Air-ground cooperative photovoltaic panel intelligent snow removal system

Abstract

The invention discloses an air-ground cooperative photovoltaic panel intelligent snow removal system, and relates to the technical field of intelligent control. A central intelligent scheduling management subsystem is used for receiving operation and maintenance requirements of a photovoltaic power station, planning a routing inspection path, receiving a routing inspection result, analyzing the routing inspection result, and generating an optimal snow removal scheme; the parking station unmanned aerial vehicle subsystem is used for collecting surface data of the photovoltaic panel according to the inspection path and analyzing the surface data of the photovoltaic panel by adopting an accumulated snow automatic identification model to obtain an inspection result; and the ground operation vehicle subsystem is used for cleaning the snow accumulation area according to the optimal snow removal scheme and uploading a cleaning result to the central intelligent scheduling management subsystem. The problems that an existing photovoltaic panel snow removal technology is low in efficiency, poor in adaptability and lack of air-ground cooperative linkage are solved, whole-process intelligentization of routing inspection, diagnosis, sweeping and dispatching is achieved, and the operation and maintenance efficiency and power generation stability of a photovoltaic power station under extreme climate and complex terrain are improved.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic panel snow removal technology, and more specifically to a smart photovoltaic panel snow removal system that coordinates air and ground operations. Background Technology

[0002] In recent years, the global photovoltaic (PV) power generation industry has expanded rapidly, and PV power generation has become an important energy form for achieving green and low-carbon development. However, with the frequent occurrence of extreme weather events (such as blizzards and freezing rain), the problem of snow accumulation and ice buildup on PV panels has become increasingly prominent. Snow and ice layers can severely obstruct the light-receiving surface of PV panels, leading to a sharp drop in power generation efficiency. At the same time, snow loads may damage the PV panel support structure, affecting the service life of the equipment. In addition, long-term accumulation of dust on the surface of PV panels will also continuously reduce power generation efficiency, further exacerbating energy waste and economic losses.

[0003] Currently, photovoltaic panel cleaning and snow removal technologies are mainly divided into two categories: one is traditional manual cleaning technology, which relies on manual hand tools. This is not only inefficient and costly, but also poses safety hazards in high-altitude and steep terrain, making it difficult to meet the needs of large-scale photovoltaic power plants. The other is mechanized automatic cleaning technology, such as track-mounted sweeping robots and fixed spraying systems. Although these technologies achieve partial automation, they are limited by fixed tracks or terrain, making them unable to flexibly cope with complex terrain (such as mountain photovoltaic power plants). Furthermore, they are prone to jamming and malfunctions under extreme conditions such as excessive snow accumulation and icing, resulting in poor operational stability.

[0004] Therefore, in view of the shortcomings of existing technologies, how to provide a smart snow removal system for photovoltaic panels that combines air and ground is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides an intelligent snow removal system for photovoltaic panels that integrates air and ground, which solves the problems of low efficiency, poor adaptability and lack of air-ground coordination in existing photovoltaic panel snow removal technologies. It realizes intelligent operation and maintenance of the entire process of inspection, diagnosis, cleaning and scheduling, and improves the operation and maintenance efficiency and power generation stability of photovoltaic power plants under extreme climate and complex terrain.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an air-ground coordinated photovoltaic panel intelligent snow removal system, comprising: a ground operation vehicle subsystem, a stationed drone subsystem, and a central intelligent dispatch and management subsystem; The central intelligent dispatch and management subsystem transmits data with the ground operation vehicle subsystem and the stationed drone subsystem, respectively. The central intelligent dispatch and management subsystem is used to receive photovoltaic power plant operation and maintenance requirements, plan inspection routes, receive inspection results, analyze the inspection results, and generate the optimal snow removal plan. The stationed drone subsystem is used to collect surface data of photovoltaic panels according to the inspection path, and to analyze the surface data of photovoltaic panels using an automatic snow recognition model to obtain inspection results; The ground operation vehicle subsystem is used to clear snow-covered areas according to the optimal snow removal plan and upload the clearing results to the central intelligent dispatch and management subsystem.

