Photovoltaic panel intelligent cleaning robot control method and system

Abstract

The invention discloses a photovoltaic panel intelligent cleaning robot control method and system, and aims to solve the problems of single data acquisition, poor environmental adaptability and the like of an existing photovoltaic panel cleaning technology. According to the invention, an eight-step closed-loop control method of data acquisition, environment monitoring, path planning, stain detection, cleaning execution, summary analysis, synchronous control and track and backup optimization is adopted, and equipment attitude, fitting pressure and environment parameters are acquired through distributed multiple sensors; combining the CAD model of the photovoltaic station with the real-time plate body characteristics to generate an accurate path; during cleaning, a cooperative strategy of airflow pre-blowing, airflow pre-blowing, brush group cleaning, secondary airflow blowing delayed for 0.3 s and secondary blowing is adopted, and parameters and backup are optimized by utilizing dual-motor synchronous control and closed-loop feedback; the control system dynamically adapts to the equipment and the environment; the cleaning coverage rate reaches 100%, the stain removal rate is larger than or equal to 98%, the double-motor dislocation recovery time is short, the cleaning efficiency and the equipment stability are remarkably improved, and the requirements of a complex layout photovoltaic field station are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic cleaning device control, and particularly relates to a photovoltaic panel intelligent cleaning robot control method and system. BACKGROUND

[0002] In recent years, with the rapid growth of global demand for clean energy, photovoltaic technology, as an important part of renewable energy, has been widely applied and developed. Small-capacity distributed photovoltaic stations are widely used in building curtain walls, open-air supports and other scenarios due to their flexible installation and superior wind resistance, becoming an important supplement to urban and remote power supply. However, these photovoltaic stations are exposed to the outdoor environment for a long time, and their surfaces are prone to accumulate dust, bird droppings, leaves and other debris, which significantly reduces the photoelectric conversion efficiency. According to research, surface stains can reduce the power generation efficiency of photovoltaic panels by more than 30%, therefore, regular cleaning of photovoltaic panels is a key measure to ensure their efficient operation.

[0003] Currently, there are various photovoltaic panel cleaning devices on the market, but most of these devices are designed for flat photovoltaic arrays and have many limitations when adapting to small-capacity distributed photovoltaic stations. Although CN120722811A, CN219052189U and CN120588224A provide certain technical improvements, they still cannot fully meet the cleaning needs of small-capacity distributed photovoltaic stations in actual application. Specifically, the existing technology mainly has the following problems: 1. Insufficient attitude perception and dynamic adjustment capability Small-capacity distributed photovoltaic stations are mostly installed vertically or at a large inclination angle. Traditional cleaning devices lack precise attitude perception and dynamic adjustment capability, and are prone to slipping and falling under unstable adsorption force. For example, some devices cannot sense the change in their own attitude in real time when working on photovoltaic panels with a large inclination angle, and cannot adjust the cleaning pressure according to the panel angle, resulting in incomplete cleaning or scratching the panel, affecting the service life of the photovoltaic panel.

[0004] 2. Poor environmental adaptability Existing cleaning devices do not have monitoring and response mechanisms designed for high-altitude strong winds, sudden changes in temperature and humidity, etc. In strong wind environments, the device is prone to lose control, such as some devices that do not have a wind speed monitoring module and cannot adjust their state in time when the wind speed exceeds a certain threshold, resulting in being blown off or colliding with the photovoltaic panel; in low-temperature and high-humidity environments, the cleaning assembly is prone to frost and ice, seriously affecting the safety of operation and the service life of the device, such as the cleaning brush of some devices which cannot rotate normally due to icing at low temperatures, and even damage the motor.

[0005] For example, CN219052189U has achieved remote control and adsorption fixing functions, but there are obvious shortcomings in environmental adaptability. The utility model does not design special monitoring and response mechanism for special environments such as high-altitude strong wind, sudden change of temperature and humidity, which may lead to unstable operation or damage of the equipment in extreme environments. At the same time, the design of the cleaning assembly does not fully consider the cleaning needs of different types of stains, and the cleaning effect of stubborn stains is limited, and there is no real-time verification and feedback mechanism for cleaning effect, which makes it difficult to ensure the cleaning quality.

[0006] 3. Path planning problem Path planning solidification: the path planning method of traditional cleaning equipment often cannot identify the special structures such as the frame and the splicing joint of small-capacity distributed photovoltaic power stations, and is easy to form cleaning blind area at the joint. For example, some devices use simple straight line or spiral path planning, which cannot effectively cover all areas for complex layout photovoltaic power stations. In addition, when encountering temporary obstacles, these devices are difficult to quickly bypass obstacles, resulting in insufficient cleaning coverage. For example, some devices will stop or repeatedly collide when encountering temporarily placed obstacles, and cannot autonomously plan a new path.

