Intelligent cleaning robot for photovoltaic module and path planning method thereof
By constructing a pollution situation map in real time and planning dynamic priority areas, the photovoltaic cleaning robot can achieve efficient and thorough cleaning in complex pollution scenarios, solving the problems of low efficiency and high energy consumption in existing technologies, adapting to non-rectangular array layouts and responding to sudden pollution.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 华能海南发电股份有限公司南山电厂
- Filing Date
- 2026-01-30
- Publication Date
- 2026-05-12
AI Technical Summary
Existing photovoltaic cleaning robots are incomplete and inefficient in complex and uneven pollution scenarios, and cannot cope with non-rectangular arrays and sudden pollution, resulting in high energy consumption and severe mechanical wear.
The system uses an environmental perception module to build a pollution status map in real time, dynamically divides priority areas, and plans paths through a biomimetic detour coverage path and real-time perception feedback to prioritize the treatment of highly polluted areas.
It achieves efficient cleaning in complex pollution scenarios, reduces ineffective coverage and repeated paths, improves cleaning quality and equipment lifespan, and reduces energy consumption.
Smart Images
Figure CN122018503A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of photovoltaic power plant operation and maintenance technology, and in particular to an intelligent cleaning robot for photovoltaic modules and its path planning method. Background Technology
[0002] With the rapid development of the photovoltaic industry, the scale and number of photovoltaic power plants have increased dramatically. The cleanliness of the photovoltaic module surface directly affects the power generation efficiency, and regular cleaning has become a necessary part of power plant operation and maintenance. Due to its automation and high efficiency, the photovoltaic module intelligent cleaning robot has been widely used.
[0003] Currently, most photovoltaic cleaning robots on the market use preset fixed trajectory patterns for path planning, the most typical being "bow" or "return" shaped full-coverage paths; however, this path planning method is based on the ideal assumption that the photovoltaic array is a regular rectangle and that surface contaminants are evenly distributed.
[0004] However, in large-scale photovoltaic power plants, pollution distribution is often extremely uneven. For example, due to factors such as bird activity, tree obstruction, or oil dripping from equipment, localized heavily polluted areas (such as dense bird droppings) may form, while most other areas are relatively clean. Robots with fixed paths will traverse the entire array at the same speed and cleaning intensity, resulting in a significant waste of time and energy in clean areas, while residues may remain in truly heavily polluted areas due to incomplete cleaning in a single pass, leading to low overall operational efficiency. At the same time, for complex array layouts that are not rectangular and have corners or gaps, fixed "bow"-shaped paths will generate a large number of invalid coverages and repetitive paths, requiring frequent sharp turns, which not only reduces cleaning efficiency but also increases mechanical wear and energy consumption of the robots. Furthermore, existing robot paths are pre-set and cannot respond to sudden pollution during operation (such as bird droppings in flight). They lack the ability to make dynamic decisions based on real-time environmental perception, resulting in limited intelligence. Because they cannot distinguish the degree of pollution, the robots operate at a fixed power consumption throughout the process, leading to a low cost-effectiveness ratio between overall energy consumption and cleaning results.
[0005] Therefore, to address the above problems, we propose an intelligent cleaning robot for photovoltaic modules and its path planning method. By constructing a pollution map, dividing priority areas, executing biomimetic detour cleaning based on queue scheduling, and using real-time perception feedback and dynamic replanning, it can achieve efficient and high-quality cleaning in complex and uneven pollution scenarios. Summary of the Invention
[0006] To overcome the problems of incomplete cleaning leading to low efficiency and poor quality in complex and unevenly polluted scenarios, and the poor adaptability of photovoltaic cleaning robots in the current technology, which also result in a large number of invalid coverages and repeated paths in complex array layouts with non-rectangular corners or gaps.
[0007] The technical solution of this invention is: a photovoltaic module intelligent cleaning robot, comprising:
[0008] A moving mechanism for moving the photovoltaic module array; Cleaning actuators are used to remove dirt from the surface of photovoltaic modules; The environmental sensing module is used to acquire surface image information and / or dirt concentration information of the photovoltaic module array in real time; The control unit is communicatively connected to the moving mechanism, the cleaning execution mechanism, and the environmental sensing module. The control unit is configured as follows: Based on the information obtained by the environmental perception module, a global pollution status map is constructed and dynamically updated. This map divides the photovoltaic array surface into multiple grids and assigns a comprehensive pollution level to each grid. Based on the comprehensive pollution level, the entire work area is dynamically divided into at least two sub-areas with different cleaning priorities, and a dynamically changing area priority queue is generated. The path planning module controls the mobile mechanism to prioritize the highest priority sub-region based on the regional priority queue, and executes a non-preset fixed trajectory biomimetic detour coverage path within that sub-region.