[0007] Preferably, the stationed UAV subsystem includes: a multi-source sensor inspection module, an intelligent diagnostic module, and a collaborative communication module; The multi-source sensing inspection module uses an infrared thermal imager, an image acquisition sensor, and a lidar to simultaneously collect data on snow thickness, ice distribution, and surface contamination on the photovoltaic panel surface. The intelligent diagnostic module includes a data preprocessing unit, a feature extraction unit, an automatic snow accumulation recognition model, and a result generation unit connected in sequence. The data preprocessing unit is used to preprocess the data collected by the multi-source sensor inspection module; The feature extraction unit is used to extract snow / ice layer features and surface contamination features based on the preprocessed data; The automatic snow recognition model uses an improved YOLOv8 deep learning network, which adds an attention mechanism to the YOLOv8 deep learning network structure to identify snow, ice and polluted areas. The result generation unit uses a non-maximum suppression algorithm to remove overlapping recognition regions and generate inspection results.

[0008] Preferably, the collaborative communication module transmits data with the central intelligent dispatch and management subsystem and the ground operation vehicle subsystem respectively; The collaborative communication module adopts a dual-mode communication network architecture; the main communication link adopts 5G private network communication based on the 5G micro base stations deployed in the photovoltaic power station. The backup communication link uses LoRa wide area network communication.

[0009] Preferably, the central intelligent scheduling and management subsystem includes an intelligent scheduling and task allocation module and a visual operation and maintenance and prediction module; The intelligent scheduling and task allocation module plans the drone inspection path according to the operation and maintenance needs of the photovoltaic power station and the distribution of the photovoltaic panel array. After receiving the inspection results transmitted by the stationed drone subsystem, a multi-objective optimization model is constructed by combining data on snow coverage, snow thickness, ice distribution range and pollution level, and the optimal snow removal plan is generated based on the operation data of the ground operation vehicle. The visualization operation and maintenance and prediction module is used to display the operation and maintenance management interface and predict the equipment status of drones and ground operation vehicles.

[0010] Preferably, based on the distribution of the photovoltaic panel array, a partition traversal algorithm is used to divide the entire inspection area into several continuous sub-regions. At the same time, combined with the constraints of the UAV's endurance and sensor acquisition efficiency, the shortest path and the optimal inspection sequence are planned.

[0011] Preferably, the ground operation vehicle subsystem includes: a driving navigation module and a multi-functional sweeping module; the driving navigation module includes a positioning unit and a path tracking control unit; The positioning unit is used for path navigation and real-time calibration of the distance between the work vehicle and the photovoltaic panel; The path tracking control unit receives the snow removal path from the central intelligent scheduling and management subsystem and dynamically adjusts the driving parameters based on the real-time location, speed and steering angle of the work vehicle. The multi-functional cleaning module is used to select different cleaning components according to the snow removal plan to complete the cleaning work.

[0012] Preferably, the cleaning assembly includes: a rotary brush cleaning assembly, a scraper de-icing assembly, and a hot air snow melting assembly.

[0013] Preferably, the ground operation vehicle subsystem further includes a status monitoring module, which is used to collect the operation data of the operation vehicle in real time during the sweeping operation and to determine the malfunction of the operation vehicle based on the operation data.

[0014] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a smart snow removal system for photovoltaic panels that integrates air and ground, which solves the problems of low efficiency, poor adaptability and lack of air-ground coordination in existing photovoltaic panel snow removal technology, realizes intelligent operation and maintenance of the entire process of inspection, diagnosis, cleaning and scheduling, and improves the operation and maintenance efficiency and power generation stability of photovoltaic power stations under extreme climate and complex terrain. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of a smart snow removal system using photovoltaic panels that integrates air and ground operations, as provided by the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This invention discloses an air-ground coordinated photovoltaic panel smart snow removal system, such as... Figure 1 As shown, the central intelligent dispatch and management subsystem transmits data with the ground operation vehicle subsystem and the stationed drone subsystem, respectively; The central intelligent dispatch and management subsystem is used to receive photovoltaic power plant operation and maintenance requirements, plan inspection routes, receive inspection results, analyze the inspection results, and generate the optimal snow removal plan. The stationed drone subsystem is used to collect surface data of photovoltaic panels according to the inspection path, and to analyze the surface data of photovoltaic panels using an automatic snow recognition model to obtain inspection results; The ground operation vehicle subsystem is used to clear snow-covered areas according to the optimal snow removal plan and upload the clearing results to the central intelligent dispatch and management subsystem.

