A photovoltaic module self-adaptive cleaning method, system, computer device and medium

Abstract

The application provides a photovoltaic module adaptive cleaning method, system, computer device and medium, and belongs to the field of photovoltaic module automatic cleaning control. The method comprises the following steps: dividing the photovoltaic module into operation units, generating a pollution judgment graph by fusing unmanned aerial vehicle aerial photography and unmanned vehicle shooting images, obtaining each unit pollution pixel set through spatial mapping, and statistically and weightedly aggregating to obtain comprehensive weighted pollution intensity and pollution coverage ratio; generating control parameters of unmanned aerial vehicle spray washing and unmanned vehicle rolling brush cleaning according to the pollution judgment graph, and sequentially executing spray washing and rolling brush operation; reacquiring images after cleaning to update the pollution intensity and the coverage ratio, determining units exceeding the standard as units not meeting the standard, extracting residual pollution areas and generating cleaning control parameters to execute local cleaning until all units meet the standard. The application realizes detection, decision, execution and re-detection, solves the problems of over-cleaning and under-cleaning coexisting in the prior art, lack of quality verification and cleaning ability, and improves the cleaning refinement level and resource utilization efficiency.

Description

Technical Field

[0001] This invention belongs to the field of automated cleaning control of photovoltaic modules, and specifically relates to an adaptive cleaning method, system, computer equipment and medium for photovoltaic modules. Background Technology

[0002] Photovoltaic modules are exposed to the outdoor environment for extended periods, making their surfaces highly susceptible to the accumulation of contaminants such as dust, bird droppings, oil, and water stains. These contaminants significantly reduce the light transmittance of the module's glass cover, resulting in reduced light intensity reaching the solar cells and directly causing a decrease in the power generation efficiency of the photovoltaic power station. Studies have shown that in areas with heavy dust storms or industrial pollution, annual power generation losses due to dust accumulation in photovoltaic power stations can range from 10% to 30%. Furthermore, persistent localized contamination can easily trigger hot spot effects, further reducing power output and potentially burning out solar cells, shortening module lifespan, and increasing operation and maintenance costs. Therefore, regular, efficient, and thorough cleaning of photovoltaic module surfaces is a crucial aspect of operation and maintenance to ensure the power generation revenue of photovoltaic power stations and extend equipment lifespan.

[0003] Existing photovoltaic cleaning technologies generally adopt an open-loop control mode: First, drones or ground vision systems are used to detect and identify surface contamination of photovoltaic modules, generating a contamination distribution map or cleanliness index. However, the detection results are only used as a reference for inspection or as a basis for judging the start of cleaning, and are not directly linked to the cleaning process. Subsequently, the cleaning robot performs a one-time cleaning of the entire row of modules according to a preset fixed trajectory and constant parameters (such as travel speed, brush rotation speed, water flow rate, etc.). The operation ends when the path is completed, lacking a mechanism for verifying the cleaning effect and making decisions on re-cleaning.

[0004] The fixed parameters of this traditional approach cannot adapt to the spatial differences in pollution levels, resulting in over-cleaning of lightly polluted areas and under-cleaning of heavily polluted areas. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides an adaptive cleaning method, system, computer device, and medium for photovoltaic modules.

[0006] To achieve the above objectives, the present invention provides an adaptive cleaning method for photovoltaic modules, comprising: The photovoltaic modules to be cleaned are divided into multiple work units, and aerial images of the photovoltaic modules to be cleaned are collected by drones and unmanned vehicles.

[0007] Edge texture features are extracted from two images respectively, and contaminated pixels are identified based on these features. The contamination type and degree of each contaminated pixel are determined, and a fused contamination determination map is generated based on the contaminated pixels in the two images. The fused contamination determination map is spatially mapped to each work unit to obtain a set of contaminated pixels associated with each work unit. Based on the set of contaminated pixels in each work unit, the number of contaminated pixels is counted, and the overall weighted contamination intensity and contamination coverage ratio of each work unit are obtained by weighted aggregation based on the contamination type and degree. For each overall weighted contamination intensity, corresponding UAV spraying control parameters and unmanned vehicle roller brush cleaning control parameters are generated for the work unit. Based on the UAV spraying control parameters and unmanned vehicle roller brush cleaning control parameters, UAV spraying and unmanned vehicle roller brush cleaning are executed sequentially within the same unit.

[0008] Once all units have been sprayed and brushed, one cleaning cycle is complete.