[0007] Optimization without combining station features: without combining photovoltaic power station CAD model and real-time panel features for path planning, the cleaning coverage of key areas such as splicing joint and frame is insufficient, and there is no flexible obstacle bypassing and intervention response capability. For example, CN120588224A control method, system and robot of a photovoltaic panel cleaning robot combines A* algorithm and reinforcement learning for adaptive path planning, but in actual photovoltaic power stations, due to the complex and variable layout of photovoltaic panels, the implementation difficulty and calculation cost of the path planning algorithm are high, which may affect the cleaning efficiency, and at the same time, in extreme environmental conditions, the stability of the sensor and the accuracy of data fusion may be affected, leading to deviation in obstacle avoidance and path planning.

[0008] 4. Cleaning strategy problem Single cleaning strategy: existing equipment cannot dynamically adjust the cleaning intensity according to the type of stain, and stubborn stains cannot be completely removed, while ordinary dust may be over-cleaned, damaging the panel coating and increasing maintenance costs. For example, some devices use fixed cleaning pressure and frequency, regardless of the type of stain, the same cleaning method is used, which makes it difficult to effectively remove stubborn stains, and over-cleaning of light dust may scratch the surface of the photovoltaic panel.

[0009] Lack of real-time verification and collaborative control: Lack of real-time verification of cleaning effect, unable to dynamically adjust cleaning method and intensity according to stain residue, and lack of "pre-blowing-cleaning-second blowing" collaborative control, high stain residue rate. For example, some equipment does not have a corresponding detection link to verify the cleaning effect after cleaning is completed, and if there is stain residue, it cannot be cleaned again in time; at the same time, there is no pre-blowing link to blow away the surface dust, and direct cleaning will affect the cleaning effect, and there is no second blowing to further remove the residual stains.

[0010] Low motor control precision: The cleaning robot driven by double motors is prone to displacement deviation, causing equipment running jitter, cleaning trajectory deviation and other problems, affecting the stability of the cleaning trajectory, and thus reducing the cleaning coverage rate. For example, some equipment driven by double motors has inconsistent speed of the two motors due to poor motor synchronization, causing the equipment to jitter during operation and the cleaning trajectory to deviate, making it impossible to clean according to the predetermined path.

[0011] For example, CN120722811A proposes a high-efficiency collaborative mechanism of constant-speed cruise and manual intervention, but its dependence on manual intervention is high, and in complex or special environments, high manual cost is still required to ensure cleaning effect and robot safety, and it lacks accurate identification of stain type and distribution, making it difficult to dynamically adjust cleaning strategy to achieve optimal resource allocation.

[0012] 5. Data acquisition and interaction problems Single data acquisition: Only local operating parameters of the cleaning equipment are concerned, and robot posture, adhesion state and historical fault data are not integrated, resulting in one-sided equipment state judgment and easy cleaning deviation or failure. For example, some equipment only collects parameters such as motor speed and current, without collecting robot posture information, adhesion state of photovoltaic panel and historical fault records, etc. When the equipment fails, it cannot accurately determine the cause of the failure, affecting maintenance efficiency.

[0013] Insufficient data interaction reliability: Data loss is prone to occur in extreme environments, affecting remote monitoring and equipment scheduling. For example, the control method, system and robot of a photovoltaic panel cleaning robot disclosed in CN120588224A integrate multiple complex components such as multi-modal visual sensor and multi-sensor fusion module. In extreme environmental conditions, the stability of the sensor and the accuracy of the data fusion may be affected, resulting in loss or error during data transmission, affecting the accurate judgment of the equipment state by remote monitoring personnel and the timeliness of equipment scheduling.

[0014] Lack of data-based state evaluation and optimization mechanism: Equipment operation data and fault records are scattered and difficult to predict failures or optimize cleaning processes through data analysis. This leads to high maintenance costs, long equipment downtime, and affects the overall operation efficiency of the photovoltaic station. For example, the operation data and fault records of some equipment are stored in different systems and are not integrated and analyzed. It is difficult to extract valuable information from a large amount of data, predict potential equipment problems in advance, and only repair the equipment after it fails, increasing maintenance costs and equipment downtime.

[0015] 6. Stain detection problem Low stain detection accuracy: Traditional detection methods rely solely on a single image or visual angle, which cannot effectively distinguish between different types of stains, leading to improper cleaning methods. For example, some equipment can only use simple image recognition to determine whether there is a stain, but cannot distinguish between different types of stains such as dust, bird droppings, or leaves, making it difficult to choose the appropriate cleaning method and intensity, affecting cleaning effectiveness.