[0009] Preferably, the device moves on the photovoltaic module array via a moving mechanism; a cleaning execution mechanism removes dirt from the surface of the photovoltaic modules; an environmental perception module acquires real-time surface image information and / or dirt concentration information of the photovoltaic module array; a control unit constructs and dynamically updates a global pollution status map based on the information acquired by the environmental perception module, the map dividing the photovoltaic array surface into multiple grids and assigning a comprehensive pollution level to each grid; the control unit dynamically divides the entire work area into at least two sub-areas with different cleaning priorities based on the comprehensive pollution level, and generates a dynamically changing area priority queue; a path planning module controls the moving mechanism to prioritize the highest priority sub-area according to the area priority queue, and executes a non-preset fixed trajectory biomimetic detour coverage path within that sub-area; thus, through an intelligent decision-making algorithm based on the dynamic pollution status map and priority queue, the robot is driven to execute the biomimetic detour coverage path.
[0010] Preferably, when the control unit is configured to construct and dynamically update the pollution status map, it is specifically used for: Based on the initial rapid scan data from the environmental sensing module, each grid is assigned an initial pollution level and initial uncertainty parameters. During the cleaning process, the pollution level of the cleaned grid is updated based on the real-time monitoring data of the environmental sensing module, and its uncertainty parameters are reduced. For grids that are not directly observed but are adjacent to observed areas, their pollution level is estimated based on the data and spatial correlation of their neighboring grids and assigned a high uncertainty parameter.
[0011] Preferably, when the control unit is configured to generate a dynamically changing region priority queue, it is specifically used for: Grids with pollution levels higher than the first threshold and that are spatially adjacent are merged into a first-priority sub-region. Grids with pollution levels below the first threshold but above the second threshold and that are spatially adjacent are merged into a second priority sub-region. The remaining grid cells are classified as third priority sub-regions by default. Among them, the queue order of the first priority sub-region is always higher than that of the second priority sub-region, and the queue order of the second priority sub-region is always higher than that of the third priority sub-region.
[0012] Preferably, the control unit is further configured as follows: If, during the operation of the environmental perception module in the current sub-area, a new pollution event is detected that causes the pollution level of a certain grid to rise sharply to exceed the first threshold, the current path is immediately interrupted, the new area containing the grid is inserted as the highest priority task at the head of the area priority queue, and the mobile mechanism is controlled to proceed to perform cleaning.
[0013] Preferably, the biomimetic detour coverage path specifically refers to controlling the moving mechanism to cover the area within the boundary of the sub-region by walking along the boundary of the region and combining it with local random walks.
[0014] Preferably, the control unit is further configured as follows: The working parameters of the cleaning actuator and / or the moving speed of the moving mechanism are dynamically adjusted according to the priority or average pollution level of the current sub-region. Specifically, for high-priority or high-pollution-level sub-regions, a combination mode of low moving speed and high cleaning intensity is adopted; for low-priority or low-pollution-level sub-regions, a combination mode of high moving speed and low cleaning intensity is adopted.
[0015] A path planning method for an intelligent cleaning robot for photovoltaic modules includes the following steps: S1: The photovoltaic module array is scanned through the environmental sensing module to obtain initial pollution data; S2: Based on the initial pollution data, construct a global pollution status map; S3: Based on the pollution situation map, dynamically divide the area into sub-areas with different cleaning priorities and form an area priority queue; S4: The path planning module controls the robot to move to the highest priority sub-region based on the queue; S5: Within the current highest priority sub-region, control the robot to execute a biomimetic detour coverage path that is not a preset fixed trajectory for cleaning; S6: During the cleaning process, the pollution situation map and area priority queue are dynamically updated based on real-time data from the environmental perception module. S7: Repeat steps S4 to S6 until the pollution level of all sub-regions is lower than the preset completion threshold.
