Photovoltaic panel automatic cleaning system based on drone

Abstract

The present invention discloses an automatic cleaning system for photovoltaic panels based on drones, which relates to the field of photovoltaic technology. The system comprises: a communication link establishment module for establishing a communication network between a drone control center and a plurality of photovoltaic self-propelled cleaning robots; a real-time monitoring data acquisition module for connecting to the monitoring equipment of the photovoltaic power station, collecting images of photovoltaic panel stains and extracting stain characteristics; an initial cleaning strategy generation module for outputting an initial cleaning strategy based on the stain characteristics; a cleaning strategy synchronization module for synchronizing the initial cleaning strategy to an idle cleaning robot via a communication network; a collaborative operation setting module for setting a collaborative operation process based on the layout of the photovoltaic power station, robot position information and cleaning task allocation; and an operation adaptive adjustment module for introducing environmental conditions and setting a delayed takeoff mechanism to adaptively update the initial cleaning strategy. Thus, the technical effect of optimizing the collaboration level and improving the cleaning efficiency and risk resistance is achieved.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic technology, and in particular to an automatic cleaning system for photovoltaic panels based on an unmanned aerial vehicle (UAV). Background Art

[0002] In large-scale photovoltaic power plants, cleaning and maintenance of photovoltaic panels are crucial for ensuring efficient power generation. The layout and environment of photovoltaic panels are not conducive to traditional manual cleaning methods, and it is difficult to cover large areas of photovoltaic panels. While self-propelled cleaning robots have improved cleaning efficiency, the complexities of collaborative control between multiple robots in large-scale photovoltaic power plants make it difficult to quickly and comprehensively complete cleaning tasks. This presents technical challenges such as high coordination difficulty, low cleaning efficiency, and poor risk mitigation. Summary of the Invention

[0003] The present invention provides an automatic cleaning system for photovoltaic panels based on drones to solve the technical problems of high coordination difficulty, low cleaning efficiency and poor risk resistance in the existing technology, and achieve the technical effect of optimizing the coordination level and improving cleaning efficiency and risk resistance.

[0004] The present invention provides a photovoltaic panel automatic cleaning system based on a drone, comprising:

[0005] A communication link establishment module is used to establish a two-way communication link between the drone control center and multiple photovoltaic self-propelled cleaning robots in a target photovoltaic power station, and set up a communication network.

[0006] A real-time monitoring data acquisition module is used to connect to the real-time monitoring equipment of the target photovoltaic power station, collect photovoltaic panel stain images, and extract photovoltaic panel stain features, wherein the photovoltaic panel stain features include stain distribution features and pollution level features.

[0007] An initial cleaning strategy generation module is used to output an initial cleaning strategy based on the photovoltaic panel stain characteristics. The initial cleaning strategy includes a cleaning speed and customized cleaning tasks for key cleaning areas.

[0008] A cleaning strategy synchronization module is used to synchronize the initial cleaning strategy to the multiple photovoltaic self-propelled cleaning robots based on the communication network when the drone status information received by the drone control center is in idle standby.

[0009] The collaborative operation setting module is used to set the collaborative operation process according to the layout information of the target photovoltaic power station, the position information of the photovoltaic self-propelled cleaning robot, and the allocation information of the customized cleaning task.

[0010] The operation adaptive adjustment module is used to introduce environmental conditions. When the environmental conditions are not conducive to the cleaning operation, a delayed takeoff mechanism is set and the initial cleaning strategy is adaptively updated with the safety of the cleaning operation as a constraint condition.

[0011] In a feasible implementation, the environmental conditions include wind speed and rainfall.

[0012] The environmental conditions are used to evaluate the degree of impact on cleaning operations and to obtain impact factors.

[0013] The influencing factors and the environmental conditions are integrated to obtain an influencing index, and the influencing index is compared with a preset influencing index to determine whether to trigger the delayed takeoff mechanism.

[0014] In a feasible implementation, comparing the impact index with a preset impact index to determine whether to trigger the delayed takeoff mechanism includes:

[0015] Connecting to a data storage module, extracting the photovoltaic panel power generation efficiency of the target photovoltaic power station.

[0016] Based on the photovoltaic panel contamination characteristics and the photovoltaic panel power generation efficiency, a cleaning emergency constraint is generated, where the cleaning emergency constraint includes a cleaning priority mapping table.

[0017] The preset impact index is corrected by the cleaning emergency constraint.

[0018] In a feasible implementation, based on the photovoltaic panel stain characteristics and the photovoltaic panel power generation efficiency, a cleaning emergency constraint is generated. The cleaning emergency constraint includes a cleaning priority mapping table, including:

[0019] Based on the stain distribution characteristics in the photovoltaic panel stain characteristics and in combination with the photovoltaic panel power generation efficiency, a first associated node aggregation structure is generated.