[0009] Preferably, after all units have completed spraying and brushing, and one round of cleaning is finished, the process further includes: cyclically acquiring two images, updating the comprehensive weighted pollution intensity and pollution coverage ratio of each work unit, identifying work units with a pollution coverage ratio exceeding a threshold as substandard units; extracting residual pollution areas from substandard units, generating re-cleaning control parameters based on the updated comprehensive weighted pollution intensity, and performing local re-cleaning according to the residual pollution areas; and ending cleaning when all units meet the standards. The re-cleaning control parameters are recalculated based on the updated comprehensive weighted pollution intensity, or can be generated by multiplying the previous UAV spraying control parameters or UAV brush cleaning control parameters by an enhancement coefficient, wherein the enhancement coefficient is determined based on the degree of pollution in the residual pollution areas.

[0010] Preferably, the drone spraying control parameters include the drone's water flow rate, foam flow rate, and spraying residence time; the unmanned vehicle roller brush cleaning control parameters include the unmanned vehicle's roller brush rotation speed, travel speed, and roller brush downward pressure; wherein, the water flow rate, foam flow rate, spraying residence time, roller brush rotation speed, and roller brush downward pressure are positively correlated with the pollution intensity, and the travel speed is negatively correlated with the pollution intensity.

[0011] Preferably, the step of extracting residual contamination areas from non-compliant units specifically includes: acquiring the latest drone aerial images and unmanned vehicle images, generating an updated fused contamination determination map; extracting the set of residual contamination pixels within the non-compliant units from the latest fused contamination determination map, performing connected component analysis on the set of residual contamination pixels, and obtaining a set of residual contamination areas.

[0012] Preferably, for each comprehensive weighted pollution intensity, the generation of corresponding UAV spraying control parameters and unmanned vehicle roller brush cleaning control parameters for the corresponding work unit specifically includes: calculating the UAV water spray flow rate, foam spray flow rate, and spraying residence time based on the comprehensive weighted pollution intensity and a preset minimum value and a first preset adjustment coefficient; and calculating the unmanned vehicle roller brush rotation speed, travel speed, and roller brush downward pressure based on the comprehensive weighted pollution intensity and a preset extreme value and a second preset adjustment coefficient.

[0013] Preferably, the first preset adjustment coefficient includes the water spray adjustment gain coefficient, foam adjustment gain coefficient, and time adjustment gain coefficient for drone spray washing, and the second preset adjustment coefficient includes the rotation speed adjustment coefficient, speed adjustment coefficient, and pressure adjustment coefficient for unmanned vehicle roller brush cleaning.

[0014] Preferably, the pollution type includes at least one of dust, water stains, bird droppings and oil stains, with different pollution types corresponding to different preset weights; the pollution degree is classified into levels according to the adhesion strength or coverage thickness of the pollutants, with different levels corresponding to different preset weights.

[0015] The present invention also provides an adaptive cleaning system for photovoltaic modules, comprising: The data acquisition module is used to divide the photovoltaic modules to be cleaned into multiple work units and collect aerial images of the photovoltaic modules to be cleaned by drones and images taken by unmanned vehicles.

[0016] The calculation module is used to extract edge texture features from two images respectively, identify contaminated pixels based on the edge texture features, determine the contamination type and degree of each contaminated pixel, and generate a fused contamination determination map based on the contaminated pixels in the two images; map the fused contamination determination map to the spatial mapping of each work unit to obtain a set of contaminated pixels associated with each work unit; count the number of contaminated pixels based on the set of contaminated pixels in each work unit, and combine the contamination type and degree with weighted aggregation to obtain the comprehensive weighted contamination intensity and contamination coverage ratio of each work unit; for each comprehensive weighted contamination intensity, generate the corresponding UAV spraying control parameters and UAV roller brush cleaning control parameters for the work unit; based on the UAV spraying control parameters and UAV roller brush cleaning control parameters, execute UAV spraying and UAV roller brush cleaning sequentially within the same unit.

[0017] The cleaning module is used to end a cleaning cycle after all units have finished spraying and brushing.

[0018] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any of the steps in the photovoltaic module adaptive cleaning method.

[0019] The present invention also provides a computer-readable storage medium storing a computer program that, when loaded by a processor, is capable of executing any of the steps in the photovoltaic module adaptive cleaning method.

