body tilt angle, and flight speed are obtained. Based on pollution intensity, flight speed and relative altitude, the target liquid flow rate, target injection pressure and injection pattern are determined by consulting the injection parameter mapping table. Obtain the current actual liquid flow rate and actual injection pressure, adjust the opening of the liquid circuit proportional valve, the speed of the pressure stabilizing pump (94) and the opening of the gas circuit proportional valve, so that the actual injection flow rate and pressure gradually reach the target value.

8. A multi-mode intelligent cleaning drone system suitable for photovoltaic panels according to claim 7, characterized in that, The method for performing region segmentation to obtain sub-regions includes: Real-time acquisition of visible light grayscale images of the photovoltaic panel surface and infrared temperature distribution matrix corresponding to the grayscale images, where each pixel corresponds to both a grayscale value and a temperature value. Median filtering and grayscale normalization are performed on the grayscale image and the infrared temperature matrix respectively. The two types of images are synchronized in time based on the timestamp, so that grayscale pixels and temperature pixels acquired at the same time form fused pixels to merge the image. For each pixel, a feature vector containing temperature and grayscale values ​​is constructed. When the temperature difference between adjacent pixels is less than a preset temperature similarity threshold and the grayscale difference is less than a preset grayscale similarity threshold, the adjacent pixels are merged into the same sub-region. The above merging process is repeated until there are no more adjacent pixels that can be merged, thereby obtaining multiple sub-regions.

9. A multi-mode intelligent cleaning drone system suitable for photovoltaic panels according to claim 6, characterized in that, The method for correcting liquid flow rate and injection pressure when the lateral wind speed is greater than a preset wind speed threshold includes: The heading angle and ground speed vector of the UAV (1) are obtained based on the GPS positioning unit and the IMU inertial measurement unit. The lateral deviation of the flight path caused by the wind field is calculated based on the deviation angle between the heading angle and the ground speed vector, and the lateral wind speed component is obtained. When the lateral wind speed component exceeds the preset wind speed threshold, it is determined that there is significant lateral wind disturbance in the current flight segment, and the system enters the wind disturbance compensation control state.

10. A multi-mode intelligent cleaning drone system suitable for photovoltaic panels according to claim 9, characterized in that, The method for correcting liquid flow rate and injection pressure when the lateral wind speed is greater than a preset wind speed threshold also includes: The flow compensation and pressure compensation are calculated based on the crosswind speed component. The flow compensation is determined by multiplying the crosswind speed by a coefficient kw1, and the pressure compensation is determined by multiplying the crosswind speed by a coefficient kw2. The flow compensation is added to the original target fluid flow rate, and the pressure compensation is added to the original target injection pressure to obtain the corrected target fluid flow rate and the corrected target injection pressure. The corrected target fluid flow rate and target injection pressure are sent to the control terminals of the liquid circuit proportional valve, the pressure stabilizing pump (94) and the gas circuit proportional valve. By increasing the opening of the liquid circuit proportional valve and increasing the speed of the pressure stabilizing pump (94), the initial injection velocity at the injection outlet is increased synchronously, thereby shortening the flight time of the droplets in the air and reducing the drift distance under the action of cross wind. The lateral trajectory compensation is calculated based on the lateral wind speed component. The lateral trajectory compensation is determined by multiplying the lateral wind speed by a coefficient kw3. The lateral trajectory compensation is superimposed on the lateral position command of the current flight segment, causing the UAV (1) to generate a reverse lateral offset along the direction of the wind in the current flight segment.

Images