[0020] Based on the pollution level feature in the photovoltaic panel stain feature and in combination with the photovoltaic panel power generation efficiency, a second associated node aggregation structure is generated.

[0021] The first associated node aggregation structure and the second associated node aggregation structure are hierarchically connected to construct a cleaning requirement network, and the time series features extracted from the cleaning requirement network form the cleaning priority mapping table.

[0022] In a feasible implementation, based on the stain distribution characteristics in the photovoltaic panel stain characteristics and in combination with the photovoltaic panel power generation efficiency, a first associated node aggregation structure is generated, including:

[0023] The stain aggregation feature and the stain dispersion feature are determined by the stain distribution feature in the photovoltaic panel stain feature.

[0024] Based on the photovoltaic panel cleaning example, the neighboring samples of the stain aggregation feature are searched as a type of neighbor node.

[0025] Based on the photovoltaic panel cleaning example, neighboring samples of the stain dispersion characteristics are searched as second-class neighbor nodes.

[0026] A first associated node aggregation structure is generated by combining the first-category neighbor nodes and the second-category neighbor nodes with the photovoltaic panel power generation efficiency.

[0027] In a feasible implementation method, based on a photovoltaic panel cleaning example, neighbor nodes are collected as second-level neighbor nodes based on the first-level neighbor nodes in the first class of neighbor nodes, wherein the correlation strength between the first-level neighbor nodes in the first class of neighbor nodes and the photovoltaic panel power generation efficiency is greater than the correlation strength between the second-level neighbor nodes in the first class of neighbor nodes and the photovoltaic panel power generation efficiency.

[0028] Taking the second-level neighbor nodes as a reference, repeatedly collect N-level neighbor nodes to determine the first-level neighbor nodes, second-level neighbor nodes, ..., N-level neighbor nodes in the first category of neighbor nodes.

[0029] In a feasible implementation, a first associated node aggregation structure is generated by using the first-category neighbor nodes and the second-category neighbor nodes in combination with the photovoltaic panel power generation efficiency, including:

[0030] A positive stain influence gradient is established through the first-level neighbor nodes, the second-level neighbor nodes, ..., the N-level neighbor nodes in the first type of neighbor nodes.

[0031] A reverse stain influence gradient is established through the first-level neighbor nodes, the second-level neighbor nodes, ..., the M-level neighbor nodes in the second type of neighbor nodes.

[0032] The positive stain influence gradient and the reverse stain influence gradient are merged to generate the first associated node aggregation structure.

[0033] In a feasible implementation, a collaborative operation process is set up based on the layout information of the target photovoltaic power station, the location information of the photovoltaic self-propelled cleaning robot, and the allocation information of the customized cleaning task, including:

[0034] The photovoltaic panel inclination angle of the target photovoltaic power station is obtained according to the layout information of the target photovoltaic power station.

[0035] The position information of the photovoltaic self-propelled cleaning robot is synchronized in real time through the inclination angle of the photovoltaic panel to evaluate the sideslip risk probability and sideslip distance.

[0036] The stability of the cleaning operation is evaluated by the sideslip risk probability and sideslip distance. If the cleaning operation stability threshold in the cleaning operation safety is not met, the drone flight assistance operation is activated to perform tilt angle adaptability optimization.

[0037] The present invention discloses a drone-based automatic photovoltaic panel cleaning system, comprising: a communication link establishment module, which establishes a bidirectional communication link between a drone control center and multiple self-propelled photovoltaic cleaning robots in a target photovoltaic power station, thereby setting up a communication network; a real-time monitoring data acquisition module, which connects to the photovoltaic power station's monitoring equipment, collects images of photovoltaic panel stains and extracts stain characteristics, including distribution and contamination level; an initial cleaning strategy generation module, which outputs an initial cleaning strategy based on the stain characteristics, including cleaning speed and customized cleaning tasks; a cleaning strategy synchronization module, which synchronizes the initial cleaning strategy to the cleaning robots via the communication network when the drone is idle; a collaborative operation setting module, which sets a collaborative operation process based on the photovoltaic power station layout, robot location information, and cleaning task allocation; and an adaptive operation adjustment module, which incorporates environmental conditions and, when the environment is unfavorable for cleaning operations, implements a delayed takeoff mechanism and adaptively updates the initial cleaning strategy. The disclosed drone-based automatic photovoltaic panel cleaning system solves the technical problems of high collaborative difficulty, low cleaning efficiency, and poor risk mitigation, achieving the technical effect of optimizing collaborative levels, improving cleaning efficiency, and improving risk mitigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic structural diagram of the automatic cleaning system for photovoltaic panels based on drones of the present invention;

[0039] Figure 2 The figure is a flow chart of generating emergency cleaning constraints in the automatic cleaning system of photovoltaic panels based on drones of the present invention.