[0020] The adaptive cleaning method for photovoltaic modules provided by this invention has the following beneficial effects: This invention divides photovoltaic modules into working units, integrates air and ground dual-view images to identify pollution type and degree, and weights and aggregates them to obtain a comprehensive weighted pollution intensity and pollution coverage ratio. Based on this, adaptive control parameters for drone spraying and unmanned vehicle roller brushing are generated differently for each unit, realizing on-demand cleaning. This fundamentally avoids the problem of over-cleaning in lightly polluted areas and under-cleaning in heavily polluted areas, upgrading photovoltaic cleaning from open-loop fixed parameter operation to data-driven intelligent control, significantly improving cleaning consistency, resource utilization efficiency and automation level. Attached Figure Description

[0021] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of an adaptive cleaning method for photovoltaic modules according to an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0024] This invention proposes a comprehensive intelligent enhancement method for the detection, decision-making, and execution of a heterogeneous photovoltaic cleaning system based on air-ground architecture. The cleaning execution is collaboratively performed by a drone and an unmanned vehicle: the drone is used for pre-wetting the panel surface with water spray and applying cleaning foam, while the unmanned vehicle performs mechanical brushing of the panel surface. This method integrates panel-level contamination distribution maps. Using the photovoltaic surface cleanliness index CISS as a unified sensing and evaluation input, a parameter generation mechanism for coordinated control of aerial spraying and ground scrubbing is constructed to realize a complete control logic of pollution sensing, cleaning execution, effect verification, and re-cleaning control.

[0025] Based on this, the present invention provides an adaptive cleaning method for photovoltaic modules, specifically as follows: Figure 1 As shown, it includes the following steps: S1. Input acquisition and spatial consistency processing of evaluation results Before the cleaning operation begins, the system directly acquires the contamination distribution and quantitative evaluation data on the surface of the photovoltaic modules, which serves as the input to the cleaning decision module of this invention. The inputs include a fusion contamination determination map. Comprehensive weighted pollution intensity at the operational unit level Panel-level cleanliness evaluation indicators .in, This represents the pixels or grid points on the surface of the photovoltaic array in a unified world coordinate system. Indicates the number of divisions already made One cleaning operation unit.

[0026] To ensure consistency between cleaning control and evaluation results, this step spatially maps the contamination map to the work unit index, so that each work unit... All are associated with corresponding pollution distribution information and overall pollution intensity. .

[0027] S2. Generation of UAV water spraying and foam cleaning parameters based on weighted pollution intensity The pollution intensity is calculated by combining the work units input in step one. As a control state variable, this invention constructs a model for generating control parameters for UAV spraying and washing.

[0028] Drones in the work unit When the water spray cleaning is performed above, the water spray flow control parameters are set as follows: ; in, For drones in the work unit The internal water flow rate Let l be the l-th work unit, where l is the physical location index of the work unit, l=1,2,3,…,L (L is the total number of units), and all l have this meaning. The minimum allowable water flow rate of the system is w, where w is the water flow rate. The gain coefficient for water spray adjustment; This represents the weighted pollution intensity obtained in step one.

[0029] To enhance the removal capabilities of oil stains, water stains, and adhesive contaminants, this invention simultaneously controls the drone to spray cleaning foam, with the foam spray flow rate set as follows: ; in, For foam spraying flow rate; Minimum foam spray flow rate; This is the gain coefficient for foam adjustment, where f stands for Foam.

[0030] To ensure adequate spray coverage, the effective spray dwell time for the drone within this work unit is set as follows: ; in, This refers to the spraying residence time; Minimum stay time; The gain coefficient is adjusted for time, u is the abbreviation for UAV (Unmanned Aerial Vehicle), and t is time.

[0031] This constitutes a three-parameter linkage adjustment mechanism for the drone's water spray, foam, and dwell time.

[0032] S3. Generation of unmanned vehicle roller brush cleaning parameters based on weighted pollution intensity After the drone completes the spraying operation, the unmanned vehicle carries the roller brush device into the corresponding work unit to perform mechanical brushing.

[0033] Similarly, the weighted pollution intensity of the input is used. As a control basis, the brush rotation speed is set as follows: ; in, This refers to the rotational speed of the roller brush. Minimum rotational speed; is the speed adjustment coefficient, and n is the speed.

[0034] The travel speed of the unmanned vehicle within this work unit is set as follows: ; in, Speed ​​of travel; The maximum permissible speed; , where v is the speed adjustment coefficient, and v represents the speed.

[0035] The pressure exerted by the roller brush on the surface of the photovoltaic module is set as follows: ; in, This refers to the downward pressure of the roller brush; Minimum downforce; F represents the pressure regulation coefficient and the pressure.

[0036] Therefore, a mechanical scrubbing control strategy based on adaptive adjustment of pollution intensity is constructed.