[0016] Preferably, the dynamic division of sub-regions in step S3 includes: One or more grid clusters with the highest pollution level and spatial contiguousness are identified as heavily polluted areas and given the highest priority. One or more grid clusters with moderate pollution levels and spatial contiguousness are identified as moderately polluted areas and assigned a medium priority. The remaining grid cells are classified as background cleanup areas by default and given the lowest priority.
[0017] Preferably, the dynamic update in step S6 includes: When the robot is cleaning in a sub-area, if the environmental perception module detects a new high-pollution point that does not belong to the current area and whose pollution level exceeds a certain threshold, the task will be replanned immediately. Pause the cleaning task in the current sub-area, set the area where the new high-contamination point is located as the temporary highest priority target, and control the robot to go there for cleaning; Once the contamination level of the temporary highest priority target drops to a safe threshold, the robot is controlled to return to the atomic region to continue the interrupted task.
[0018] Preferably, the execution of the biomimetic detour coverage path in step S5 is specifically as follows: After entering a sub-area, the robot first travels around the outer perimeter of the sub-area to complete the boundary cleaning. Subsequently, based on the random walk algorithm and combined with real-time judgment of cleaned and uncleaned areas, a filling-style roundabout cleaning is performed within the sub-region.
[0019] The beneficial effects of this invention are: This invention uses a moving mechanism to move the device on a photovoltaic module array; a cleaning execution mechanism to remove dirt from the surface of the photovoltaic modules; an environmental perception module to acquire real-time surface image information and / or dirt concentration information of the photovoltaic module array; a control unit to construct and dynamically update a global pollution situation map based on the information acquired by the environmental perception module, which divides the surface of the photovoltaic array into multiple grids and assigns a comprehensive pollution level to each grid; the control unit to dynamically divide the entire work area into at least two sub-areas with different cleaning priorities based on the comprehensive pollution level, and generates a dynamically changing area priority queue; a path planning module to control the moving mechanism to prioritize the highest priority sub-area according to the area priority queue, and execute a non-preset fixed trajectory biomimetic detour coverage path within that sub-area; thus, through an intelligent decision-making algorithm based on the dynamic pollution situation map and priority queue, the robot is driven to execute a biomimetic detour coverage path, achieving efficient and high-quality cleaning in complex and uneven pollution scenarios. Attached Figure Description
[0020] Figure 1 The diagram shown is a three-dimensional structural schematic of the intelligent cleaning robot for photovoltaic modules according to the present invention. Figure 2 The diagram shown is a flowchart illustrating the path planning method for the intelligent cleaning robot for photovoltaic modules according to the present invention. Explanation of reference numerals in the attached diagram: 1. Moving mechanism; 2. Cleaning execution mechanism; 3. Environmental sensing module; 4. Control unit. Detailed Implementation
[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0022] Example 1 Please see Figure 1 and Figure 2 This invention provides an embodiment of a photovoltaic module intelligent cleaning robot, comprising: The moving mechanism 1 is used to move on the photovoltaic module array; Cleaning actuator 2 is used to remove dirt from the surface of photovoltaic modules; Environmental sensing module 3 is used to acquire surface image information and / or dirt concentration information of the photovoltaic module array in real time; Control unit 4 is communicatively connected to moving mechanism 1, cleaning execution mechanism 2 and environmental sensing module 3; Control unit 4 is configured as follows: Based on the information obtained by the environmental perception module 3, a global pollution status map is constructed and dynamically updated. This map divides the photovoltaic array surface into multiple grids and assigns a comprehensive pollution level to each grid. Based on the comprehensive pollution level, the entire work area is dynamically divided into at least two sub-areas with different cleaning priorities, and a dynamically changing area priority queue is generated. The path planning module controls the mobile mechanism 1 to prioritize the highest priority sub-region based on the regional priority queue, and executes a non-preset fixed trajectory biomimetic detour coverage path within that sub-region.
