Large-range cleaning method for ground-air cooperative photovoltaic panel

Abstract

The invention relates to a ground-air cooperative photovoltaic panel large-range cleaning method, and belongs to the technical field of unmanned aerial vehicle path planning and photovoltaic cleaning. The method is realized based on a ground-air collaborative cleaning system composed of a multi-rotor unmanned aerial vehicle, a ground station, a photovoltaic cleaning robot and an unmanned aerial vehicle hoisting device, and the core innovation lies in that cross-regional transfer of the cleaning robot carried by the unmanned aerial vehicle is deeply fused with optimal path planning. The unmanned aerial vehicle serves as a transfer main body, carries the robot through the hoisting device and flies along the optimal path planned by the ant colony algorithm; and after the robot arrives at the center of each photovoltaic panel operation area, the robot is accurately lowered to execute ground cleaning operation, and after operation is completed, the robot is recycled and goes to the next area, and finally efficient cleaning of the photovoltaic panels in a large range is achieved. The method optimizes the heuristic function according to the regular distribution characteristics of the photovoltaic panel array, ensures that the unmanned aerial vehicle covers all target areas with the shortest path, effectively reduces the energy consumption and operation time, and is suitable for the complex operation environment of a large photovoltaic field.

Description

Technical Field

[0001] This invention belongs to the field of UAV path planning and photovoltaic cleaning technology, specifically relating to a ground-air collaborative method for large-scale cleaning of photovoltaic panels. Background Technology

[0002] With the large-scale expansion of the photovoltaic energy industry, the need for routine maintenance of large-scale photovoltaic fields is becoming increasingly urgent. Dust, sand, and other pollutants deposited on the surface of photovoltaic panels can reduce power generation efficiency by 10% to 30%. Regular and efficient cleaning has become a core element in ensuring the stable operation of photovoltaic systems. Currently, photovoltaic cleaning robots are the mainstream choice among photovoltaic cleaning equipment due to their low operating costs and good cleaning effects. However, these devices have a significant application bottleneck—weak regional transfer capability. Traditional photovoltaic cleaning robots mostly rely on ground-based wheeled or tracked mobile structures, and can only operate within a single photovoltaic panel array. When cleaning across arrays or areas is required, manual handling or ground track assistance is necessary. Frequent interruptions are required when transferring between arrays. Their flexibility is extremely poor when facing obstacles such as power poles and maintenance access routes within the photovoltaic field. In large photovoltaic fields exceeding a thousand acres, the time spent on regional transfer by the robot can even exceed the actual cleaning time, severely restricting overall operational efficiency.

[0003] To address the challenge of transferring robots across different areas, multi-rotor drones have been introduced into the photovoltaic cleaning field. Drones, with their high mobility and wide coverage, can quickly transport cleaning robots across various work areas, overcoming obstacles. However, current drone path planning technology has significant gaps in photovoltaic applications. More importantly, existing technologies lack a collaborative link between the "drone transfer path and robot cleaning operation," resulting in a disconnect between the drone transfer rhythm and the robot's operational rhythm, further reducing overall cleaning efficiency. Summary of the Invention

[0004] The purpose of this invention is to address the problems existing in the prior art by providing a method for large-scale cleaning of ground-air coordinated photovoltaic panels.

[0005] The core of the ground-air coordinated photovoltaic panel cleaning system upon which this invention relies consists of four parts:

[0006] 1. Multi-rotor UAVs: As a mobile carrier and control core, they have long endurance, high-precision hovering and autonomous navigation capabilities. Equipped with a flight control system, positioning module and communication module, they can receive path planning data and execute flight commands, providing an installation platform and power support for hoisting devices and cleaning robots.

[0007] 2. Ground Station: As the control center for multi-rotor drones and photovoltaic cleaning robots, it has the functions of controlling the movement and operation of drones and robots, real-time video monitoring and path planning, and can receive and display various equipment status parameters of drones and cleaning robots in real time.