[0040] Description of the accompanying drawings: communication link establishment module 11, real-time monitoring data acquisition module 12, initial cleaning strategy generation module 13, cleaning strategy synchronization module 14, collaborative operation setting module 15, operation adaptive adjustment module 16. DETAILED DESCRIPTION

[0041] The technical solutions provided in the embodiments of the present invention are designed to solve the technical problems of high coordination difficulty, low cleaning efficiency, and poor risk resistance in the prior art. The overall approach adopted is as follows:

[0042] First, the communication link establishment module is used to establish a two-way communication link between the drone control center and multiple photovoltaic self-propelled cleaning robots in the target photovoltaic power station, and set up a communication network; then, the real-time monitoring data acquisition module connects to the real-time monitoring equipment of the target photovoltaic power station, collects the photovoltaic panel stain image, and extracts the stain characteristics, including stain distribution characteristics and pollution level characteristics; then, the initial cleaning strategy generation module outputs the initial cleaning strategy based on the photovoltaic panel stain characteristics. The strategy includes the cleaning speed and the key areas of the customized cleaning task; then, the cleaning strategy synchronization module synchronizes the initial cleaning strategy to multiple photovoltaic self-propelled cleaning robots based on the communication network when the drone status information is in idle standby; finally, the collaborative operation setting module sets up the collaborative operation process according to the layout information of the target photovoltaic power station, the location information of the cleaning robot and the customized cleaning task allocation information, and introduces environmental conditions through the operation adaptive adjustment module, and adaptively updates the initial cleaning strategy with the safety of the cleaning operation as a constraint.

[0043] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention, rather than all, are shown in the drawings. Example

[0044] Figure 1 The following is a schematic diagram of the structure of the automatic cleaning system for photovoltaic panels based on drones according to the present invention, wherein the system includes:

[0045] The communication link establishing module 11 is used to establish a two-way communication link between the UAV control center and a plurality of photovoltaic self-propelled cleaning robots in the target photovoltaic power station, and set up a communication network.

[0046] Optionally, establish two-way communication between the drone control center and multiple photovoltaic self-propelled cleaning robots, including communication network architecture design, equipment connection, data transmission setting, information security protection, etc.

[0047] For example, first, select an appropriate communication protocol (such as LoRa, Zigbee, or Wi-Fi) to ensure the coverage and signal stability of the bidirectional communication link. Next, configure a communication module for each photovoltaic self-propelled cleaning robot to ensure data exchange with the drone control center. Finally, define the data transmission format and frequency, as well as information security measures, to ensure the security of the communication link and prevent unauthorized access.

[0048] The real-time monitoring data acquisition module 12 is used to connect to the real-time monitoring equipment of the target photovoltaic power station, collect photovoltaic panel stain images, and extract photovoltaic panel stain features, wherein the photovoltaic panel stain features include stain distribution features and pollution level features.

[0049] Specifically, first connect a real-time monitoring device. This device is a high-resolution camera suitable for outdoor environments, designed to capture clear images in varying lighting conditions. Optionally, this device can include multiple fixed cameras installed in appropriate locations around the PV panels, or dynamic cameras mounted on mobile inspection platforms (such as drones, balloons, or satellites) to ensure coverage of the entire PV panel area.

[0050] Specifically, according to the set image acquisition frequency (such as once per hour), images of the photovoltaic panels of the target photovoltaic power station are collected to obtain images of photovoltaic panel stains to ensure the effectiveness of real-time monitoring.

[0051] Furthermore, the PV panel stain image is preprocessed (e.g., denoising, contrast enhancement, etc.) to extract PV panel stain features. These PV panel stain features include stain distribution features and contamination level features. Optionally, the stain distribution features include the stain's area, shape, and location (including the absolute position of the PV panel and the corresponding relative position of the stain on the panel). The contamination level feature characterizes the severity of the stain on the PV panel and is determined based on the stain's area, location, type, and duration.

[0052] For example, to detect photovoltaic panel stain features, a grayscale image of the panel stain is first acquired and the Canny edge detection algorithm is used to identify the stain edges within the panel stain image. Next, an image segmentation algorithm (such as Otsu threshold segmentation or K-means clustering) is used to segment the stained area from the background, generating multiple stained area images and extracting the stain distribution features. Next, the full-color images corresponding to the multiple stained area images are retrieved to obtain the stain's color features. Combined with the stain distribution features, the stain's contamination level is determined.