[0019] Specifically, the station-based UAV subsystem includes: a multi-source sensor inspection module, an intelligent diagnostic module, and a collaborative communication module; The multi-source sensing inspection module uses an infrared thermal imager, an image acquisition sensor, and a lidar to simultaneously collect data on snow thickness, ice distribution, and surface contamination on the photovoltaic panel surface; the image acquisition sensor uses a high-definition visible light camera. The infrared thermal imager uses a high-sensitivity infrared detector to accurately capture temperature difference areas on the surface of photovoltaic panels caused by snow and ice, and indirectly invert the snow thickness and ice distribution.

[0020] The image acquisition sensor uses an industrial-grade visible light camera to clearly capture the surface texture details of the photovoltaic panel, such as the uniformity of snow cover and the patchy distribution of surface dust accumulation, providing visual basis for subsequent pollution level determination.

[0021] The lidar uses a mechanical rotating lidar to acquire three-dimensional point cloud data of the photovoltaic panel surface in real time. It distinguishes snow-covered areas by the difference in point cloud density and directly calculates the snow thickness, which is the vertical distance between the point cloud surface and the photovoltaic panel reference plane.

[0022] By employing timestamp synchronization technology, the data collected by the three types of devices are bound to a unified timestamp, ensuring that the infrared temperature data, visible light image data and lidar point cloud data are accurately matched at the same time and in the same area, thus avoiding analysis errors caused by asynchronous data.

[0023] Based on the distribution of the photovoltaic power station array, the inspection area is divided into several sub-areas. The drone traverses each sub-area in turn, staying in each sub-area for 2 to 3 seconds, and simultaneously triggering three types of sensors to collect data, thereby improving inspection efficiency.

[0024] The intelligent diagnostic module includes a data preprocessing unit, a feature extraction unit, an automatic snow accumulation recognition model, and a result generation unit connected in sequence. The data preprocessing unit is used to preprocess the data collected by the multi-source sensor inspection module; Specifically, the infrared thermal imaging data is denoised: a median filtering algorithm is used to eliminate environmental noise (such as image jitter caused by wind), and the original infrared image is converted into a standardized temperature field distribution map through temperature calibration correction to highlight the temperature difference between snow, ice and the photovoltaic panel itself.

[0025] The visible light image is enhanced by histogram equalization to improve image contrast, Gaussian blur algorithm to suppress image noise, and image cropping to remove invalid edge areas (such as the photovoltaic panel border and ground background), retaining only the image data of the effective light-receiving area of ​​the photovoltaic panel.

[0026] Denoising and registration of lidar point cloud data: Statistical filtering algorithms are used to remove outliers (such as false point clouds caused by birds or dust interference), and the point cloud data is registered with the preset 3D model of the photovoltaic power station to determine the position coordinates of the photovoltaic panel corresponding to each point cloud, providing a basis for subsequent positioning.

[0027] The feature extraction unit is used to extract snow / ice layer features and surface contamination features based on the preprocessed data.

[0028] Specifically, snow / ice layer feature extraction: extract the outline and area of ​​the low-temperature region from the infrared temperature field distribution map, calculate the proportion of the low-temperature region to the area of ​​a single photovoltaic panel, and obtain the snow coverage; extract the average point cloud height of the snow-covered region from the lidar point cloud data as a quantitative indicator of snow thickness. For example, when the average point cloud height is ≥5cm, it is judged as moderate snow cover, and when it is ≥15cm, it is judged as heavy snow cover.

[0029] Surface contamination feature extraction: From the preprocessed visible light image, the gray-level co-occurrence matrix algorithm is used to extract image texture features. Due to dust accumulation, the texture contrast of the contaminated area is 20%~30% lower than that of the clean area. At the same time, the pixel ratio of the contaminated area is calculated, and the contamination level is divided into light (contamination ratio ≤10%), moderate (10% < contamination ratio ≤30%), and heavy (contamination ratio >30%).