[0037] S4, Air-Ground Coordinated Cleaning Scheduling Execution The control system schedules drones and unmanned vehicles to work together according to the order of the work unit index.

[0038] Within the same work unit, drones are prioritized to complete water spraying and foam pretreatment, followed by unmanned vehicles to complete roller brush cleaning, so that the foam can fully soften the contaminants and reduce the brushing resistance.

[0039] Both the drone flight path and the unmanned vehicle travel path are based on work units. The spatial boundary serves as a constraint, ensuring that the two operating methods can be executed collaboratively within the same spatial area.

[0040] S5. Citation of evaluation results after cleaning and acquisition of re-inspection input. After all work units have been cleaned, the system's visual perception module scans and analyzes the cleaned area again to obtain a contamination assessment map. and cleanliness evaluation indicators after cleaning "post" indicates the state after cleaning is complete. The above results will serve as feedback input for subsequent re-cleaning decisions.

[0041] S6. Cleaning compliance unit judgment Based on the contamination determination chart after cleaning For each work unit Conduct a compliance assessment.

[0042] The set of non-compliant work units is defined as follows: ; in, This indicates the percentage of contamination coverage after cleaning of the work unit. The "fail" value represents the cleaning quality threshold, indicating failure and non-compliance.

[0043] S7. Extraction of Residual Contamination Areas (Input Reference) For any non-compliant work unit Extract the set of residual contaminated pixels from the input contamination determination image after cleaning.

[0044] ;in, Let be the set of residual contaminated pixels in the l-th non-compliant work unit, and p be the pixel points on the photovoltaic array surface in a unified coordinate system.

[0045] And based on connected component analysis, a set of residual contamination areas was obtained: .

[0046] This step is used to locate the specific areas of space that need to be cleaned.

[0047] S8, Generation of Drone Cleaning Enhancement Parameters For each residual contamination area Where l is the physical location index of the work unit, l=1,2,3,…,L (L is the total number of units). r=1,2,3,…,R is the number of the l-th work unit after it has been divided.

[0048] Based on the original cleaning parameters of the work unit, re-cleaning enhancement parameters are generated.

[0049] The drone's water spray flow rate is set as follows: ; The foam spray flow rate is set as follows: ; The spray wash residence time is set as follows: ; in, , All are enhancement coefficients greater than 1, and re represents re-cleaning.

[0050] S9, Unmanned Vehicle Cleaning Enhancement Parameter Generation When performing re-cleaning and scrubbing on areas with residual contamination, the roller brush speed should be set as follows: ; The travel speed is set as follows: ; Downforce setting: ; in, .

[0051] S10, Collaborative Implementation of Cleaning and Re-cleaning Drones and unmanned vehicles assemble according to the areas of residual pollution. A localized cleaning path is generated, and spraying and brushing operations are performed only within the corresponding spatial area to achieve targeted cleaning.

[0052] S11. The evaluation results after cleaning are cited again. After completing the re-cleaning operation, obtain new contamination assessment maps and cleanliness indicators for the re-cleaned area: This is used for final, complete verification.

[0053] S12, Complete Termination Determination When all work units satisfy The complete cleaning control process ends when the overall cleanliness index meets the preset requirements, or equivalently; otherwise, it re-enters S7 to continue local re-cleaning.

[0054] Traditional methods, which rely on manual sampling to determine compliance after cleaning due to the inability to automatically assess quality, are inefficient and prone to omissions. Furthermore, localized residual contamination necessitates complete rewashing of the entire row, wasting water and energy. The disconnect between the detection and cleaning modules hinders feedback control, negatively impacting both cleaning effectiveness and operational efficiency. This invention addresses these issues by automatically acquiring images and updating contamination coverage ratios after cleaning all units. It identifies substandard units and extracts areas of residual contamination through threshold determination, performing targeted re-cleaning only on these localized areas based on the updated contamination intensity. This achieves online automatic verification of cleaning effectiveness and precise, point-to-point re-cleaning, completely eliminating the inefficient manual sampling and complete row rewashing. The entire process cycles through detection, decision-making, execution, and re-detection until all units meet the standards.