[0008] 3. Photovoltaic Cleaning Robot: As a ground cleaning execution unit, it adopts a tracked mobile structure adapted to the surface of photovoltaic panels, and integrates a roller brush cleaning mechanism, power and other modules. It can complete the cleaning operation of photovoltaic panels in a designated area and receive control signals such as start, stop and return to position of the ground docking device sent by the ground station.

[0009] 4. Drone hoisting device: Fixed to the underside of the drone, it consists of an electric winch, an end effector, and tension monitoring sensors. It has hoisting, lowering, and hovering functions, and can control the raising and lowering of the cleaning robot by retracting the hoisting rope. The end effector ensures that the robot does not fall off during the transfer process.

[0010] The cleaning approach of this invention revolves around "path optimization - precise transfer - efficient cleaning": First, a grid map is constructed based on the photovoltaic field environment. The take-off point, landing point of the drone, and the center of the photovoltaic panel operation area are set as target points. The shortest path covering all target points is solved using an ant colony algorithm to ensure that the drone's flight energy consumption is minimized. The drone carries a cleaning robot and flies along the optimal path. After reaching the target point, it uses high-precision hovering positioning to control the hoisting device to smoothly lower the robot onto the photovoltaic panel surface. The robot performs the cleaning operation and returns to the landing point after completion. The operator uses ground station images to control the hoisting device to retrieve the robot. The drone carries the robot to the next target point and repeats the above process until all areas are cleaned and then returns to the landing point.

[0011] The core of this invention lies in transforming the drone-assisted cleaning operation of photovoltaic panels into the Traveling Salesman Problem (TSP), and using the ant colony algorithm to find the shortest path passing through all target points (takeoff point, landing point, and the center of each photovoltaic array area). The specific technical solution is as follows:

[0012] Step 1: Constructing a photovoltaic field grid map

[0013] Based on the actual dimensions, photovoltaic panel array distribution, and obstacle location information of the imported photovoltaic field map data, a two-dimensional rectangular coordinate system is established with the physical boundary of the photovoltaic field as the range, and the origin of the coordinate system is set as the lower left vertex of the photovoltaic field. The scale factor k is set to 1:1, that is, 1 unit in the raster map corresponds to 1m in the actual scene, and the photovoltaic field area is divided into a 20×20 square raster array.

[0014] Grid state definition: The photovoltaic panel array and its internal passageways are marked as gray passable areas, with the grid borders of the photovoltaic panel array highlighted in bold black. Grids containing obstacles such as utility poles and fences are marked as obstacle grids, with a state value of 1. Areas where drones can fly safely, such as passageways, are marked as passable grids, with a state value of 0. The conversion formula between grid coordinates and actual coordinates is:

[0015]

[0016] Where, x real y real For actual coordinates (unit: m), x grid y grid INT represents the raster coordinates (unit: grid), and INT is the integer function.

[0017] Step 2: Determine the path target point

[0018] Based on the operational needs of the photovoltaic field, two types of path target points were determined: one is the drone operation point, including the takeoff point P0 and the landing point P. n Both are set in open areas at the edge of the photovoltaic field, and their coordinates are obtained through actual measurement and grid conversion formulas; the second is the target point of the photovoltaic panel operation area, which is divided into operation units consisting of 3×7 photovoltaic panels and their internal spacing channels, with the geometric center of each unit taken as the target point P. i (i=1,2,...,n), ensuring that the drone can complete the deployment and retrieval of the cleaning robot at this point.

[0019] The set of coordinates of all target points is denoted as G = {P0, P1, P2, ..., P...}. n}, where each target point corresponds to a unique coordinate (x, y) in the raster map.

[0020] any point The relationship between the grid number N and the grid number N can be expressed by the following formula:

[0021]

[0022] in, It is the range of coordinate values ​​for the raster. It represents the size of the raster grains, and INT is the integer function.