[0030] The automatic snow recognition model uses an improved YOLOv8 deep learning network, which adds an attention mechanism to the YOLOv8 deep learning network structure to identify snow, ice and polluted areas. An improved YOLOv8 deep learning model was selected as the core recognition framework. An attention mechanism module was added to the original model to enhance the ability to recognize small features such as snow edges and ice cracks.

[0031] Specifically, a multi-class labeled dataset was constructed: multi-source data of photovoltaic panels were collected under different climatic conditions (such as 1 hour after snowfall, 6 hours after snowfall, and icing conditions), and labeled into four categories: snow-covered areas, ice-covered areas, polluted areas, and normal areas. A total of 100,000+ samples were labeled and divided into training set, validation set, and test set in a ratio of 7:2:1.

[0032] Model training optimization: The stochastic gradient descent (SGD) optimizer was used, with an initial learning rate of 0.01, which was decayed to 1 / 10 of the original rate every 5 epochs. The training was conducted for 50 epochs, with the cross-entropy loss function as the training objective. Training was stopped when the recognition accuracy on the validation set remained stable above 97% for 3 consecutive epochs. The final model achieved recognition accuracies of 98.2% for snow, 97.5% for ice, and 96.8% for pollution.

[0033] The non-maximum suppression (NMS) algorithm is used to remove overlapping identification regions and generate inspection results. It is ensured that the same area is labeled as only one type (e.g., when snow-covered areas overlap with polluted areas, they are preferentially labeled as snow-covered areas).

[0034] The inspection results include the location coordinates of each photovoltaic panel (based on lidar registration results), snow coverage, maximum / average snow thickness, ice distribution range, and pollution level. Areas with severe, moderate, and mild problems are marked in red, yellow, and blue, respectively, and the inspection results are transmitted to the central intelligent dispatch and management subsystem.

[0035] Specifically, the collaborative communication module transmits data with the central intelligent dispatch and management subsystem and the ground operation vehicle subsystem, respectively; The collaborative communication module undertakes the data transmission tasks between the station-based UAV subsystem and the central intelligent dispatch and management subsystem, as well as the ground operation vehicle subsystem, ensuring the stability, real-time performance, and reliability of data transmission. The specific technical solution is as follows: The collaborative communication module adopts a dual-mode communication network architecture; the main communication link adopts 5G private network communication based on the 5G micro base station deployed in the photovoltaic power station; it realizes high-speed data transmission between the stationed drone subsystem and the central intelligent dispatch and management subsystem, meets the real-time backhaul requirements of lidar point cloud data, and controls the transmission delay within 50ms. The backup communication link uses LoRa low-power wide-area network communication. As a redundant backup for 5G communication, it is suitable for emergency data transmission when 5G signals are blocked (such as complex terrain or extreme weather causing signal attenuation), ensuring that critical diagnostic results (such as the location of heavily snow-covered areas) are not lost.

[0036] A priority queue mechanism is used to divide the transmitted data into three levels: Level 1 data: Location data of heavily snow-covered and ice-covered areas generated by the intelligent diagnostic module. Priority is given to ensuring bandwidth usage, so that the central intelligent scheduling and management subsystem can obtain information on high-urgency tasks as soon as possible.

[0037] Secondary data: Diagnostic results for areas with moderate snow cover and pollution, and real-time status data of UAVs (such as battery level and flight attitude), are transmitted every 5 seconds to support scheduling decisions and equipment monitoring.

[0038] Level 3 data: Complete infrared images, visible light images, and raw LiDAR point cloud data are uploaded in batches during the idle time of Level 1 and Level 2 data transmission for subsequent data tracing and model optimization.

[0039] Real-time monitoring of communication link status: By periodically sending heartbeat packets (once every 1 second), it is determined whether the communication between the UAV and the central intelligent dispatch and management subsystem is normal. If no heartbeat packet reply is received for 3 consecutive times, it is determined that the communication is interrupted.