[0053] Through the above steps, real-time monitoring and feature extraction of photovoltaic panel stains can be achieved, so that stain problems can be identified and handled in a timely manner to ensure the efficient operation of photovoltaic power stations.

[0054] The initial cleaning strategy generating module 13 is configured to output an initial cleaning strategy based on the photovoltaic panel stain characteristics. The initial cleaning strategy includes a cleaning speed and customized cleaning tasks for key cleaning areas.

[0055] Specifically, the initial cleaning strategy includes cleaning speed and customized cleaning tasks for key areas. The initial cleaning strategy is determined based on the characteristics of photovoltaic panel stains. For example, the cleaning speed of higher-level or more difficult-to-clean stains (such as oil stains or dirt) is slower to ensure the cleaning effect. For lighter stains such as dust, a faster cleaning speed is set to improve work efficiency.

[0056] Specifically, customized cleaning tasks are set based on the specific stain characteristics of different PV panel areas. This includes increasing cleaning intensity or extending cleaning time for areas likely to accumulate more dirt, such as edges, older construction (where the PV panel surface is rougher), or horizontal panels. Panel areas with higher maintenance requirements are prioritized.

[0057] Optionally, after the cleaning task is completed, the cleaning effect is evaluated, and the cleaning efficiency and stain removal rate are recorded to further improve and adjust the cleaning strategy. Examples include modifying the cleaning speed, adjusting the cleaning priority, adjusting customized cleaning tasks, and optimizing the division of key cleaning areas.

[0058] The cleaning strategy synchronization module 14 is configured to synchronize the initial cleaning strategy to the plurality of photovoltaic self-propelled cleaning robots based on the communication network when the drone status information received by the drone control center is in idle standby.

[0059] Specifically, the drone control center synchronizes the initial cleaning strategy to multiple PV self-propelled cleaning robots in idle state in real time via the communication network. If the drone's status is not idle, the initial cleaning strategy is synchronized with the PV self-propelled cleaning robots in idle state after the current task is completed. This ensures that the initial cleaning strategy is applied efficiently and promptly to multiple PV self-propelled cleaning robots, while also ensuring that the robots currently performing their tasks are not disturbed, maintaining their efficiency and safety.

[0060] The collaborative operation setting module 15 is used to set the collaborative operation process according to the layout information of the target photovoltaic power station, the position information of the photovoltaic self-propelled cleaning robot, and the allocation information of the customized cleaning task.

[0061] Specifically, collaborative operation involves a PV self-propelled cleaning robot and a drone working together to complete cleaning tasks based on the real-time posture characteristics of the PV panels, ensuring that all critical areas are cleaned promptly and effectively. This is because the angles and orientations of PV panels in actual PV power plants vary, and these different angles and orientations can affect the operational safety, cleaning efficiency, and effectiveness of the PV self-propelled cleaning robot. Furthermore, due to location constraints, PV self-propelled cleaning robots may struggle to meet the timeliness requirements of customized cleaning tasks in key areas within the specified timeframe.

[0062] In some embodiments, a collaborative operation process is set based on the layout information of the target photovoltaic power station, the location information of the photovoltaic self-propelled cleaning robot, and the allocation information of the customized cleaning task, including:

[0063] The photovoltaic panel inclination angle of the target photovoltaic power station is obtained according to the layout information of the target photovoltaic power station.

[0064] The position information of the photovoltaic self-propelled cleaning robot is synchronized in real time through the inclination angle of the photovoltaic panel to evaluate the sideslip risk probability and sideslip distance.

[0065] The stability of the cleaning operation is evaluated by the sideslip risk probability and sideslip distance. If the cleaning operation stability threshold in the cleaning operation safety is not met, the drone flight assistance operation is activated to perform tilt angle adaptability optimization.

[0066] Specifically, the system first acquires the inclination angle data of each photovoltaic panel in real time based on the layout of the target photovoltaic power station. Then, the system synchronizes the position of the self-propelled photovoltaic cleaning robot in real time and combines it with the inclination angle of the photovoltaic panels to determine the real-time inclination angle of the self-propelled photovoltaic cleaning robot.

[0067] Furthermore, based on the panel inclination angle, the sideslip risk probability and sideslip distance are calculated. For example, based on the ideal friction coefficient between the photovoltaic self-propelled cleaning robot and the photovoltaic panel (the friction coefficient of a clean photovoltaic panel), the ideal critical condition for sideslip is calculated to obtain the upper limit sideslip critical angle; then, based on the most unfavorable friction coefficient between the photovoltaic self-propelled cleaning robot and the photovoltaic panel (such as the friction coefficient of a photovoltaic panel covered with simulated dust), the most unfavorable critical condition for sideslip is calculated to obtain the lower limit sideslip critical angle; then, based on the lower limit sideslip critical angle and the upper limit sideslip critical angle, a sideslip risk interval is constructed. If the photovoltaic panel inclination angle is less than the lower limit sideslip critical angle, it can be considered that there is no sideslip risk. Correspondingly, in the sideslip risk interval, the larger the photovoltaic panel inclination angle (the closer to the upper limit sideslip critical angle), the greater the sideslip risk.