[0016] To solve the above problems, the present application provides a photovoltaic panel cleaning robot control method and system, which integrates advanced data collection, environmental monitoring, path planning, cleaning execution, and summary analysis modules to effectively solve many problems in the prior art, providing a more high-quality and efficient solution for small-capacity distributed photovoltaic station cleaning, achieving efficient and intelligent cleaning of small-capacity distributed photovoltaic stations, and significantly improving the safety of high-altitude work and equipment service life. SUMMARY

[0017] The technical problem to be solved by the present application is to provide a photovoltaic panel intelligent cleaning robot control method and system to solve the specific technical problems existing in the cleaning process of small-capacity distributed photovoltaic stations, i.e. the existing photovoltaic panel cleaning equipment lacks precise attitude perception and dynamic adjustment capability, has poor environmental adaptability, fixed path planning, single cleaning strategy, and lacks data-based state evaluation and optimization mechanism, resulting in incomplete cleaning, low operation safety, short equipment life, and high maintenance costs.

[0018] To solve the above technical problems, the present application provides a photovoltaic panel intelligent cleaning robot control method and system, specifically including: I. A photovoltaic panel cleaning robot control method The control method of the present application realizes efficient cleaning of small-capacity distributed photovoltaic stations through eight-step closed-loop design of "data collection-environmental monitoring-path planning-stain detection-cleaning execution-summary analysis-synchronous control-trajectory and backup optimization", and the specific steps are as follows: (I) Data collection Sensor layout: Adopting a distributed layout, the six-axis gyroscope, high-precision tilt sensor, and pressure sensor are installed on the robot chassis and the contact area with the adsorption surface, ensuring the authenticity and comprehensiveness of the collected data. Data collection content: Real-time collection of three types of core data - robot operation state data (tilt angle, adhesion pressure with photovoltaic panels, motion acceleration raw data), equipment historical data (historical fault parameters, recovery parameters); Data storage: Classified storage of historical fault parameters and recovery parameters, establishment of "fault-recovery" database to provide data support for subsequent summary and analysis.

[0019] (2) Environmental monitoring Monitoring parameters: Real-time monitoring of wind speed, environmental temperature, relative humidity, and dust density on the surface of photovoltaic panels in the working environment; Collection method: Using a polling collection strategy, wind speed and temperature and humidity data are updated every 1s, and dust density data are collected every 3s; Dynamic response: Real-time analysis of the rate of change of environmental parameters, when the wind speed > 10m / s or the dust density > 1mg / m³ is detected, automatically trigger high-frequency monitoring mode, shorten the sampling interval to 0.5s, ensure the timeliness of data collection in extreme environment.

[0020] (3) Path planning Data fusion: Fusion of photovoltaic power station CAD model (including standby position coordinates) and real-time image of panel body shot by industrial camera, accurate identification of panel body frame, splicing seam and obstacles; Path generation: Establish a local coordinate system to generate a cleaning path containing straight line segments and turning segments, and encrypt the path within 5cm on both sides of the splicing seam; Synchronously plan the flight path from "standby position → target photovoltaic panel → next photovoltaic panel → standby position", and clearly define the two flight blowing points; Path adjustment mechanism: Obstacle handling: When temporary obstacles are detected, start local path re-planning to generate a barrier-avoiding path with a safety distance ≥3cm from the edge of the obstacle; Intervention response: If an intervention signal is received during path planning, immediately suspend the current planning process and enter the manual intervention instruction receiving state, and continue execution after the instruction is confirmed; Backup optimization: When receiving the excellent signal sent by the summary and analysis module, adjust the backup node settings of the path planning according to the compensation backup interval time to improve the safety of the path data.

[0021] (4) Stain detection Image acquisition: Obtain two sets of image data of the photovoltaic panel detection area under near-infrared light source off / on state through the image acquisition module, provide the basis for stain identification; Light transmission index calculation: the control unit compares the two groups of images pixel by pixel and calculates the light transmission index of each pixel (i.e. the difference between the pixel gray value of the target image and the pixel gray value at the corresponding position of the background image); Stain area segmentation: the deviation of the light transmission index of each pixel from the representative index of a standard photovoltaic panel is calculated, and abnormal light transmission points with a deviation greater than a preset threshold are screened out to obtain the target stain area through a connected domain extraction algorithm; Cleaning effectiveness verification: after each cleaning is completed, the above image acquisition and analysis steps are repeated to calculate the change in the stain characteristic index, and if the change is ≤ a preset threshold (i.e. the stain residue does not meet the standard), the cleaning mode is automatically changed, and secondary cleaning is triggered when local cleaning does not meet the standard.