[0040] After communication is interrupted, the system automatically switches from the 5G primary link to the LoRa backup link and sends a link switching alarm to the central intelligent scheduling and management subsystem. Once the 5G signal is restored, the system automatically switches back to the primary link. The switching process does not affect the transmission of data with secondary or higher priority.

[0041] For the three levels of raw data, a fragmented transmission and verification mechanism is adopted. An MD5 check is performed every 10MB of data transmitted. If a fragment transmission fails, only the previously failed fragment will be retransmitted after the next reconnection, thus avoiding bandwidth waste caused by repeated transmission.

[0042] It receives sweeping progress data from ground operation vehicles and uploads it synchronously to the central intelligent dispatch and management subsystem, providing a basis for the central intelligent dispatch and management subsystem to determine whether additional equipment needs to be dispatched or the operation sequence needs to be adjusted.

[0043] Specifically, the central intelligent scheduling and management subsystem includes an intelligent scheduling and task allocation module and a visual operation and maintenance and prediction module; The intelligent scheduling and task allocation module plans the drone inspection path according to the operation and maintenance needs of the photovoltaic power station and the distribution of the photovoltaic panel array. The intelligent scheduling and task allocation module is guided by the operation and maintenance needs of photovoltaic power plants, realizing intelligent scheduling of the entire process of inspection path planning and snow removal task optimization allocation.

[0044] After receiving the inspection results transmitted by the stationed drone subsystem, a multi-objective optimization model is constructed by combining data on snow coverage, snow thickness, ice distribution range and pollution level, and the optimal snow removal plan is generated based on the operation data of the ground operation vehicle. Specifically, upon receiving the inspection results transmitted by the stationed drone subsystem, the module first performs in-depth analysis of the results. Combining indicators such as snow cover, snow thickness, ice distribution, and pollution level, a multi-objective optimization model is constructed. This model comprehensively considers factors such as the snow removal efficiency of ground vehicles (e.g., operating speed under different snow thicknesses), equipment distribution location, and remaining power to generate the optimal snow removal plan. During the plan formulation process, snow removal resources are prioritized for areas with heavy snow cover (marked in red) and ice coverage, while avoiding overlapping paths for operating vehicles to reduce scheduling conflicts.

[0045] The visualization operation and maintenance and prediction module is used to display the operation and maintenance management interface and predict the equipment status of drones and ground operation vehicles.

[0046] The visualized operation and maintenance (O&M) and prediction module provides photovoltaic power plant O&M personnel with an intuitive and comprehensive O&M management interface, enabling the prediction of snow accumulation risks and equipment status, thus improving the foresight and convenience of O&M work. In terms of visualization, the module is based on Geographic Information System (GIS) technology to construct a 3D visualization model of the photovoltaic power plant. It dynamically overlays information such as problem areas (marked with red, yellow, and blue for severe, moderate, and minor problems respectively), the real-time location and progress of ground operation vehicles, and the drone inspection trajectory onto the 3D model. O&M personnel can intuitively view the status of each area through the interface and quickly locate the problem.

[0047] Meanwhile, the visualization operation and prediction module performs health assessments on the equipment status of ground operation vehicles and drones. By analyzing equipment operating parameters (such as the engine condition of the operation vehicle and the battery cycle count of the drone), it predicts potential equipment failure risks and promptly reminds operation and maintenance personnel to carry out inspections and maintenance, ensuring the long-term stable operation of the system.

[0048] Specifically, based on the distribution of the photovoltaic panel array, a partitioning traversal algorithm is used to divide the entire inspection area into several continuous sub-regions. Simultaneously, considering constraints such as the drone's endurance and sensor acquisition efficiency, the shortest path and optimal inspection sequence are planned. This ensures that the drone completes full-coverage inspection in the shortest possible time, reducing unnecessary flight losses.

[0049] Specifically, the ground operation vehicle subsystem includes a driving navigation module and a multi-functional sweeping module; ensuring the precise driving of the operation vehicle between photovoltaic panel arrays and avoiding collisions with photovoltaic panel supports or deviation from the work area.