[0023] Step 3: Ant colony algorithm parameter initialization

[0024] The core parameters of the ant colony algorithm are set as follows: Ant count m = 30 (positively correlated with the number of target points, taken as 3-5 times the number of target points); Pheromone evaporation coefficient ρ = 0.95 (controls the pheromone decay rate, balancing global and local search capabilities); Pheromone weight factor α = 1.0 (affects the strength of pheromone's effect on path selection); Distance heuristic factor β = 15.0 (highlights the guiding role of distance factors in path selection); Maximum number of iterations G = 50; Pheromone release constant Q = 100; Initial pheromone concentration. =5.

[0025] Initialize the tabu table tabo for each ant. k (k=1, 2, ..., m), the tabu list is used to record the target points visited by the ants. In the initial state, it only contains the starting position of the ants (all ants are randomly distributed in each target point).

[0026] Step 4, Path Search

[0027] In each iteration, the ant selects the next target point to visit based on the pheromone concentration along the path and the distance between the target point, using a probability-based approach. The path selection probability formula is expressed as follows:

[0028]

[0029] in, Let be the probability that the k-th ant travels from target point i to target point j at time t; Let be the pheromone concentration on path ij at time t; The heuristic function takes the value d, which is the Euclidean distance between target points i and j. ij The reciprocal of, that is =1 / d ij J k (i) represents the set of target points that the k-th ant has not visited. d ij The calculation formula is:

[0030]

[0031] After the ant has visited all the target points, it stores the current path in the path set and calculates the path length L. k (k=1,2,...,m), the path length is the sum of the Euclidean distances between each adjacent target point.

[0032] Step 5: Pheromones Update

[0033] A pheromone update mechanism combining local and global updates is employed: local updates are performed after an ant completes a jump to the target point to prevent premature convergence; global updates are performed after all ants have completed a full path search to reinforce pheromone guidance for the optimal path. The update formula is as follows:

[0034] Local pheromone update formula: (t+1)=(1-ρ)· (t)+ρ· ,in The initial pheromone concentration;

[0035] Global pheromone update formula: (t+1)=(1-ρ)· (t)+Δ , where Δ For pheromone increment, Δ =Q / L best Q is the pheromone release constant, L best This is the globally optimal path length for this iteration.

[0036] Step 6: Algorithm convergence judgment

[0037] If the current iteration count reaches the preset maximum iteration count G, or the change in the optimal path length over 10 consecutive iterations is less than 0.1m, the algorithm is considered to have converged, and the globally optimal path is output; otherwise, the tabu list of all ants is cleared, and the algorithm returns to step 4 to continue iterating.

[0038] Step 7: Output the optimal path

[0039] After outputting the optimal path, the path information of the optimal path can be transmitted to the UAV from the ground station control system, and the UAV will move along the optimal path to the corresponding target point to carry out the operation.

[0040] The present invention also provides an electronic device, including a memory, a processor, and computer program instructions stored in the memory and executable by the processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0041] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] (1) Highly targeted: Focusing on the path planning of UAVs in the photovoltaic panel area, adapting to the precise needs of ground-air collaborative operations; combining the regular distribution characteristics of photovoltaic panel arrays to divide the operation area, improving the rationality of target point selection.

[0044] (2) Good path optimization effect: Based on the TSP problem model, the path covers all target points without repetition. The ant colony algorithm balances the global search capability and convergence speed through weight factor adjustment and pheromone update mechanism, effectively avoiding local optima.

[0045] (3) High practicality: The grid map construction method is simple and easy to implement. It can be quickly generated from photovoltaic field design drawings. The 20×20 grid size takes into account both accuracy and calculation efficiency, reduces the difficulty of drone operation, and adapts to complex photovoltaic field environments. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments will be briefly described below, wherein the black grid is an obstacle grid, the white grid is a passable grid, the gray grid is an internal passable channel, and the gray grid with a black thick border is a photovoltaic array.

[0047] Figure 1 This is a schematic diagram of the overall algorithm flow of the present invention.

[0048] Figure 2 This is the optimal path graph for each segment of the path search algorithm in this invention.