[0022] (Five) Cleaning execution Actuator start: start two independent lifting cleaning brushes, high-pressure airflow injection units, and four sets of symmetrically arranged propellers; Flight pre-blowing: the propellers are rotated in the forward direction to provide lift, and when flying to the starting end of the photovoltaic panel, the angle is adjusted to blow downward to loosen the surface dust; Adsorption fixation: fly to 10-20 cm above the target photovoltaic panel, and the propellers are switched to reverse rotation to generate downward pressure, which is dynamically adjusted through pressure feedback to stabilize the adsorption pressure at 0.2-0.5 MPa; Dynamic parameter adjustment: dynamically adjust the cleaning pressure, brush set speed, and propeller speed according to the easy / hard signal; Cleaning timing control: use a "five-section" cooperative logic - airflow pre-blowing → brush set cleaning → delay 0.3s secondary airflow blowing; after cleaning is completed, the propellers are switched to forward rotation to disengage, and fly to the next panel to perform secondary dust blowing.

[0023] (Six) Summary and analysis Data reception: receive multi-dimensional data in real time, including cleaning effect data (path coverage, stain removal rate), device operation data (energy consumption, number of abnormal stoppages), and historical fault parameters and recovery parameters from the data acquisition module; First-level signal generation: normalize the historical fault parameters and recovery parameters to obtain a data recovery effect coefficient, and compare it with a preset data recovery effect threshold: When the coefficient ≥ the threshold, generate an "easy signal" (the device is in good condition and can operate normally); When the coefficient < the threshold, generate a "hard signal" (the device needs to be adjusted); Second-level signal generation: after obtaining the "hard signal", normalize the device resource parameters (such as remaining power and brush wear) and performance parameters (such as cleaning efficiency and fault recovery speed) to obtain an operation coefficient, and compare it with an operation threshold: When the coefficient is greater than or equal to the threshold value, a "good signal" is generated (the operation can continue through parameter adjustment); When the coefficient is less than the threshold value, an "intervention signal" is generated (manual intervention is required); Backup strategy optimization: obtain the current data backup interval time, and comprehensively calculate the data recovery effect coefficient, the data recovery effect threshold value and the database backup time interval to obtain a compensation backup interval time, and replace the original backup interval time to realize dynamic optimization of the data backup strategy.

[0024] (Seven) Dual-motor synchronous control Displacement difference collection: a non-contact angle sensor installed on the dual-motor transmission assembly is used to collect the displacement difference of the upper and lower independent driving mechanisms in real time; Synchronous regulation: a fuzzy controller built in the control unit generates a real-time adjustment signal according to the displacement difference, dynamically adjusts the motor speed, ensures that the time for the motor to resume stable operation after misalignment is less than or equal to 1.2 seconds, and ensures that the robot moves stably along the planned path.

[0025] (Eight) Trajectory and backup optimization Trajectory update: the intensity of the longitudinal control signal and the transverse control signal is compared to dynamically adjust the cleaning trajectory priority to ensure comprehensive coverage; Backup node optimization: the backup node density and frequency of the path planning are updated according to the compensation backup interval time generated by the summary analysis module to avoid data loss and improve operation continuity.

[0026] II. A photovoltaic panel cleaning robot control system To realize the above control method, the corresponding control system is designed, each module corresponds to the method steps one by one, and intelligent cleaning is realized in cooperation, specifically including the following core modules: (I) Data acquisition module Hardware configuration: including a distributed six-axis gyroscope, a high-precision tilt sensor and a pressure sensor, installed on the contact area between the robot chassis and the adsorption surface, to ensure the comprehensiveness of data acquisition; Function: real-time acquisition of robot tilt angle, photovoltaic panel adhesion pressure and motion acceleration raw data, and configuration of a data storage unit for classified storage of historical fault parameters and recovery parameters to establish a "fault-recovery" database.

[0027] (II) Environmental monitoring module Hardware integration: integrating an ultrasonic wind speed sensor, a temperature and humidity integrated sensor and a laser dust sensor, and uniformly installing them in the robot top wind shield to avoid environmental interference; Function: Adopting polling collection method to obtain wind speed, temperature, relative humidity and dust density data on the surface of photovoltaic panel, wind speed and temperature and humidity data are updated every 1s, dust density data are collected every 3s; when the wind speed is greater than 10m / s or the dust density is greater than 1mg / m³, automatically switch to high-frequency monitoring mode (sampling interval 0.5s), and upload the environmental data to the control unit in real time.