[0050] The driving navigation module includes a positioning unit and a path tracking control unit; The positioning unit is used for path navigation and real-time calibration of the distance between the work vehicle and the photovoltaic panel; the positioning unit integrates Beidou positioning and UWB positioning, with Beidou positioning used for global path navigation and UWB positioning used for real-time calibration of the distance between the work vehicle and the photovoltaic panel. The path tracking control unit receives the snow removal path from the central intelligent scheduling and management subsystem and dynamically adjusts the driving parameters based on the real-time location, speed and steering angle of the work vehicle. After receiving the snow removal path from the central intelligent scheduling and management subsystem, the path tracking control unit uses the model predictive control (MPC) algorithm to dynamically adjust the driving parameters based on the real-time position, speed and steering angle of the work vehicle fed back by the positioning unit, so as to ensure that the deviation of the work vehicle along the planned path is minimized and adapts to the dense layout of the photovoltaic panel array.

[0051] The multi-functional cleaning module is used to select different cleaning components according to the snow removal plan to complete the cleaning work.

[0052] Specifically, the cleaning components include: a rotary brush cleaning component, a scraper de-icing component, and a hot air snow melting component.

[0053] Cleaning solutions were designed to address varying snow thicknesses, ice conditions, and photovoltaic panel surface characteristics, including: The sweeping unit integrates a rotary brush sweeping component, a scraper de-icing component, and a hot air snow melting component. These three components can automatically switch or operate in combination according to the snow removal plan. Rotary brush cleaning component: It uses flexible nylon bristles to avoid scratching the surface of the photovoltaic panel. The brush speed can be adjusted according to different snow thicknesses. It uses centrifugal force to throw loose snow into the snow collection troughs on both sides of the work vehicle.

[0054] Scraper de-icing assembly: It adopts a polyurethane elastic scraper that fits the surface of the photovoltaic panel without damaging the coating layer. The scraper angle is adjusted by a hydraulic system. Different scraper angles are set according to different ice thicknesses. It works with a vibrating motor to break the ice layer and push it to the snow collection trough.

[0055] Hot air snow melting component: When the ice layer thickness is >5cm or the snow is frozen into a hard block, it is activated by electric heating tubes to generate hot air. The hot air is blown evenly onto the ice and snow surface through the air guide hood (the width of the air outlet matches the width of the photovoltaic panel). The melted water is discharged through the guide groove on the edge of the photovoltaic panel to prevent the water from freezing again.

[0056] Specifically, the ground operation vehicle subsystem also includes a status monitoring module, which is used to collect the operation data of the operation vehicle in real time during the sweeping operation and to determine the malfunction of the operation vehicle based on the operation data.

[0057] Specifically, the status monitoring module collects real-time operating data of the work vehicle to provide early warning of equipment failures and feedback on work progress, including: Operating parameter monitoring unit: Collects parameters such as engine speed, battery level, cleaning component operating status, and fuel / electricity consumption via CAN bus, and uploads the data to the central intelligent dispatch and management subsystem in real time.

[0058] Fault diagnosis unit: It adopts the fault tree analysis (FTA) algorithm to identify common faults based on the monitoring parameters of the work vehicle, such as rotating card swiping lag, scraper wear, and positioning signal loss. When abnormal data is detected, it immediately triggers local audible and visual alarms, and the buzzer and warning light flash red. At the same time, it sends fault codes and location information to the central intelligent dispatch and management subsystem.

[0059] Work progress feedback unit: The positioning unit records the number of photovoltaic panels that the work vehicle has cleaned. Each time a photovoltaic panel is cleaned, it is marked as completed. Combined with the total task volume issued by the central intelligent dispatch and management subsystem, the work progress is calculated and updated every 5 minutes. The progress is displayed simultaneously on the work vehicle's cab display screen and the central intelligent dispatch and management subsystem's visualization interface.

[0060] In one specific embodiment of the present invention, the complete operation flow of an air-ground coordinated photovoltaic panel smart snow removal system is as follows: Photovoltaic power plant operation and maintenance personnel can manually initiate snow removal requests through the visual interface of the central intelligent dispatch and management subsystem, select the target area, and set a completion deadline. Task priorities are divided into emergency (current snow thickness ≥15cm or ice), normal (snow thickness 5-15cm), and low priority (snow thickness <5cm).