[0055] This invention addresses the scenario of heterogeneous collaborative photovoltaic (PV) cleaning operations between air and ground. It constructs a complete control method encompassing detection, decision-making, execution, and re-cleaning, using correlated detection and evaluation results as input and drone water spraying and foam spraying, as well as unmanned vehicle roller brush cleaning, as execution means. By establishing a mapping model between the contamination state and the cleaning parameters of multiple actuators, it achieves a transformation of the PV module cleaning process from traditional open-loop control to data-driven complete control. Compared to existing fixed-parameter, feedback-free cleaning schemes, this invention has at least the following advantages:

[0056] First, this invention effectively avoids the coexistence of over-cleaning and under-cleaning, enabling on-demand allocation and refined utilization of cleaning resources. By using the obtained pollution coverage ratio R of a single photovoltaic panel, weighted pollution intensity S, and cleanliness index CISS as inputs for cleaning decisions, an adaptive adjustment model for cleaning parameters is constructed. This allows the drone's water spray volume, foam spray intensity, and the unmanned vehicle's roller brush rotation speed, travel speed, and brush pressure to dynamically change with the degree of pollution. For lightly dusty areas, the system automatically reduces the water and foam spray intensity and simultaneously reduces the contact intensity of the unmanned vehicle's roller brush, thereby reducing water and energy consumption and brush wear. For heavily polluted or stubbornly polluted areas, the system automatically increases the spray intensity and roller brush action force, significantly improving the success rate of a single cleaning cycle. This fundamentally solves the problems of wasted cleaning resources and insufficient localized cleaning caused by traditional uniform parameters.

[0057] Secondly, this invention establishes an online verification and complete adjustment mechanism for cleaning quality, achieving intelligent re-cleaning capabilities with immediate judgment after washing and immediate replenishment of areas not yet clean. By acquiring the panel's contamination evaluation results again after a cleaning operation and comparing them with the evaluation input before cleaning, the system can automatically determine whether the current panel meets the set cleanliness target. When the judgment result shows that there is still unacceptable contamination in some areas, the system automatically generates re-cleaning areas and re-cleaning parameters, and triggers a drone to perform local water spraying or foam enhancement treatment on that area, while simultaneously dispatching an unmanned vehicle to perform low-speed, high-pressure roller brush re-cleaning in the corresponding area. This mechanism transforms the cleaning process from a single-operation mode that ends upon execution to an evaluation-driven complete operation mode, effectively eliminating the inefficiency problems caused by relying on manual sampling and rework in existing technologies.

[0058] Furthermore, this invention enhances the cleaning adaptability in complex pollution scenarios by introducing an air-ground collaborative execution mechanism. Drones offer advantages in rapid coverage and flexible positioning, making them suitable for precise water and foam spraying in heavily polluted areas; unmanned vehicle roller brushes provide stable contact and mechanical decontamination capabilities, suitable for continuous scrubbing of large areas of dust and adhesive contaminants. This invention uses a unified cleaning decision model at its core, coordinating the aerial and ground execution units at the parameter level. This enables the two types of execution units to form a collaborative operation mechanism of softening-stripping-scrubbing within the same polluted area, thereby significantly improving the removal capacity for complex pollution types such as bird droppings, mixed water stains and dust, and oil deposits.

[0059] Furthermore, this invention directly transforms the acquired evaluation results into cleaning control inputs, achieving a continuous mapping from contamination status to execution parameters, thus enabling the cleaning process to possess excellent scalability and engineering feasibility. Since the cleaning parameter generation model uses standardized contamination indicators as input, this invention does not rely on specific models of water spray devices, foam devices, or roller brush mechanisms. It only requires calibrating the parameter ranges of different hardware to reuse the same control logic, facilitating rapid deployment on photovoltaic power plants of different scales and on different models of cleaning robot platforms, demonstrating good system compatibility and promotional value.

[0060] From an economic perspective, this invention can effectively reduce water consumption, electricity consumption, and mechanical wear by reducing unnecessary water spraying, foam spraying, and high-intensity brush operation time, thus extending the service life of cleaning equipment. At the same time, it reduces the number of manual inspections and rework, and has significant potential for saving operation and maintenance costs in large-scale photovoltaic power plants.

[0061] To address the issues that existing photovoltaic cleaning robots generally employ an open-loop control mode with fixed trajectories and fixed cleaning parameters, which cannot dynamically adjust cleaning strategies based on the degree of surface contamination and spatial distribution of contamination in photovoltaic modules, and lack the ability to verify cleaning effects and control re-cleaning, the present invention aims to propose an intelligent improvement method for the complete detection, decision-making, and execution of photovoltaic module cleaning and re-cleaning. Based on existing high-precision contamination detection and quantitative evaluation results, this method effectively transforms contamination information into cleaning control strategies, and constructs a complete cleaning control mechanism that can adaptively adjust cleaning parameters and has the ability to provide cleaning quality feedback.

[0062] Specifically, the present invention aims to solve the following key technical problems.