[0028] (Three) Path planning module Composition unit: configure industrial camera (for real-time shooting of panel image), CAD model analysis unit (for analyzing photovoltaic power station CAD model), coordinate system establishment unit, backup node adjustment unit, obstacle detection unit and intervention processing unit; Core function: Fuse CAD model and real-time panel image, identify the frame and splice seam, generate cleaning path with the upper left corner of the panel as the origin, and encrypt the path within 5cm range on both sides of the splice seam; when the obstacle detection unit detects temporary obstacles, generate a detour path with a safety distance of ≥3cm from the edge of the obstacle; when the intervention processing unit receives an intervention signal, pause the current planning and trigger the artificial intervention instruction receiver mechanism; the backup node adjustment unit adjusts the backup node setting according to the good signal.

[0029] (Four) Cleaning execution module Hardware structure: including two independent lifting cleaning brushes (symmetric layout), high-pressure airflow injection unit and four groups of symmetrically arranged propellers; Control unit: configure parameter adjustment unit, cleaning timing control unit and propeller control unit; parameter adjustment unit dynamically adjusts cleaning pressure, speed and propeller speed according to the easy / hard signal; cleaning timing control unit realizes the cooperative control of "flight pre-blowing → adsorption fixation → airflow pre-blowing → brush group cleaning → secondary blowing → flight secondary dust removal"; propeller control unit realizes forward and reverse rotation switching, angle adjustment and speed-adsorption pressure closed-loop control.

[0030] (Five) Summary and analysis module Composition unit: including data receiving unit, normalization processing unit, signal generation unit and backup strategy optimization unit; Core function: Data receiving unit receives cleaning path coverage, stain removal rate, equipment energy consumption, abnormal shutdown times and historical fault / recovery parameters; normalization processing unit normalizes fault parameters and recovery parameters to obtain data recovery effect coefficient, and normalizes resource parameters and performance parameters to obtain operation coefficient; signal generation unit compares the coefficient with the corresponding threshold value to generate easy signal, hard signal, good signal or intervention signal; backup strategy optimization unit calculates the compensation backup interval time according to the data recovery effect coefficient, data recovery effect threshold and database backup time interval, and optimizes the data backup strategy.

[0031] (Six) image acquisition module The near-infrared light source and the image sensor are configured to obtain image data of the near-infrared light source closing / opening state of the photovoltaic panel to be detected, and to provide original image support for the stain detection step.

[0032] (Seven) control unit As the core of the system, it is connected with the data acquisition module, the environment monitoring module, the path planning module, the cleaning execution module, the summary analysis module and the image acquisition module in a bidirectional communication connection; the image processing algorithm, the fuzzy control algorithm and the signal analysis algorithm are built in, which are used to process the data uploaded by each module, generate trajectory control instructions, cleaning parameter adjustment instructions and motor synchronous control instructions, and realize the collaborative control of the whole system.

[0033] (Eight) double-motor transmission assembly It includes an upper end independent driving mechanism and a lower end independent driving mechanism, each of which is provided with a non-contact angle sensor for real-time collection of motor displacement difference; the synchronous adjustment signal of the control unit is received to dynamically adjust the motor speed to compensate for the displacement difference, and the propeller is adjusted to maintain the adsorption stability, so as to ensure the smooth movement of the robot along the planned trajectory.

[0034] (Nine) communication module The communication module supports the data interaction between the control unit and each module, the remote terminal and the cloud platform, has the function of network interruption and continuous transmission, ensures the stable transmission of environment data, equipment state data and cleaning effect data, and provides support for remote monitoring and scheduling.

[0035] The present application realizes all-round collection of equipment and environment data, dynamic adaptation of environment, accurate planning of path, intelligent collaborative cleaning through the cooperative design of "flying + cleaning" and multi-module closed-loop control, guarantees the operation stability and data security, significantly improves the cleaning efficiency and quality, and adapts to the complex operation requirements of small-capacity distributed photovoltaic stations.

[0036] The control method and system of the intelligent cleaning robot for photovoltaic panels provided by the present application have the following beneficial effects: 1. The present application provides a comprehensive solution to the problems of lack of precise attitude perception and dynamic adjustment capability, poor environmental adaptability, fixed path planning, single cleaning strategy and lack of data-based state evaluation and optimization mechanism in the cleaning process of small-capacity distributed photovoltaic stations, which significantly improves the cleaning efficiency and quality.

[0037] 2. The present application realizes real-time perception of the inclination angle and adhesion pressure of the robot through the data acquisition module, dynamically adjusts the adsorption force combined with the driving control logic, effectively avoids the risk of slipping and falling, and significantly improves the safety of high-altitude operation.