[0023] The device moves on the photovoltaic module array via the moving mechanism 1; the cleaning execution mechanism 2 removes dirt from the surface of the photovoltaic modules; the environmental perception module 3 acquires real-time surface image information and / or dirt concentration information of the photovoltaic module array; the control unit 4 constructs and dynamically updates a global pollution status map based on the information acquired by the environmental perception module 3, which divides the surface of the photovoltaic array into multiple grids and assigns a comprehensive pollution level to each grid; the control unit 4 dynamically divides the entire work area into at least two sub-areas with different cleaning priorities based on the comprehensive pollution level, and generates a dynamically changing area priority queue; the path planning module controls the moving mechanism 1 to prioritize the highest priority sub-area according to the area priority queue, and executes a non-preset fixed trajectory biomimetic detour coverage path within that sub-area; thus, through an intelligent decision-making algorithm based on the dynamic pollution status map and priority queue, the robot is driven to execute the biomimetic detour coverage path.
[0024] Furthermore, when the control unit 4 is configured to build and dynamically update the pollution status map, it is specifically used for: Based on the initial rapid scan data from the environmental sensing module 3, each grid is assigned an initial pollution level and initial uncertainty parameters. During the cleaning process, the pollution level of the cleaned grid is updated based on the real-time monitoring data of the environmental sensing module 3, and its uncertainty parameters are reduced. For grids that are not directly observed but are adjacent to observed areas, their pollution level is estimated based on the data and spatial correlation of their neighboring grids and assigned a high uncertainty parameter.
[0025] Control unit 4 logically rasterizes the photovoltaic array surface into a two-dimensional grid map; each grid cell stores two core state variables: pollution level value P (e.g., a value normalized to 0-1 by image grayscale analysis or laser scattering sensor readings) and uncertainty parameter U (indicating the age and reliability of the data in that cell; areas not directly observed after the initial scan have higher U values). Its update mechanism follows these principles: Initial Scan: After the robot starts, it first moves quickly along the edge of the array or a simplified path, and the environment perception module 3 performs the first wide-area scan. Based on the scan data, each cell that falls into the field of view is assigned an initial P value and its U value is reduced to the minimum. For cells that are not directly scanned but are in the neighborhood of scanned cells, their P value is estimated by spatial interpolation algorithm and assigned a higher U value. Real-time updates: During the cleaning process, the environmental perception module 3 continues to operate; for the cell where the robot is currently located, its P value is updated in real time to the latest sensor reading, and its U value is reset to the lowest value; simultaneously, through the sensor's field of view, the P and U values of cells within a certain range in front of and to the side are also updated; thus, the map can reflect the latest pollution status and gradually reduce global uncertainty. Furthermore, when the control unit 4 is configured to generate a dynamically changing region priority queue, it is specifically used for: Grids with pollution levels higher than the first threshold and that are spatially adjacent are merged into a first-priority sub-region. Grids with pollution levels below the first threshold but above the second threshold and that are spatially adjacent are merged into a second priority sub-region. The remaining grid cells are classified as third priority sub-regions by default. Among them, the queue order of the first priority sub-region is always higher than that of the second priority sub-region, and the queue order of the second priority sub-region is always higher than that of the third priority sub-region.
[0026] Specifically, control unit 4 periodically or when the pollution status map changes significantly, divides the entire map into regions; a threshold-based region growing algorithm is used. First, find all cells whose P-values exceed the first threshold P_high as seed points; Then, spatially adjacent (quad-connected or oct-connected) cells with P values exceeding the second threshold P_medium are merged to form a continuous "heavily polluted area". Similarly, grids with pollution levels below the first threshold but above the second threshold and that are spatially adjacent are merged to form a "moderately polluted zone". The remaining cells are treated as the "background cleanup area"; Subsequently, the system generates a task queue, with the head of the queue always being the sub-area with the highest overall contamination level (which can be defined as the weighted sum of the average P value and the average U value). This ensures that the robot can always proactively find and prioritize the dirtiest and most uncertain areas, thereby maximizing cleaning efficiency and avoiding ineffective stays in clean areas.
[0027] Furthermore, the control unit 4 is also configured as follows: If, during the operation of the environmental perception module 3 in the current sub-area, a new pollution event is detected that causes the pollution level of a certain grid to rise sharply to exceed the first threshold, the current path is immediately interrupted, the new area containing the grid is inserted as the highest priority task at the head of the area priority queue, and the moving mechanism 1 is controlled to go there to perform cleaning.