[0068] Furthermore, if the photovoltaic panel is located in the above-mentioned sideslip risk range, the sideslip risk probability of the photovoltaic self-propelled cleaning robot is calculated based on the relative position of the photovoltaic panel inclination angle in the sideslip risk range; at the same time, the sideslip distance is obtained, which determines the impact force received by the photovoltaic self-propelled cleaning robot when it sideslips to the boundary of the photovoltaic panel. If the sideslip distance is too large, the photovoltaic self-propelled cleaning robot is at risk of overturning.

[0069] The safety of the cleaning operation is then assessed using a weighted approach based on the calculated sideslip risk probability and sideslip distance. The cleaning operation stability threshold is a confidence threshold determined based on statistical analysis of sideslip cases. For example, the cleaning operation stability threshold is defined as the 95% confidence level for sideslip. If sideslip risk exists, appropriate safety measures must be implemented, such as adjusting the robot's travel path, reducing operating speed, or using a drone for tilt adjustment.

[0070] The operation adaptive adjustment module 16 is used to introduce environmental conditions. When the environmental conditions are not conducive to the cleaning operation, a delayed takeoff mechanism is set and the initial cleaning strategy is adaptively updated with the safety of the cleaning operation as a constraint condition.

[0071] In some embodiments, the environmental conditions include wind speed and rainfall.

[0072] The environmental conditions are used to evaluate the degree of impact on cleaning operations and to obtain impact factors.

[0073] The influencing factors and the environmental conditions are integrated to obtain an influencing index, and the influencing index is compared with a preset influencing index to determine whether to trigger the delayed takeoff mechanism.

[0074] Specifically, sensors are used to monitor relevant environmental conditions in real time, such as wind speed and rainfall. Then, based on the monitored environmental conditions, the degree of impact of these conditions on the cleaning operation is analyzed based on regression, and multiple influencing factors are obtained. Then, the influencing factors are used as fusion weights to obtain the impact index under the current environmental conditions. The impact index is used to quantitatively represent the impact of the environment on the safety of the cleaning operation. If the impact index is greater than the preset impact index, the drone's takeoff needs to be delayed to ensure safety.

[0075] By following these steps, you can ensure the safety and effectiveness of cleaning operations under adverse environmental conditions, reduce potential risks, and improve overall operational efficiency.

[0076] In some embodiments, as Figure 2 As shown, the impact index is compared with a preset impact index to determine whether to trigger the delayed takeoff mechanism, including:

[0077] Connecting to a data storage module, extracting the photovoltaic panel power generation efficiency of the target photovoltaic power station.

[0078] Based on the photovoltaic panel contamination characteristics and the photovoltaic panel power generation efficiency, a cleaning emergency constraint is generated, where the cleaning emergency constraint includes a cleaning priority mapping table.

[0079] The preset impact index is corrected by the cleaning emergency constraint.

[0080] Specifically, the system first connects to the data storage module to extract the target PV power plant's photovoltaic panel power generation efficiency data. This efficiency data can indirectly reflect the necessity of the current cleaning operation. Specifically, the photovoltaic panel power generation efficiency is affected by both contamination and ambient light intensity. Next, based on the contamination characteristics of the photovoltaic panel (such as contamination type and coverage area), the impact on power generation efficiency is evaluated to obtain the cleaning emergency constraint. For example, the cleaning emergency constraint is the efficiency reduction impact corresponding to the contamination characteristics of the photovoltaic panel.

[0081] Specifically, the cleaning urgency constraint further includes a cleaning priority mapping table, which is used to map different stains to corresponding cleaning priorities. A larger stain corresponds to a higher cleaning priority.

[0082] Furthermore, the cleaning emergency constraint information is used to correct the preset impact index to more accurately reflect the urgency of the cleaning operation. The higher the cleaning priority of the stain, the smaller the preset impact index corresponding to it. This ensures that when the cleaning operation is affected by the external environment, it can still give priority to power generation efficiency and optimize the cleaning strategy.

[0083] In some embodiments, based on the photovoltaic panel contamination characteristics and the photovoltaic panel power generation efficiency, a cleaning emergency constraint is generated, and the cleaning emergency constraint includes a cleaning priority mapping table, including:

[0084] Based on the stain distribution characteristics in the photovoltaic panel stain characteristics and in combination with the photovoltaic panel power generation efficiency, a first associated node aggregation structure is generated.