[0063] Firstly, this invention addresses the problem of converting spatial distribution information of contaminants into cleaning control commands in existing cleaning control systems. Recognizing that existing technologies only use contaminant detection results as inspection references and do not establish a direct link with the cleaning process, this invention uses a high-precision contaminant distribution map obtained through multi-view fusion of air and ground perspectives as a foundation to construct a mapping model between contaminated areas and cleaning operation units. This enables automatic association between panel-level and local area-level contaminant status and cleaning paths and operation units, providing reliable spatial constraints for subsequent cleaning strategy generation.

[0064] Secondly, it addresses the problem that cleaning intensity cannot be adaptively adjusted according to the severity of contamination. In existing technologies, key cleaning parameters such as brushing speed, water volume, and travel speed remain constant during operation. This invention uses the photovoltaic surface cleanliness index (CISS) and contamination type information as the core adjustment basis to establish an adaptive mapping mechanism between contamination level and cleaning parameters. This allows the cleaning robot to dynamically adjust the cleaning intensity based on the contamination level of different panels and different local areas, achieving differentiated cleaning control. This avoids resource waste in lightly contaminated areas and insufficient cleaning in heavily contaminated areas.

[0065] Third, this invention addresses the problem of existing systems lacking the ability to automatically verify cleaning quality. Addressing the issue of the inability to determine whether cleaning results meet standards after a cleaning operation, and the reliance on manual sampling, this invention acquires images of the component surface again after cleaning is completed and re-evaluates the cleaning results using the same contamination detection and evaluation model. This establishes a mechanism for comparing contamination states before and after cleaning, enabling automatic judgment and quantitative evaluation of cleaning effectiveness, thereby achieving online verification of cleaning quality.

[0066] Fourth, it solves the problem of automatic re-cleaning of localized residual contamination. Addressing the issue in existing technologies where incomplete cleaning usually necessitates re-cleaning the entire row of components, this invention uses a post-cleaning residual contamination distribution map to automatically identify substandard areas and generate localized re-cleaning tasks. It automatically plans re-cleaning paths for these contaminated areas and correspondingly increases the re-cleaning operation parameters, achieving targeted re-cleaning and parameter-enhanced re-cleaning control for locally contaminated areas.

[0067] Fifth, it addresses the long-standing disconnect between the detection system and the cleaning execution system. By constructing a unified data interface and a complete decision-making process, it organically integrates pollution detection results, cleanliness evaluation indicators, cleaning control parameter generation modules, cleaning execution modules, and effect verification modules. This forms a complete control system covering pre-cleaning strategy generation, in-cleaning parameter adjustment, and post-cleaning effect verification and re-cleaning decisions, thereby improving the intelligence level, resource utilization efficiency, and cleaning consistency of photovoltaic cleaning operations at the system level.

[0068] By achieving the above objectives, this invention can realize refined cleaning control and intelligent re-cleaning operations for complex pollution scenarios by upgrading algorithms and control strategies without changing the existing hardware structure of cleaning robots, thus providing a complete intelligent cleaning control solution that can be engineered and implemented for large-scale photovoltaic power plants.

[0069] The existing technical solutions closest to this invention mainly fall into the following two categories: The first type of solution is a photovoltaic contamination detection system based on drone inspection. This type of system typically uses deep learning models (such as YOLO and Faster R-CNN) to detect contaminants on the surface of photovoltaic modules and generate a contamination distribution map to guide subsequent cleaning. The main drawback of this solution is that detection is only used as a front-end input; the detection results are not involved in the feedback control during the cleaning process, and there is a lack of re-inspection and re-evaluation of the cleaning effect after cleaning. This deficiency arises because this type of system designs the detection module and the cleaning module as independent subsystems, failing to establish a complete connection between cross-system data calculation and control logic.

[0070] The second type of solution is a cleaning scheduling scheme based on fixed rules. For example, it may set whether to clean based on a threshold of contamination area, or perform a one-time cleaning of an entire row of components according to a preset path. The drawbacks of this type of solution are: the re-cleaning decision relies solely on human experience thresholds; it cannot adaptively adjust cleaning parameters according to different types of contamination; and it cannot adjust the path and areas to be cleaned repeatedly based on the residue after cleaning.

[0071] The root cause of these shortcomings lies in the fact that such solutions lack a decision model with the cleaning effect as the objective function, and the cleaning process lacks self-learning and self-optimization capabilities. Compared with the existing technologies, this invention does not simply improve detection accuracy, but rather focuses on the feedback of cleaning effect to construct a complete control mechanism encompassing detection, decision-making, execution, and re-detection.