[0028] Specifically, when the robot is performing a cleaning task in a sub-area, the environmental perception module 3 suddenly detects a new, independent contamination point, and its P value instantly exceeds P_high. At this time, the control unit 4 will not stick to the original plan, but will immediately save the current path point onto the stack, mark the micro-area where the new high-contamination point is located as an "emergency task", and insert it at the head of the priority queue. The robot will then pause its current operation and switch to handling the emergency task. After the emergency task is completed, the robot will restore the previous task state and path point from the stack and continue to execute.
[0029] Furthermore, the biomimetic detour coverage path specifically refers to the control of the mobile mechanism 1 to cover the area within the boundary of the sub-region by favoring walking along the boundary of the region and combining it with local random walks.
[0030] Specifically, within the designated sub-region, the robot no longer travels along regular straight lines. Its path planning is manifested as follows: Boundary following: The robot first walks around the logical boundary of the sub-region, which can quickly define the work area and remove dust from the edges; Internal filling random detour: Inside the boundary, the robot uses an algorithm that combines random walk and exclusion of cleaned areas; specifically, at each waypoint, the robot's next movement direction has a certain degree of randomness, but is affected by a virtual pheromone - it will tend to move towards uncleaned cells (high U or P values) and avoid immediately returning to the cleaned area.
[0031] Compared to the single path planning of existing devices, this approach can better adapt to the irregular shapes of sub-regions, reducing the number of turns and repeated paths. For stubborn, irregularly shaped stains, this roundabout approach is more likely to achieve thorough removal than unidirectional back-and-forth cleaning.
[0032] Furthermore, the control unit 4 is also configured as follows: Based on the priority or average pollution level of the current sub-area, the working parameters of the cleaning actuator 2 and / or the moving speed of the moving mechanism 1 are dynamically adjusted; for high priority or high pollution level sub-areas, a combination mode of low moving speed and high cleaning intensity is adopted; for low priority or low pollution level sub-areas, a combination mode of high moving speed and low cleaning intensity is adopted.
[0033] Specifically, the control unit 4 dynamically outputs control instructions to the moving mechanism 1 and the cleaning execution mechanism 2 according to the average P value of the current sub-region. For example, when the average P > P_high, the control unit 4 controls the moving mechanism 1 to travel at a speed of V_slow, and at the same time controls the cleaning execution mechanism 2 to operate at a rotational speed of R_high and / or maximum suction; when the average P < P_low, it switches to a speed of V_fast and a rotational speed of R_low. Thus, while ensuring the cleaning quality, time and energy are maximally saved, and the service life of the equipment is extended.
[0034] A path planning method for an intelligent cleaning robot of a photovoltaic module includes the following steps: S1: The environmental perception module 3 scans the photovoltaic module array to obtain initial pollution data. S2: Based on the initial pollution data, a global pollution situation map is constructed. S3: According to the pollution situation map, sub-regions with different cleaning priorities are dynamically divided, and a regional priority queue is formed. S4: The path planning module controls the robot to move to the current highest-priority sub-region according to the queue. S5: Within the current highest-priority sub-region, the robot is controlled to execute a bionic detour coverage path that is not a preset fixed trajectory for cleaning. S6: During the cleaning process, according to the real-time data of the environmental perception module 3, the pollution situation map and the regional priority queue are dynamically updated. S7: Repeat steps S4 to S6 until the pollution levels of all sub-regions are lower than the preset completion threshold.
[0035] Further, the dynamic division of sub-regions in step S3 includes: One or more contiguous grid clusters with the highest pollution level are identified as severely polluted areas and given the highest priority. One or more contiguous grid clusters with a medium pollution level are identified as moderately polluted areas and given medium priority. The remaining grids are default classified as background cleaning areas and given the lowest priority.
[0036] Specifically, the step of dynamically dividing sub-regions is an image processing technology based on connected component analysis. The pollution situation map is regarded as a grayscale image (the P value is the grayscale), and through setting different thresholds for binaryzation, and then using the connected component labeling algorithm, all connected pollution blocks are found, and the centroid, area and average pollution level of each block are calculated to provide a basis for priority sorting.
[0037] Further, the dynamic update in step S6 includes: When the robot is cleaning in a sub-area, if the environmental perception module 3 detects a new high-pollution point that does not belong to the current area and whose pollution level exceeds a certain threshold, the task will be replanned immediately. Pause the cleaning task in the current sub-area, set the area with the new high-contamination point as the temporary highest priority target, and control the robot to go there for cleaning; Once the contamination level of the temporary highest priority target drops to a safe threshold, the robot is controlled to return to the atomic region to continue the interrupted task.