[0085] Based on the pollution level feature in the photovoltaic panel stain feature and in combination with the photovoltaic panel power generation efficiency, a second associated node aggregation structure is generated.

[0086] The first associated node aggregation structure and the second associated node aggregation structure are hierarchically connected to construct a cleaning requirement network, and the time series features extracted from the cleaning requirement network form the cleaning priority mapping table.

[0087] Specifically, based on the stain distribution data on photovoltaic panels and their power generation efficiency, a graph neural network (GraphSAGE) was used to generate a first-association node aggregation structure. In this structure, each node represents a stain distribution feature, and edges connect to corresponding power generation efficiency values, forming a dynamic weighted graph.

[0088] Specifically, the impact of the pollution level of the photovoltaic panels (e.g., mild, moderate, or severe pollution) on power generation efficiency was evaluated. GraphSAGE was then used to generate a second, connected node aggregation structure describing the relationship between pollution level and power generation efficiency. Each pollution level node was connected to other nodes based on its characteristics and power generation efficiency loss, forming a connected network.

[0089] Furthermore, the first associated node aggregation structure is hierarchically connected with the second associated node aggregation structure to form a cleaning demand network. In this cleaning demand network, the connections between nodes reflect the complex relationships of cleaning needs, including multi-dimensional information such as stain characteristics, pollution level, and power generation efficiency. Through the aggregation function of GraphSAGE, the above steps can extract features at each level, thereby forming a higher-level representation of cleaning needs.

[0090] Furthermore, time series features are extracted from the cleaning demand network to analyze changes in soiling and fluctuations in power generation efficiency, providing a dynamic basis for cleaning decisions. This allows the construction of a cleaning priority mapping table to prioritize cleaning tasks. For example, panels with high soiling and low power generation efficiency are prioritized for cleaning.

[0091] In some embodiments, based on the stain distribution characteristics in the photovoltaic panel stain characteristics and in combination with the photovoltaic panel power generation efficiency, generating a first associated node aggregation structure includes:

[0092] The stain aggregation feature and the stain dispersion feature are determined by the stain distribution feature in the photovoltaic panel stain feature.

[0093] Based on the photovoltaic panel cleaning example, the neighboring samples of the stain aggregation feature are searched as a type of neighbor node.

[0094] Based on the photovoltaic panel cleaning example, neighboring samples of the stain dispersion characteristics are searched as second-class neighbor nodes.

[0095] A first associated node aggregation structure is generated by combining the first-category neighbor nodes and the second-category neighbor nodes with the photovoltaic panel power generation efficiency.

[0096] Specifically, the system first analyzes the distribution of dirt on the photovoltaic panel to identify concentration characteristics (e.g., dense areas of dirt) and dispersion characteristics (e.g., sparse distribution of dirt). Based on the concentration characteristics, it then retrieves historical cleaning examples with similar characteristics. These examples serve as references for understanding the impact of concentration on power generation efficiency. Similarly, based on the dispersion characteristics, it retrieves similar historical cleaning examples to explore the impact of dispersion on power generation efficiency.

[0097] Specifically, the first and second category neighbor nodes are integrated to form a preliminary node set. Each node contains soiling characteristics and corresponding cleaning effect and power generation efficiency data. This information is then combined with the node set information to calculate a weighted value for each node based on its actual impact on power generation efficiency. This creates a first associated node aggregation structure, which effectively represents the relationship between soiling distribution characteristics and photovoltaic panel power generation efficiency, supporting subsequent cleaning strategy optimization.

[0098] In some embodiments, based on a photovoltaic panel cleaning example, neighbor nodes are collected as second-level neighbor nodes based on the first-level neighbor nodes in the first class of neighbor nodes, wherein the correlation strength between the first-level neighbor nodes in the first class of neighbor nodes and the photovoltaic panel power generation efficiency is greater than the correlation strength between the second-level neighbor nodes in the first class of neighbor nodes and the photovoltaic panel power generation efficiency.

[0099] Taking the second-level neighbor nodes as a reference, repeatedly collect N-level neighbor nodes to determine the first-level neighbor nodes, second-level neighbor nodes, ..., N-level neighbor nodes in the first category of neighbor nodes.

[0100] Specifically, from a first-class neighboring node, nodes with the strongest correlation with the photovoltaic panel's power generation efficiency are selected as first-class neighboring nodes. These nodes represent the clean instances with the strongest correlation to the current stain characteristics. Next, using the first-class neighboring nodes as a benchmark, the more highly correlated neighboring nodes are analyzed. The correlation strength of these nodes needs to be smaller than the correlation between the first-class neighboring nodes and power generation efficiency, but still maintain a strong correlation. These nodes are then collected as second-class neighboring nodes, forming a broader sample set for further analysis.