[0061] After receiving the request, the central intelligent scheduling and task allocation module uses a partition traversal algorithm to plan the inspection path based on the distribution of the photovoltaic panel array. For example, if 100 photovoltaic panels are divided into 10 sub-regions with 10 panels in each sub-region, the drone will inspect them in the order of sub-region 1 → sub-region 2 → … → sub-region 10, and at the same time, the inspection parameters will be issued and the sensors will collect data synchronously.

[0062] After the stationed drone subsystem takes off, the multi-source sensor inspection module collects data according to the inspection parameters, and the intelligent diagnosis module completes data preprocessing, feature extraction and snow identification to generate inspection results (including snow thickness, coverage and pollution level of each photovoltaic panel). The results are then uploaded to the central intelligent dispatch and management subsystem through the collaborative communication module (prioritizing 5G private network).

[0063] After receiving the inspection results, the intelligent scheduling and task allocation module of the central intelligent scheduling and management subsystem uses a multi-objective optimization model. The objective functions include maximizing snow removal efficiency (minimizing the total travel distance of the work vehicles), minimizing equipment wear (prioritizing the scheduling of work vehicles closest to the target area with sufficient remaining power), and prioritizing tasks (allocating resources to emergency tasks first). The model is solved using a genetic algorithm to generate the optimal snow removal plan, including: work vehicle allocation, the travel path of each work vehicle (avoiding already assigned areas and avoiding intersections), and the sweeping mode. For example, in sub-area 3 with a snow thickness of 18cm, a rotary brush and hot air snow melting mode are specified.

[0064] The central intelligent dispatch and management subsystem distributes snow removal plans to the corresponding ground operation vehicles. After receiving the route, the driving navigation module initiates precise driving, and the multi-functional sweeping module carries out snow removal operations according to the specified mode. During the operation, the status monitoring module uploads operating parameters and operation progress in real time. If the drone detects an increase in snow thickness in a certain area, it immediately updates the inspection results and uploads them to the central intelligent dispatch and management subsystem. The central intelligent dispatch and management subsystem then re-optimizes the plan (e.g., adjusts the sweeping mode of that area to "rotary brush + high-power hot air") and distributes it to the operation vehicles, achieving dynamic adjustment.

[0065] After the ground vehicles complete their snow removal tasks, the central intelligent dispatch and management subsystem dispatches drones to conduct a secondary inspection of the target area. If the secondary inspection shows that the snow thickness is ≤2cm and there is no ice residue, the inspection is deemed successful. If there are any areas that have not been cleared, a supplementary snow removal task is issued. At the same time, the visualization operation and maintenance and prediction module generates a snow removal report, including total time, equipment energy consumption, fault records, and snow removal rate, for subsequent operation and maintenance optimization.

[0066] The stationed drone subsystem is equipped with a backup drone. When a drone malfunctions (such as running out of battery or damaged sensors), the central intelligent dispatch and management subsystem automatically dispatches the backup drone to take over its inspection task, ensuring that the inspection is not interrupted.

[0067] If a work vehicle malfunctions (such as engine stalling), the central intelligent dispatch and management subsystem will real-time assign its unfinished tasks to other idle work vehicles, or adjust the routes of the remaining work vehicles to supplement the work area of ​​the malfunctioning vehicle, ensuring that the overall snow removal task is completed on time.

[0068] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0069] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An aerial-ground cooperative photovoltaic panel intelligent snow removal system, characterized in that, The system comprises a ground operation vehicle subsystem, a station-based unmanned aerial vehicle subsystem, and a central intelligent scheduling and management subsystem. The central intelligent scheduling and management subsystem is configured to transmit data to the ground operation vehicle subsystem and the station-based unmanned aerial vehicle subsystem. The central intelligent scheduling and management subsystem is configured to receive photovoltaic power station operation and maintenance requirements, plan an inspection path, receive inspection results, analyze the inspection results, and generate an optimal snow removal plan. The station-based unmanned aerial vehicle subsystem is configured to collect photovoltaic panel surface data according to the inspection path and analyze the photovoltaic panel surface data using a snow automatic identification model to obtain the inspection results. The ground operation vehicle subsystem is configured to clean the snow-covered area according to the optimal snow removal plan and upload the cleaning results to the central intelligent scheduling and management subsystem. The station-based unmanned aerial vehicle subsystem comprises a multi-source sensing inspection module, an intelligent diagnosis module, and a cooperative communication module.