[0049] Figure 3 This is an integrated diagram of the optimal path of the algorithm in this invention. Detailed Implementation

[0050] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] This invention provides a ground-air collaborative method for large-scale cleaning of photovoltaic panels, implemented based on a ground-air collaborative photovoltaic panel cleaning system. The cleaning system includes a multi-rotor drone, a ground station, a photovoltaic cleaning robot, and a hoisting device mounted on the drone's fuselage. The hoisting device has lifting, lowering, and hovering / fixing functions, used to facilitate the transfer of the photovoltaic cleaning robot between photovoltaic panel work areas assisted by the drone. The cleaning method uses the multi-rotor drone as the execution entity, constructs a drone path optimization model based on the Traveling Salesman Problem (TSP), and employs an ant colony algorithm to achieve optimal path planning for the drone carrying the photovoltaic cleaning robot within a grid map of the photovoltaic panel area.

[0052] The following are specific implementation examples of the present invention.

[0053] This invention embodiment selects an array of 84 photovoltaic panels in a large photovoltaic field as the work object. Each photovoltaic panel measures 1.6m × 1.0m, with a spacing of 1m between panels. There are four utility poles and other obstacles around the array. The described embodiment is merely a part of the embodiments of this invention, not all of them. Various modifications and improvements made to the technical solutions of this invention by those skilled in the art without departing from the inventive concept should fall within the protection scope of this invention.

[0054] Example 1:

[0055] like Figure 1 As shown, a method for large-scale cleaning of ground-air coordinated photovoltaic panels includes the following steps:

[0056] Step 1: Raster Map Construction. Operators import map data of the photovoltaic area where cleaning operations will be performed into the ground station system. The system will automatically generate a two-dimensional raster map based on the provided map data, such as... Figure 2 , Figure 3 The four gray grid areas shown represent the four photovoltaic arrays and their internal passageways. At this point, the operator uses a ground station to lift the photovoltaic cleaning robot from the drone and hover it approximately 10m above the ground at the takeoff point P0. The actual size of the photovoltaic field operation area is 20m × 20m. A 20 × 20 grid map is constructed using a 1:1 scale factor (each grid corresponds to 1m in actual dimensions). Utility pole obstacles are located at actual coordinates (5m, 10m), (15m, 2m), (10m, 15m), and (18m, 12m), corresponding to grid coordinates (5, 10), (15, 2), (10, 15), and (18, 12), and are marked as obstacle grids. Considering the possibility of other obstacles in the actual photovoltaic field operation area, other obstacles are also marked on the grid map, and these obstacle grids are uniformly represented in black. A 20 × 20 grid map is generated according to the above parameters, marking obstacle grids and passable grids. The result is as follows: Figure 2 , Figure 3 As shown.

[0057] Step 2: Initialize the ant colony algorithm parameters. On the ground station, the software's built-in algorithm is set by default with the following parameters: number of ants m=30, ρ=0.95, α=1.0, β=15.0, G=50, Q=100, τ0=5. The 30 ants are randomly distributed across 6 target points, and the tabu list for each ant initially only contains the starting target point.

[0058] Step 3: Setting Path Target Points. Operators set a series of target points along the path using the ground station. For example... Figure 2 , Figure 3The actual coordinates of the takeoff point P0 are (1m, 20m), corresponding to the grid coordinates (1, 20); the actual coordinates of the landing point P5 are (20m, 1m), corresponding to the grid coordinates (20, 1); the 84 photovoltaic panels are divided into four 3×7 working areas, with the actual coordinates of the center of each area being P1(5m, 15m), P2(15m, 15m), P3(5m, 5m), and P4(15m, 5m), corresponding to the grid coordinates P1(5, 15), P2(15, 15), P3(5, 5), and P4(15, 5). The target point set G = {P0(1, 20), P1(5, 15), P2(15, 15), P3(5, 5), P4(15, 5), P5(20, 1)}.

[0059] Step 4, Path Search. The first ant starts from P0(1,20) and calculates the probability of reaching each unvisited target point according to the path selection probability formula. Since the initial pheromone concentration is the same, it is mainly affected by the distance heuristic factor, so it prioritizes P1(5,15) as the next target point. After reaching P1, the taboo table is updated and the search for the next target point continues until all target points are visited and P5(20,1) is reached. The recorded path is P0→P1→P2→P3→P4→P5.