[0101] Specifically, starting with each second-level neighbor node, the data is collected from its neighboring nodes to generate third-level neighbor nodes. This process continues until the preset N levels of neighbor nodes are reached. During the node collection process, the correlation strength between each level of node and power generation efficiency decreases gradually.

[0102] The complete neighbor node hierarchy constructed through the above steps can provide more data support for analyzing the cleaning needs of photovoltaic panels and optimizing cleaning strategies, helping cleaning strategies to better adapt to the complex relationship between different stain characteristics and power generation efficiency, thereby improving the overall cleaning efficiency and power generation performance of photovoltaic power stations.

[0103] In some embodiments, generating a first associated node aggregation structure by using the first type of neighbor nodes and the second type of neighbor nodes in combination with the photovoltaic panel power generation efficiency includes:

[0104] A positive stain influence gradient is established through the first-level neighbor nodes, the second-level neighbor nodes, ..., the N-level neighbor nodes in the first type of neighbor nodes.

[0105] A reverse stain influence gradient is established through the first-level neighbor nodes, the second-level neighbor nodes, ..., the M-level neighbor nodes in the second type of neighbor nodes.

[0106] The positive stain influence gradient and the reverse stain influence gradient are merged to generate the first associated node aggregation structure.

[0107] Specifically, a positive contamination impact gradient is established for each type of neighboring node (first-level neighboring nodes, second-level neighboring nodes, and so on, up to N-level neighboring nodes). This positive contamination impact gradient determines the extent to which contamination concentration affects the efficiency of photovoltaic panels by calculating each node's contribution to pollution. For example, the positive contamination impact gradient decreases as the distance between nodes increases.

[0108] Specifically, based on the same idea, a reverse contamination influence gradient is established using two types of neighbor nodes (first-level neighbor nodes, second-level neighbor nodes, ..., M-level neighbor nodes). This gradient can be expressed as the contribution of each node corresponding to the contamination dispersion to the photovoltaic panel's power generation efficiency.

[0109] Specifically, first, the magnitudes of the positive and negative contamination gradients are compared and normalized so that they can be compared on the same scale. Next, the positive and negative contamination gradients are merged, such as by simply adding them together or performing a weighted average. Finally, the merged gradients are used to generate the first associated node aggregation structure.

[0110] In summary, the drone-based photovoltaic panel automatic cleaning system provided by the present invention has the following technical effects:

[0111] The communication link establishment module establishes a two-way communication link between the drone control center and multiple self-propelled photovoltaic cleaning robots in the target photovoltaic power station, setting up a communication network. The real-time monitoring data acquisition module connects to the photovoltaic power station's monitoring equipment, collects images of photovoltaic panel stains, and extracts stain characteristics, including distribution and contamination level. The initial cleaning strategy generation module outputs an initial cleaning strategy based on the stain characteristics, including cleaning speed and customized cleaning tasks. The cleaning strategy synchronization module synchronizes the initial cleaning strategy to the cleaning robots via the communication network when the drone is idle. The collaborative operation setting module sets the collaborative operation process based on the photovoltaic power station layout, robot location information, and cleaning task allocation. The operation adaptive adjustment module introduces environmental conditions. When the environment is not conducive to cleaning operations, it sets a delayed takeoff mechanism and adaptively updates the initial cleaning strategy to ensure operation safety. This achieves the technical effect of optimizing the level of collaboration, improving cleaning efficiency and risk resistance.

[0112] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the above embodiments or replace some of the technical features therein with equivalents; and such modifications or replacements do not deviate from the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention and are all included in the scope of protection of the present invention.