[0038] 3. The environmental monitoring module of this invention accurately monitors and responds frequently to parameters such as wind speed, temperature, and humidity, enabling timely triggering of protective measures in scenarios such as strong winds and extreme temperature and humidity, effectively extending the service life of the equipment. Simultaneously, through dynamic response to environmental parameters and adaptive adjustment of cleaning parameters, it avoids overloading the equipment in extreme environments, extending the service life of core components (cleaning brushes and motors) by more than 30%, and reducing equipment maintenance costs.

[0039] 4. This invention employs a distributed array of multiple sensor types (six-axis gyroscope, high-precision tilt sensor, and pressure sensor) to collect robot posture, contact status, and motion data from all angles. It also adds a dedicated adsorption pressure sensor, integrating multi-dimensional data on equipment operation, environment, and historical faults. This increases data support dimensions by 60%, making the data more comprehensive and achieving an accuracy rate of over 98% in equipment status assessment. Furthermore, it categorizes and stores historical faults and recovery parameters, providing complete data support for control strategy optimization.

[0040] 5. This invention adopts a dynamic sampling frequency switching mechanism, which automatically triggers high-frequency monitoring under extreme environments (wind speed > 10m / s, dust density > 1mg / m³), ensuring that the operation efficiency retention rate is ≥86%; dynamically switching the monitoring sampling frequency, triggering the standby position return command under extreme environments (wind speed 12m / s, dust density 1.1mg / m³), the operation efficiency retention rate is ≥88%, solving the problem of data lag and operation interruption of traditional equipment in complex environments.

[0041] 6. This invention integrates CAD models with real-time image recognition of borders and seams, densifies the path within a 5cm range on both sides of the seam, avoids the metal bracket area of ​​the border, and improves the cleaning coverage to over 99%; it integrates CAD models with real-time images, simultaneously plans the flight path and cleaning path, densifies the path in key areas and designs the flying dust blowing node, achieving a cleaning coverage of 100%, completely eliminating cleaning blind spots in gaps and other areas, and ensuring no borders or seams are missed.

[0042] 7. When replanning the local path, this invention maintains a safe distance of ≥3cm from the edge of the obstacle. Combined with the manual intervention response mechanism, the obstacle avoidance success rate reaches 100%. The obstacle avoidance path maintains a safe distance of ≥3cm from the edge of the obstacle and supports remote intervention signal response. The obstacle avoidance success rate is 100%, effectively avoiding the damage caused by the collision between the equipment and temporary obstacles.

[0043] 8. Based on near-infrared dual-state images and transmittance index quantitative analysis, this invention achieves an accuracy rate of ≥98% in identifying loose dust and stubborn stains, providing a precise basis for targeted cleaning and avoiding the waste of traditional "one-size-fits-all" cleaning methods.

[0044] 9. This invention adopts a five-stage collaborative logic of "flying pre-blowing → airflow pre-blowing + brush group cleaning + 0.3s delayed secondary airflow blowing → flying secondary dust blowing", combined with dynamic adjustment of cleaning parameters, achieving a stain removal rate of 98.7%, a stubborn stain removal rate of ≥96%, and a residue rate of <2%; the "five-stage" cleaning sequence + dynamic parameter adjustment achieves a floating dust removal rate of ≥65% during flying pre-blowing, an overall stain removal rate of over 99%, a stubborn stain removal rate of ≥97%, and a residue rate of <1%, which is significantly better than traditional single cleaning methods.

[0045] 10. This invention, through the collaboration of fuzzy control algorithm and non-contact angle sensor, achieves an average motor misalignment recovery time of only 0.9s, with a maximum of 1.1s, and a cleaning trajectory deviation of <±2mm; the fuzzy control algorithm combined with real-time feedback from the angle sensor also achieves an average motor misalignment recovery time of 0.9s, with a maximum of 1.1s, and a cleaning trajectory deviation of <±2mm, effectively avoiding cleaning trajectory deviation caused by motor misalignment and significantly improving operational stability.

[0046] 11. The summary and analysis module of this invention generates multi-dimensional control signals (feasible / difficult / excellent / intervention) to achieve dynamic optimization of cleaning strategies, path backup and equipment parameters, improving work continuity by 50% and reducing ineffective work time; the summary and analysis module generates four types of control signals to achieve dynamic optimization of cleaning strategies and backup schemes, improving work continuity by 50% and reducing ineffective work time by 40%.