[0038] When a new high-priority region appears, the system immediately assesses its positional relationship with the current task and the difference in its level of contamination. It then calculates the benefit of interrupting the current task using a cost function. If the benefit is greater than a preset threshold, a replanning is immediately triggered.
[0039] Furthermore, in step S5, a biomimetic detour coverage path is executed, specifically as follows: After entering a sub-area, the robot first travels around the outer perimeter of the sub-area to complete the boundary cleaning. Subsequently, based on the random walk algorithm, and combined with real-time judgment of cleaned and uncleaned areas, a filling-style roundabout cleaning is performed within the sub-region. Specifically, a random coverage algorithm is adopted. This algorithm first decomposes the complex sub-polygon into simple units. Then, within each unit, instead of strictly straight lines, random perturbations and intelligent connection sequences between units are introduced, thereby greatly reducing the regularity and repetition of paths while ensuring coverage.
[0040] Through the above steps, the mobile mechanism 1 moves the device on the photovoltaic module array; the cleaning execution mechanism 2 removes dirt from the surface of the photovoltaic modules; the environmental perception module 3 acquires real-time surface image information and / or dirt concentration information of the photovoltaic module array; the control unit 4 constructs and dynamically updates a global pollution situation map based on the information acquired by the environmental perception module 3. This map divides the surface of the photovoltaic array into multiple grids and assigns a comprehensive pollution level to each grid; the control unit 4 dynamically divides the entire work area into at least two sub-areas with different cleaning priorities based on the comprehensive pollution level and generates a dynamically changing area priority queue; the path planning module controls the mobile mechanism 1 to prioritize the highest priority sub-area according to the area priority queue and executes a non-preset fixed trajectory biomimetic detour coverage path within that sub-area; thus, the intelligent decision-making algorithm based on the dynamic pollution situation map and priority queue drives the robot to execute the biomimetic detour coverage path.
[0041] Example 2 Optionally, the present invention provides another embodiment, in which the device hardware implementation is configured as follows: Its control unit 4 is preferably an embedded system; The environmental perception module 3 includes a front-mounted global vision sensor and multiple local dirt detection sensors; The mobile mechanism 1 uses two sets of independently driven track wheels, driven by DC servo motors, and equipped with encoders to achieve accurate odometer calculation; The cleaning actuator 2 includes a roller brush driven by a DC motor; All modules communicate with control unit 4 via CAN bus.
[0042] The specific implementation steps of the path planning method of this invention are as follows: S201: System Initialization and Global Coarse Scan After the robot is powered on, the control unit 4 initializes each sensor and actuator; then, the control moving mechanism 1 performs a rapid global coarse scan along the edge of the photovoltaic array; during this process, the global vision sensor captures images at a fixed frequency; the image processing thread in the control unit 4 preprocesses the images, including grayscale conversion, noise reduction and image binarization based on Otsu's method, and initially distinguishes between clean and contaminated areas; S202: Constructing and updating dynamic pollution status maps Control unit 4 maintains a two-dimensional array in memory as a pollution status map; the resolution of the map can be set according to the array size. For the grid cells directly captured by the camera during coarse scanning, the pollution level value P is obtained by calculating the average gray value of the image in that area and normalizing it to the [0,1] interval; its uncertainty parameter U is initialized to a low value, such as U=0.1. For grid cells that are not directly observed but are adjacent to observed grid cells, their P-values are calculated using the inverse distance weighted interpolation method; the specific formula is as follows: ; in, The pollution level of the grid to be estimated. It is the first The pollution level of each observed neighboring grid cell. It is the Euclidean distance between the centers of two grid cells; the U of these grid cells is given a high initial value, for example, U=0.8; During subsequent detailed cleaning, the P-values of the grid within the robot's current position and sensor field of view are updated in real time to reflect the latest sensor readings; the U-values are attenuated based on observation time and distance, using the formula: ; S203: Dynamic Region Partitioning and Priority Queue Generation Control unit 4 periodically (e.g., after each sub-area is cleaned or every 30 seconds) executes the area division algorithm: First, set a high threshold. and a medium threshold The pollution status map is binarized using these two thresholds to obtain the "heavy pollution mask" and the "moderate pollution mask". Perform a connected component labeling algorithm on each mask image to identify all connected contaminated blocks; For each identified block, calculate its overall priority score. : ; in, It is the average pollution level within the block. It is the average