Claims

1. The automatic cleaning system of photovoltaic panels based on drones is characterized by: The system is in communication with a drone control center, which is used to receive drone status information, including: a communication link establishing module, the communication link establishing module being used to establish a two-way communication link between the drone control center and a plurality of photovoltaic self-propelled cleaning robots at a target photovoltaic power station, thereby setting up a communication network; A real-time monitoring data acquisition module, which is used to connect to the real-time monitoring equipment of the target photovoltaic power station, collect photovoltaic panel stain images, and extract photovoltaic panel stain features, wherein the photovoltaic panel stain features include stain distribution features and pollution level features; An initial cleaning strategy generation module, the initial cleaning strategy generation module is used to output an initial cleaning strategy based on the photovoltaic panel stain characteristics, the initial cleaning strategy including a cleaning speed and customized cleaning tasks for key cleaning areas; a cleaning strategy synchronization module, configured to synchronize the initial cleaning strategy to the plurality of photovoltaic self-propelled cleaning robots based on the communication network when the drone status information received by the drone control center is in idle standby; A collaborative operation setting module, which is used to simultaneously set a collaborative operation process based on the layout information of the target photovoltaic power station, the location information of the photovoltaic self-propelled cleaning robot, and the allocation information of the customized cleaning task; An operation adaptive adjustment module, which is used to introduce environmental conditions and, when the environmental conditions are not conducive to the cleaning operation, set a delayed takeoff mechanism and use the safety of the cleaning operation as a constraint to adaptively update the initial cleaning strategy; The environmental conditions include wind speed and rainfall; Evaluate the degree of impact on cleaning operations based on the environmental conditions and obtain impact factors; fusing the influencing factors and the environmental conditions to obtain an influencing index, and comparing the influencing index with a preset influencing index to determine whether to trigger the delayed takeoff mechanism; The impact index is compared with a preset impact index to determine whether to trigger the delayed takeoff mechanism, including: Connecting to a data storage module to extract the photovoltaic panel power generation efficiency of the target photovoltaic power station; generating a cleaning emergency constraint based on the photovoltaic panel stain characteristics and the photovoltaic panel power generation efficiency, wherein the cleaning emergency constraint includes a cleaning priority mapping table; Correcting the preset impact index by using the cleaning emergency constraint; Based on the photovoltaic panel stain characteristics and the photovoltaic panel power generation efficiency, a cleaning emergency constraint is generated, wherein the cleaning emergency constraint includes a cleaning priority mapping table, including: Based on the stain distribution characteristics in the photovoltaic panel stain characteristics and in combination with the photovoltaic panel power generation efficiency, a first associated node aggregation structure is generated; Based on the pollution level feature in the photovoltaic panel stain feature and in combination with the photovoltaic panel power generation efficiency, a second associated node aggregation structure is generated; hierarchically connecting the first associated node aggregation structure and the second associated node aggregation structure to construct a cleaning demand network, wherein the time series features extracted from the cleaning demand network form the cleaning priority mapping table; Based on the stain distribution characteristics in the photovoltaic panel stain characteristics and in combination with the photovoltaic panel power generation efficiency, a first associated node aggregation structure is generated, including: Determining the stain aggregation characteristics and the stain dispersion characteristics according to the stain distribution characteristics in the photovoltaic panel stain characteristics; Based on the photovoltaic panel cleaning example, the neighboring samples of the stain aggregation feature are searched as a type of neighbor node; Based on the photovoltaic panel cleaning example, the neighboring samples of the stain dispersion characteristics are searched as the second-class neighbor nodes; Generate a first associated node aggregation structure by combining the first-class neighbor nodes and the second-class neighbor nodes with the photovoltaic panel power generation efficiency; Based on a photovoltaic panel cleaning example, taking the first-level neighbor nodes in the first category of neighbor nodes as a benchmark, collecting neighbor nodes as second-level neighbor nodes, wherein the correlation strength between the first-level neighbor nodes in the first category of neighbor nodes and the photovoltaic panel power generation efficiency is greater than the correlation strength between the second-level neighbor nodes in the first category of neighbor nodes and the photovoltaic panel power generation efficiency; Taking the second-level neighbor node as a reference, repeatedly collecting N-level neighbor nodes to determine the first-level neighbor node, the second-level neighbor node, ..., the N-level neighbor node in the first category of neighbor nodes; By using the first-category neighbor nodes and the second-category neighbor nodes and combining the photovoltaic panel power generation efficiency, a first associated node aggregation structure is generated, including: Establishing a positive stain influence gradient through the first-level neighbor nodes, the second-level neighbor nodes, ..., the N-level neighbor nodes in the first type of neighbor nodes; Establishing a reverse stain influence gradient through the first-level neighbor nodes, the second-level neighbor nodes, ..., the M-level neighbor nodes in the second-type neighbor nodes; Merging the positive stain influence gradient and the negative stain influence gradient to generate the first associated node aggregation structure; According to the layout information of the target photovoltaic power station, the location information of the photovoltaic self-propelled cleaning robot, and the allocation information of the customized cleaning task, a collaborative operation process is set up, including: Acquiring the photovoltaic panel inclination angle of the target photovoltaic power station according to the layout information of the target photovoltaic power station; The photovoltaic panel inclination angle is used to synchronize the position information of the photovoltaic self-propelled cleaning robot in real time to assess the side slip risk probability and side slip distance; The stability of the cleaning operation is evaluated by the sideslip risk probability and sideslip distance. If the cleaning operation stability threshold in the cleaning operation safety is not met, the drone flight assistance operation is activated to perform tilt angle adaptability optimization.

Images