[0047] 12. This invention optimizes the backup interval time based on the data recovery effect coefficient, and combines the communication module's offline resume function (offline storage time of 15 minutes) to achieve a data transmission success rate of 99.9%. With 4G+LoRa dual communication mode and offline resume function, the data transmission success rate is 99.9%. The backup strategy is dynamically optimized to avoid data loss, and the problem of data loss in extreme environments is completely solved.

[0048] 13. This invention is designed for the panel layout and environmental characteristics of small-capacity distributed photovoltaic power stations. It can be adapted to photovoltaic panels of different specifications (such as 1.6m×1m; 1.0m×1.6m~1.2m×2.0m in the example), and can be put into use without large-scale modification, thus improving the versatility and flexibility of the equipment.

[0049] 14. This invention features fully automated control with zero abnormal downtime. Manual intervention is only required when an intervention signal is received. Compared with traditional manual cleaning solutions, labor costs are reduced by more than 70%. The fully automated operation with zero abnormal downtime and reduced manual intervention costs by more than 70% improves the efficiency and economic benefits of cleaning operations.

[0050] 15. Test data of this invention shows that the energy consumption per unit area is only 12Wh / m², and the ineffective energy consumption is reduced by path optimization and parameter adjustment, which is 25% lower than the energy consumption of traditional cleaning robots; the energy consumption per unit area is only 15Wh / m², which is 25% lower than the energy consumption of traditional cleaning robots. The collaborative design of flight and sweeping reduces ineffective energy consumption, and the long-term operation is economical. It is in line with the development trend of green energy conservation and has outstanding long-term operation economy.

[0051] 16. The present invention classifies and stores historical fault parameters and recovery parameters, establishes a "fault-recovery" database, achieves 100% fault tracing accuracy, and improves fault recovery efficiency by 40% by combining data recovery effect coefficient analysis, thereby improving equipment reliability and maintenance efficiency.

[0052] 17. The propeller forward and reverse rotation switching + adsorption pressure closed-loop control of the present invention stabilizes the pressure at 0.2-0.5MPa with a fluctuation error of ±0.02MPa, and maintains the adsorption state with a 100% success rate.

[0053] 18. The modules of this invention adopt a modular design, which supports the upgrading and replacement of sensors and actuators (such as adding a water quality sensor to adapt to different cleaning liquids). The functions can be flexibly expanded according to the needs of different photovoltaic power stations, adapting to future technology iterations, and have high foresight and scalability.

[0054] 19. Through reasonable control and design, this invention reduces the wear and tear of core components (propeller, cleaning brush, motor, etc.) during operation, and further extends the overall service life of the equipment.

[0055] 20. This invention, through innovative designs such as flight-coordinated cleaning, multi-module closed-loop control, and dynamic parameter adaptation, perfectly adapts to the complex operational needs of small-capacity distributed photovoltaic power stations, significantly improving cleaning efficiency, quality, and equipment reliability, and possesses significant engineering application value and promotion prospects. Attached Figure Description

[0056] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart illustrating the technical process of the method of the present invention. Figure 2 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0057] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments: Example 1: like Figure 1 As shown in the figure, this embodiment provides a control method for an intelligent cleaning robot for photovoltaic panels, which is applied in a distributed photovoltaic power station, as detailed below: Step 1: Data Collection In a distributed photovoltaic (PV) power plant, a smart PV panel cleaning robot is first placed at the starting position of the PV panel array to be cleaned. Using a distributed array of six-axis gyroscopes, high-precision tilt sensors, and pressure sensors located in the contact area between the robot's chassis and the adsorption surface, real-time data on the robot's tilt angle, contact pressure with the PV panels, and motion acceleration are collected. Simultaneously, historical fault and recovery parameters of the robot are categorized and stored to establish a fault-recovery database for subsequent analysis.

[0058] Step 2: Environmental Monitoring After the robot is started, the environmental monitoring module begins operation. Using a polling method, the ultrasonic anemometer and integrated temperature and humidity sensor update wind speed, temperature, and relative humidity data every second, while the laser dust sensor collects dust density data from the photovoltaic panel surface every three seconds. Simultaneously, it analyzes the rate of change of environmental parameters in real time. When the wind speed exceeds 10 m / s or the dust density exceeds 1 mg / m³, a high-frequency monitoring mode is triggered, shortening the sampling interval to 0.5 seconds to ensure data real-time performance and accuracy.