uncertainty. This is the block area (number of grid cells), which is the weighting factor; all blocks are weighted according to... Sort the data from highest to lowest priority and store it in a priority queue; the background cleaning area is placed at the end of the queue as a default low-priority block. S204: Priority-based regional scheduling and path execution The path planning module retrieves the highest priority block from the head of the priority queue; Using A The algorithm calculates the shortest path from the robot's current position to the geometric center (centroid) of the target block; it controls the mobile mechanism 1 to travel along this path until it reaches the boundary of the target block; Upon entering the block, the robot first performs boundary following, walking around the convex hull boundary of the block. Then, it fills the block using an improved random walk algorithm. Specifically, based on its current location, the robot uses breadth-first search to detect surrounding uncleaned grids (U or P values above a threshold). Then, from these available directions, it moves towards the direction with the highest P value with a high probability, while randomly selecting other directions with a low probability, thus achieving roundabout coverage. S205: Adaptive Cleaning and Interruption Response During the cleaning process, control unit 4 monitors two key pieces of information in real time: The movement speed and cleaning intensity are dynamically adjusted based on the average P-value of the current block; for example, the following lookup table is created:
[0043] Environmental sensing module 3 continuously scans the area ahead; once it detects a localized point where the pollution level spikes instantly and exceeds [a certain threshold], [it detects a sudden increase in pollution levels]. Furthermore, since the point is not within the current task block, control unit 4 executes interrupt handling: a. Save the current path status and task progress; b. Mark the small area where the new contamination point is located as an "urgent task," and... Set the value to the maximum and insert it at the head of the priority queue; c. Immediately terminate the random walk of the current block and call A. The algorithm plans the shortest path to the emergency task point; d. Execute a compact spiral local cleanup path at the emergency task point, and restore the previously interrupted task and path state from the queue after the P value drops to a safe level; S206: Looping and Termination Repeat steps S202 to S205 until the priority queue is empty, or the P value of all grids in the entire pollution status map is lower than a preset completion threshold, which indicates that the cleaning task is successfully completed.
[0044] Specifically, in an actual photovoltaic power station test, an array of robots of this invention was deployed and compared with an adjacent array using traditional "bow" shaped path robots. Under conditions of significant bird droppings pollution, the robot of this invention reduced the total operation time by 35% by prioritizing the removal of heavily polluted areas, and the power generation efficiency of the array after cleaning was about 5% higher than that of the comparison array. At the same time, since it operated in high-speed cruising mode most of the time, the total energy consumption decreased by about 25%.
[0045] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.
Claims
1. A photovoltaic module intelligent cleaning robot, characterized in that: include: A moving mechanism (1) is used to move on the photovoltaic module array; Cleaning actuator (2) is used to remove dirt from the surface of photovoltaic modules; The environmental sensing module (3) is used to acquire surface image information and / or dirt concentration information of the photovoltaic module array in real time; The control unit (4) is communicatively connected to the moving mechanism (1), the cleaning execution mechanism (2), and the environmental sensing module (3); The control unit (4) is configured as follows: Based on the information obtained by the environmental perception module (3), a global pollution status map is constructed and dynamically updated. The map divides the photovoltaic array surface into multiple grids and assigns a comprehensive pollution level to each grid. Based on the comprehensive pollution level, the entire work area is dynamically divided into at least two sub-areas with different cleaning priorities, and a dynamically changing area priority queue is generated. The path planning module controls the mobile mechanism (1) to prioritize the current highest priority sub-region according to the regional priority queue, and executes a non-preset fixed trajectory biomimetic detour coverage path in the sub-region.
2. The intelligent cleaning robot for photovoltaic modules according to claim 1, characterized in that: When the control unit (4) is configured to build and dynamically update the pollution situation map, it is specifically used for: Based on the initial rapid scan data of the environmental sensing module (3), each grid is assigned an initial pollution level and an initial uncertainty parameter; During the cleaning process, the pollution level of the cleaned grid is updated based on the real-time monitoring data of the environmental sensing module (3), and its uncertainty parameters are reduced. For grids that are not directly observed but are adjacent to observed areas, their pollution level is estimated based on the data and spatial correlation of their neighboring grids and assigned a high uncertainty parameter.