[0072] Specifically, existing Chinese patent CN117021105B discloses a photovoltaic cleaning robot control method that determines the cleaning timing based on operating parameters and photovoltaic module image information. By analyzing image data and relevant operating parameters of the area to be cleaned, the method determines the dirt status of the photovoltaic modules and starts the cleaning robot to perform the cleaning operation when preset conditions are met. This solution mainly addresses the issue of the rationality of the cleaning start timing, avoiding efficiency reduction and equipment wear caused by frequent or untimely cleaning. Although this existing technology incorporates image information into the cleaning decision, its control logic still focuses on whether to start cleaning. It does not establish an adaptive adjustment model for cleaning parameters based on quantitative pollution assessment results, nor does it evaluate and provide feedback control on the cleaning effect after cleaning, and it lacks an automatic re-cleaning mechanism for uncleaned areas.

[0073] In contrast, this invention is not only used to determine the timing of cleaning, but also uses the quantitative evaluation results such as the high-precision pollution coverage ratio, weighted pollution intensity and cleanliness index output by the preceding sensing module as input to construct an adaptive generation model of multiple execution parameters such as the parameters of drone water spraying and foam spraying, and the speed and pressure of the unmanned vehicle's roller brush. After the cleaning is completed, the effect is judged again, and local re-cleaning is implemented for areas that do not meet the standards, thus forming a complete cleaning control method of detection-decision-execution-re-cleaning.

[0074] Therefore, the present invention is significantly different from and superior to the open-loop cleaning control scheme adopted by CN117021105B in terms of cleaning parameter generation dimension, cleaning effect feedback mechanism and intelligent re-cleaning control.

[0075] Based on the same inventive concept, the present invention also provides a photovoltaic module adaptive cleaning system, comprising: The data acquisition module is used to divide the photovoltaic modules to be cleaned into multiple work units and collect aerial images of the photovoltaic modules to be cleaned by drones and images taken by unmanned vehicles.

[0076] The calculation module is used to extract edge texture features from two images respectively, identify contaminated pixels based on the edge texture features, determine the contamination type and degree of each contaminated pixel, and generate a fused contamination determination map based on the contaminated pixels in the two images; map the fused contamination determination map to the spatial mapping of each work unit to obtain a set of contaminated pixels associated with each work unit; count the number of contaminated pixels based on the set of contaminated pixels in each work unit, and combine the contamination type and degree with weighted aggregation to obtain the comprehensive weighted contamination intensity and contamination coverage ratio of each work unit; for each comprehensive weighted contamination intensity, generate the corresponding UAV spraying control parameters and UAV roller brush cleaning control parameters for the work unit; based on the UAV spraying control parameters and UAV roller brush cleaning control parameters, execute UAV spraying and UAV roller brush cleaning sequentially within the same unit.

[0077] The cleaning module is used to end a cleaning cycle after all units have finished spraying and brushing.

[0078] This invention also provides a computer device, which, at the hardware level, includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the aforementioned adaptive cleaning method for photovoltaic modules.

[0079] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described adaptive cleaning method for photovoltaic modules.

[0080] Specific limitations regarding the computational system for the adaptive cleaning method of photovoltaic modules can be found in the limitations of the adaptive cleaning method for photovoltaic modules described above, and will not be repeated here. Each module in the aforementioned adaptive cleaning system for photovoltaic modules can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0081] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. Furthermore, the above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for adaptive cleaning of photovoltaic modules, characterized in that, Includes the following steps: The photovoltaic modules to be cleaned are divided into multiple work units, and aerial images of the photovoltaic modules to be cleaned by drones and images taken by unmanned vehicles are collected. Edge texture features are extracted from two images respectively, and contaminated pixels are identified based on these features. The contamination type and degree of each contaminated pixel are determined, and a fused contamination determination map is generated based on the contaminated pixels in the two images. The fused contamination determination map is spatially mapped to each work unit to obtain a set of contaminated pixels associated with each work unit. Based on the set of contaminated pixels in each work unit, the number of contaminated pixels is counted, and combined with the contamination type and degree, a weighted aggregation is performed to obtain the comprehensive weighted contamination intensity and contamination coverage ratio of each work unit. For each comprehensive weighted contamination intensity, corresponding UAV spraying control parameters and unmanned vehicle roller brush cleaning control parameters are generated for the work unit. Based on the UAV spraying control parameters and unmanned vehicle roller brush cleaning control parameters, UAV spraying and unmanned vehicle roller brush cleaning are performed sequentially within the same unit. Once all units have been sprayed and brushed, one cleaning cycle is complete.