2. The air-ground coordinated photovoltaic panel intelligent snow removal system according to claim 1, characterized in that, The multi-source sensing inspection module comprises an infrared thermal imager, an image acquisition sensor, and a laser radar, and is configured to synchronously collect snow thickness, ice layer distribution, and surface pollution data on the photovoltaic panel surface. The intelligent diagnosis module comprises a data preprocessing unit, a feature extraction unit, a snow automatic identification model, and a result generation unit connected in sequence. The data preprocessing unit is configured to preprocess the data collected by the multi-source sensing inspection module. The feature extraction unit is configured to extract snow / ice layer features and surface pollution features from the preprocessed data. The snow automatic identification model uses an improved YOLOv8 deep learning network, and adds an attention mechanism based on the YOLOv8 deep learning network structure to identify snow, ice layer, and pollution areas. The result generation unit uses a non-maximum suppression algorithm to remove overlapping identified areas and generates the inspection results. The cooperative communication module is configured to transmit data to the central intelligent scheduling and management subsystem and the ground operation vehicle subsystem.

3. The air-ground coordinated photovoltaic panel intelligent snow removal system according to claim 2, characterized in that, The cooperative communication module uses a dual-mode communication network architecture, wherein a main communication link uses 5G private network communication based on a 5G micro base station deployed in the photovoltaic power station, and a backup communication link uses LoRa wide area network communication. The central intelligent scheduling and management subsystem comprises an intelligent scheduling and task allocation module and a visual operation and maintenance and prediction module.

4. The air-ground coordinated photovoltaic panel intelligent snow removal system according to claim 1, characterized in that, The intelligent scheduling and task allocation module plans an unmanned aerial vehicle inspection path based on photovoltaic power station operation and maintenance requirements and photovoltaic panel array distribution. After receiving the inspection results transmitted by the station-based unmanned aerial vehicle subsystem, a multi-objective optimization model is constructed based on snow coverage, snow thickness, ice layer distribution range, and pollution level data, and an optimal snow removal plan is generated based on ground operation vehicle operation data. The visual operation and maintenance and prediction module is configured to display an operation and maintenance interface and predict the equipment state of the unmanned aerial vehicle and the ground operation vehicle. Further comprising:

5. The air-ground coordinated photovoltaic panel intelligent snow removal system according to claim 4, characterized in that, Based on the distribution of the photovoltaic panel array, the entire inspection area is divided into a plurality of continuous sub-areas using a partition traversal algorithm, and a shortest path and an optimal inspection order are planned based on the constraints of the unmanned aerial vehicle endurance and sensor collection efficiency. The ground operation vehicle subsystem comprises a driving navigation module and a multifunctional cleaning module.

6. The air-ground coordinated photovoltaic panel intelligent snow removal system according to claim 1, characterized in that, ​ The travel navigation module comprises a positioning unit and a path tracking control unit; The positioning unit is used for path navigation and real-time calibration of the distance between the working vehicle and the photovoltaic panel; The path tracking control unit receives the snow removal path of the central intelligent scheduling management subsystem, and dynamically adjusts the travel parameters according to the real-time position, travel speed and steering angle of the working vehicle; The multifunctional cleaning module is used for selecting different cleaning components according to the snow removal scheme to complete the cleaning work.

7. The air-ground coordinated photovoltaic panel intelligent snow removal system according to claim 6, characterized in that, The cleaning component comprises a rotary brush cleaning component, a scraper deicing component and a hot air snow melting component.

8. The air-ground coordinated photovoltaic panel intelligent snow removal system according to claim 6, characterized in that, The ground working vehicle subsystem further comprises a state monitoring module, which is used for collecting working vehicle operation data in real time during cleaning work, and judging the working vehicle fault condition according to the working vehicle operation data.

Images