3. The intelligent cleaning robot for photovoltaic modules according to claim 1 or 2, characterized in that: When the control unit (4) is configured to generate a dynamically changing region priority queue, it is specifically used for: Grids with pollution levels higher than the first threshold and that are spatially adjacent are merged into a first-priority sub-region. Grids with pollution levels below the first threshold but above the second threshold and that are spatially adjacent are merged into a second priority sub-region. The remaining grid cells are classified as third priority sub-regions by default. Among them, the queue order of the first priority sub-region is always higher than that of the second priority sub-region, and the queue order of the second priority sub-region is always higher than that of the third priority sub-region.
4. The intelligent cleaning robot for photovoltaic modules according to claim 3, characterized in that: The control unit (4) is further configured to: If the environmental perception module (3) detects a new pollution event during the operation of the current sub-area, causing the pollution level of a certain grid to rise sharply to exceed the first threshold, it immediately interrupts the current path, inserts the new area containing the grid as the highest priority task into the head of the area priority queue, and controls the moving mechanism (1) to go there to perform cleaning.
5. The intelligent cleaning robot for photovoltaic modules according to claim 1, characterized in that: The biomimetic detour coverage path specifically refers to controlling the moving mechanism (1) to cover the area within the boundary of the sub-region by walking along the boundary of the region and combining it with local random walks.
6. The intelligent cleaning robot for photovoltaic modules according to claim 1 or 5, characterized in that: The control unit (4) is further configured to: Based on the priority or average pollution level of the current sub-region, the working parameters of the cleaning actuator (2) and / or the moving speed of the moving mechanism (1) are dynamically adjusted; wherein, for high priority or high pollution level sub-regions, a combination mode of low moving speed and high cleaning intensity is adopted; for low priority or low pollution level sub-regions, a combination mode of high moving speed and low cleaning intensity is adopted.
7. A path planning method for an intelligent cleaning robot for photovoltaic modules, characterized in that: Includes the following steps: S1: Scan the photovoltaic module array through the environmental sensing module (3) to obtain initial pollution data; S2: Based on the initial pollution data, construct a global pollution status map; S3: Based on the pollution situation map, dynamically divide the area into sub-areas with different cleaning priorities and form an area priority queue; S4: The path planning module controls the robot to move to the highest priority sub-region based on the queue; S5: Within the current highest priority sub-region, control the robot to execute a biomimetic detour coverage path that is not a preset fixed trajectory for cleaning; S6: During the cleaning process, the pollution situation map and the area priority queue are dynamically updated based on the real-time data from the environmental perception module (3); S7: Repeat steps S4 to S6 until the pollution level of all sub-regions is lower than the preset completion threshold.
8. The path planning method for an intelligent cleaning robot for photovoltaic modules according to claim 7, characterized in that: The dynamic division of sub-regions in step S3 includes: One or more grid clusters with the highest pollution level and spatial contiguousness are identified as heavily polluted areas and given the highest priority. One or more grid clusters with moderate pollution levels and spatial contiguousness are identified as moderately polluted areas and assigned a medium priority. The remaining grid cells are classified as background cleanup areas by default and given the lowest priority.
9. The path planning method for an intelligent cleaning robot for photovoltaic modules according to claim 7, characterized in that: The dynamic update mentioned in step S6 includes: When the robot is cleaning in a sub-area, if the environmental perception module (3) discovers a new high-pollution point that does not belong to the current area and whose pollution level exceeds a certain threshold, the task replanning will be performed immediately. Pause the cleaning task in the current sub-area, set the area where the new high-contamination point is located as the temporary highest priority target, and control the robot to go there for cleaning; Once the contamination level of the temporary highest priority target drops to a safe threshold, the robot is controlled to return to the atomic region to continue the interrupted task.
10. The path planning method for an intelligent cleaning robot for photovoltaic modules according to claim 7, characterized in that: The execution of the biomimetic detour coverage path in step S5 specifically refers to: After entering a sub-area, the robot first travels around the outer perimeter of the sub-area to complete the boundary cleaning. Subsequently, based on the random walk algorithm and combined with real-time judgment of cleaned and uncleaned areas, a filling-style roundabout cleaning is performed within the sub-region.