[0060] Step 5: Pheromone Update. After all ants have completed the first path search, calculate the optimal path length L for this iteration. iter (Path P0→P1→P2→P3→P4→P5), perform a global pheromone update, increasing the pheromone increment Δ for each segment of this path. =100 / 54.97≈1.82.

[0061] Step 6: Iterative Convergence Judgment. Repeat steps 4 and 5 up to the 50th iteration. After the path length stabilizes, the final globally optimal path is obtained as P0(1,20)→P1(5,15)→P2(15,15)→P3(5,5)→P4(15,5)→P5(20,1), with an optimal travel distance of 48.28m. After obtaining the final optimized path, the ground station's human-machine interface will prompt the operator to complete the optimal path planning. After pressing the takeoff button on the ground station, the UAV departs from the takeoff point P0. Upon reaching each target point except the landing point P5, it will hover approximately 10m above the ground and lower the photovoltaic cleaning robot. The operator will then unlock the cleaning robot's ground-to-air docking device from the hoisting device's end mechanism via the ground station and control the cleaning robot to complete the cleaning operation. After completing the cleaning operation, the cleaning robot returns to the landing point. The operator will then use the hoisting device via the ground station to retrieve the cleaning robot. The UAV will then proceed to the next target point along the planned route, repeating the same process until it reaches the landing point.

[0062] Once the work on a photovoltaic field area is completed, the target points and grid map can be reinitialized through the ground station control system to quickly generate a new optimal path, demonstrating good environmental adaptability.

[0063] The present invention also provides an electronic device, including a memory, a processor, and computer program instructions stored in the memory and executable by the processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0064] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0065] The above description is merely a preferred embodiment of the present invention, intended to enable those skilled in the art to understand and implement the content of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for large-scale cleaning of photovoltaic panels using a ground-air coordinated approach, characterized in that, The system is based on a ground-air collaborative photovoltaic panel cleaning system. The cleaning system includes a multi-rotor drone, a ground station, a photovoltaic cleaning robot, and a hoisting device installed on the drone's fuselage. The hoisting device has lifting, lowering, and hovering functions to facilitate the transfer of the photovoltaic cleaning robot between photovoltaic panel work areas with the assistance of the drone. The cleaning method uses the multi-rotor drone as the execution subject, constructs a drone path optimization model based on the Traveling Salesman Problem (TSP), and uses an ant colony algorithm to achieve optimal path planning for the drone carrying the photovoltaic cleaning robot in the grid map of the photovoltaic panel area.

2. The method for large-scale cleaning of ground-air coordinated photovoltaic panels according to claim 1, characterized in that, The method includes the following steps: Step 1: Construct a two-dimensional grid map of the photovoltaic field. Using the actual size of the photovoltaic field and a preset scale factor, the photovoltaic panel array and its surrounding environment are converted into a two-dimensional grid map containing passable grids and obstacle grids. Step 2: Determine the target points for path planning. Set the take-off point, landing point of the drone, and the center of each photovoltaic panel area as the target points that must be reached on the path, forming a set of target point coordinates. In this stage, the drone will hoist the cleaning robot and hover at the take-off point. Step 3: Initialize the ant colony algorithm parameters, including the number of ants, pheromone volatility coefficient, pheromone weight factor, distance heuristic factor, and maximum number of iterations; Step 4: Execute the ant colony algorithm for path search. Ants determine their direction of movement based on the path pheromone concentration and the distance to the target point. They record the visited target points through a taboo list and form a complete path after traversing all target points. Step 5: Update path pheromones, combining the optimal path of this iteration with the global optimal path to volatilize and replenish pheromones, and strengthen the guiding role of high-quality paths; Step 6: Determine if the algorithm has converged. If the number of iterations reaches the threshold or the length of the optimal path stabilizes after multiple consecutive iterations, output the global optimal path; otherwise, return to step 4 to continue iterating. Step 7: Output the global optimal path. After obtaining the optimal path, the UAV starts from the take-off point and arrives at each target point along the optimal path. The photovoltaic cleaning robot is accurately deployed and hovered during the cleaning operation. After the operation is completed, the cleaning robot is lifted and moved to the next target point to repeat the same operation process until it reaches the landing point.