2. The adaptive cleaning method for photovoltaic modules according to claim 1, characterized in that, After all units have completed spraying and brush cleaning, and one round of cleaning is finished, the process further includes: cyclically acquiring two images, updating the comprehensive weighted pollution intensity and pollution coverage ratio of each work unit, and identifying work units with a pollution coverage ratio exceeding a threshold as substandard units; extracting residual pollution areas from substandard units, generating re-cleaning control parameters based on the updated comprehensive weighted pollution intensity, and performing local re-cleaning according to the residual pollution areas; and ending cleaning when all units meet the standards. The re-cleaning control parameters are recalculated based on the updated comprehensive weighted pollution intensity, or can be generated by multiplying the previous UAV spraying control parameters or UAV brush cleaning control parameters by an enhancement coefficient, whereby the enhancement coefficient is determined based on the degree of pollution in the residual pollution areas.

3. The adaptive cleaning method for photovoltaic modules according to claim 1, characterized in that, The drone spraying control parameters include the drone's water flow rate, foam flow rate, and spraying residence time; the unmanned vehicle roller brush cleaning control parameters include the unmanned vehicle's roller brush rotation speed, travel speed, and roller brush downward pressure; wherein, the water flow rate, foam flow rate, spraying residence time, roller brush rotation speed, and roller brush downward pressure are positively correlated with the pollution intensity, and the travel speed is negatively correlated with the pollution intensity.

4. The adaptive cleaning method for photovoltaic modules according to claim 2, characterized in that, The extraction of residual contamination areas from non-compliant units specifically includes: acquiring the latest drone aerial images and unmanned vehicle images, generating an updated fused contamination determination map; extracting the set of residual contamination pixels within the non-compliant units from the latest fused contamination determination map, and performing connected component analysis on the set of residual contamination pixels to obtain a set of residual contamination areas.

5. The adaptive cleaning method for photovoltaic modules according to claim 1, characterized in that, For each comprehensive weighted pollution intensity, corresponding control parameters for drone spraying and unmanned vehicle brush cleaning are generated for the corresponding work unit. Specifically, this includes: calculating the drone water spray flow rate, foam spray flow rate, and spraying residence time based on the comprehensive weighted pollution intensity and a preset minimum value and a first preset adjustment coefficient; and calculating the unmanned vehicle brush rotation speed, travel speed, and brush downward pressure based on the comprehensive weighted pollution intensity and a preset extreme value and a second preset adjustment coefficient.

6. The adaptive cleaning method for photovoltaic modules according to claim 5, characterized in that, The first preset adjustment coefficient includes the water spray adjustment gain coefficient, foam adjustment gain coefficient, and time adjustment gain coefficient for drone spray washing. The second preset adjustment coefficient includes the rotation speed adjustment coefficient, speed adjustment coefficient, and pressure adjustment coefficient for unmanned vehicle roller brush cleaning.

7. The adaptive cleaning method for photovoltaic modules according to claim 1, characterized in that, The pollution types include at least one of dust, water stains, bird droppings, and oil stains, with different preset weights for different pollution types; the pollution degree is classified into levels based on the adhesion strength or coverage thickness of the pollutants, with different preset weights for different levels.

8. A photovoltaic module adaptive cleaning system, characterized in that, include: The data acquisition module is used to divide the photovoltaic modules to be cleaned into multiple work units and acquire aerial images of the photovoltaic modules to be cleaned by drones and images taken by unmanned vehicles. The calculation module is used to extract edge texture features from two images respectively, identify contaminated pixels based on the edge texture features, determine the contamination type and degree of each contaminated pixel, and generate a fused contamination determination map based on the contaminated pixels in the two images; the fused contamination determination map is spatially mapped to each work unit to obtain a set of contaminated pixels associated with each work unit; based on the set of contaminated pixels in each work unit, the number of contaminated pixels is counted, and combined with the contamination type and degree of contamination for weighted aggregation to obtain the comprehensive weighted contamination intensity and contamination coverage ratio of each work unit; for each comprehensive weighted contamination intensity, the corresponding UAV spraying control parameters and UAV roller brush cleaning control parameters for the work unit are generated respectively. Based on the aforementioned drone spraying control parameters and unmanned vehicle roller brush cleaning control parameters, drone spraying and unmanned vehicle roller brush cleaning are performed sequentially within the same unit. The cleaning module is used to end a cleaning cycle after all units have finished spraying and brushing.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the method according to any one of claims 1 to 7.

Images