3. The method for large-scale cleaning of ground-air coordinated photovoltaic panels according to claim 2, characterized in that, The grid map construction method described in step 1 is as follows: A rectangular coordinate system is established with the lower left corner of the photovoltaic field as the origin. A scale factor of 1:1 is set, and the photovoltaic field area is divided into 20×20 square grids, with each grid's side length corresponding to 1m in the actual scene. The grids corresponding to the photovoltaic panel coverage area and surrounding obstacles are marked as obstacle grids, and the drone-flyable area is marked as passable grids. In the rectangular coordinate method, the center coordinates of each grid are defined as rectangular coordinates, ensuring that each grid number corresponds to its own rectangular coordinates. The relationship between any point (x, y) and the grid number N is expressed by the following formula: in, It is the range of coordinate values ​​for the raster. It represents the size of the raster grains, and INT is the integer function.

4. A method for large-scale cleaning of ground-air coordinated photovoltaic panels according to claim 2, characterized in that, The rule for constructing the target point coordinate set in step 2 is as follows: the drone's takeoff point is set to P0(x0,y0), and the landing point is set to P... n (x n ,y n The two are independent path endpoints; for a matrix arrangement of photovoltaic panels, each 3×7 photovoltaic panels and their internal spacing channels form a working area, and the geometric center of the area is taken as the target point P. i (xᵢ,yᵢ), the coordinates of the target point are calculated using the conversion formula between raster map coordinates and actual coordinates.

5. A method for large-scale cleaning of ground-air coordinated photovoltaic panels according to claim 2, characterized in that, In step 3, the initial parameters for the ant colony algorithm are: number of ants m = 10~30, pheromone evaporation coefficient ρ = 0.2~0.95, pheromone weight factor α = 1.0, distance heuristic factor β = 15.0, and maximum number of iterations G = 50; the initial pheromone concentration on all paths is also considered. (0) = c, where c is a constant.

6. A method for large-scale cleaning of ground-air coordinated photovoltaic panels according to claim 2, characterized in that, The formula for the ant path selection probability in step 4 is: in, Let be the probability that the k-th ant travels from target point i to target point j at time t; Let be the pheromone concentration on path ij at time t; The heuristic function takes the value of the reciprocal of the Euclidean distance between target points i and j, i.e. =1 / J k (i) is the set of target points that the k-th ant has not visited.

7. A method for large-scale cleaning of ground-air coordinated photovoltaic panels according to claim 2, characterized in that, Step 5's pheromone update formula includes two mechanisms: global update and local update. Local pheromone update formula: (t+1)=(1-ρ)· (t)+ρ· ,in Let be the pheromone concentration along path ij at time t. The initial pheromone concentration is ρ, and the pheromone evaporation coefficient is ρ. Global pheromone update formula: (t+1)=(1-ρ)· (t)+Δ , where Δ For pheromone increment, Δ =Q / L best Q is the pheromone release constant, L best This is the globally optimal path length for this iteration.

8. A method for large-scale cleaning of ground-air coordinated photovoltaic panels according to claim 2, characterized in that, The rule for dividing the photovoltaic panel work area in step 2 is as follows: every 3×7 photovoltaic panels and their internal spacing channels form a work unit, and the geometric center of each unit is taken as the target point P. i (i=1,2,…,n), ensuring that the drone can complete the deployment and retrieval of the photovoltaic cleaning robot at this point.

9. An electronic device, characterized in that, It includes a memory, a processor, and computer program instructions stored in the memory and executable by the processor, which, when executed by the processor, enable the implementation of the steps of the method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, It stores computer program instructions that can be executed by a processor, and when the processor executes the computer program instructions, it can implement the steps of the method as described in any one of claims 1-8.

Images