[0059] Step 3: Path Planning The path planning module integrates the CAD model of the photovoltaic power station with real-time images of the panels captured by industrial cameras to identify panel edges and seams. A coordinate system is established with the upper left corner of the panel as the origin, avoiding the metal support area of ​​the edges, and the path is densified within a 5cm radius on both sides of the seams, generating a cleaning path that includes straight and turning segments. When a good signal is received, the backup node settings of the path planning are adjusted according to the compensation backup interval to improve cleaning efficiency and reliability. During cleaning, if a temporary obstacle is detected, the obstacle detection unit generates an obstacle avoidance path that maintains a safe distance of ≥3cm from the obstacle edge; if an intervention signal is received, the intervention processing unit pauses the current planning and awaits manual intervention instructions.

[0060] Step 4: Stain Detection The image acquisition module acquires image data of the near-infrared light source in the area to be detected, with both the light source and the light source on. It calculates the pixel grayscale difference between the initial background image and the initial target image to determine the transmittance index of each pixel. Based on the deviation of the transmittance index from the representative indicator of no stains, the target stain area is extracted through connected components. After cleaning, images are re-acquired to verify the cleaning effectiveness. If the change in stain characteristic indicators is less than or equal to a preset threshold, the cleaning method is changed to ensure cleaning quality.

[0061] Step 5: Cleaning execution The cleaning execution module activates two independently lifting, symmetrically arranged cleaning brushes, a high-pressure airflow jet unit, and four symmetrically arranged propellers. During the first flight for dust removal, as the vehicle flies to the starting point of the current photovoltaic panel, the propeller angle is adjusted downwards to loosen surface dust. Upon receiving a "can proceed" signal, cleaning is performed according to standard parameters; upon receiving a "difficult to proceed" signal, parameters such as cleaning pressure and brush speed are adjusted to adapt to changes in equipment status. After identifying stains, high-pressure airflow pre-blowing is initiated first, followed by brush cleaning. After cleaning, a secondary airflow blew is performed after a 0.3-second delay. During the second flight for dust removal, after detaching from the photovoltaic panel, the propeller angle is adjusted again during flight to blow away residual dust. The propellers support forward and reverse rotation; forward rotation provides lift during flight, while reverse rotation generates 0.2~0.5MPa downward pressure during adsorption and fixation, dynamically adjusted via pressure sensor feedback. The cleaning process employs a coordinated method of "flight pre-blowing → airflow pre-blowing + brush cleaning + 0.3-second delayed secondary airflow blew → flight secondary dust removal."

[0062] Step 6: Summary and Analysis The summary and analysis module receives data on cleaning path coverage, stain removal rate, equipment energy consumption, and number of abnormal downtimes. It then normalizes this data by combining historical fault parameters and recovery parameters to obtain a data recovery effectiveness coefficient. This coefficient is compared to a data recovery effectiveness threshold to generate a feasible or infeasible signal. After obtaining the infeasible signal, resource and performance parameters are further normalized to obtain an operable coefficient, which is compared to an operable threshold to generate an excellent or intervention signal. Simultaneously, the backup strategy optimization unit calculates a compensation backup interval based on the data recovery effectiveness coefficient, the data recovery effectiveness threshold, and the database backup time interval. This compensation interval replaces the original backup interval, optimizing the data backup strategy.

[0063] Step 7: Dual-motor synchronous control The dual-motor drive assembly uses a non-contact angle sensor to collect the displacement difference between the upper and lower motors in real time. The fuzzy controller generates an adjustment signal based on the displacement difference to control the motor speed, ensuring that the time for the motor to return to stable operation after misalignment is ≤1.2 seconds, thus ensuring the stability and coordination of the robot.

[0064] Step 8: Track and Backup Optimization Based on the signal type generated by the summary and analysis module, adjust the cleaning trajectory and data backup strategy. For example, when a good signal is received, the backup interval of the cleaning path can be appropriately extended; when a difficult or intervention signal is received, the backup interval should be shortened to ensure data security and traceability.

[0065] Example 2: In another preferred embodiment, based on the above embodiment 1, such as Figure 2As shown in the figure, this embodiment provides a photovoltaic panel intelligent cleaning robot control system for application in photovoltaic power stations with complex terrain. The specific configuration is as follows: The photovoltaic panel cleaning robot control system of this embodiment includes a data acquisition module, an environmental monitoring module, a path planning module, a cleaning execution module, a summary and analysis module, an image acquisition module, a control unit, a dual-motor transmission assembly, and a communication module.

[0066] Data acquisition module: Includes a distributed layout of six-axis gyroscopes, high-precision tilt sensors and pressure sensors, installed in the contact area between the robot chassis and the adsorption surface, used to collect the robot's tilt angle, pressure against the photovoltaic panel and motion acceleration data; at the same time, it is equipped with a data storage unit to classify and store the robot's historical fault parameters and recovery parameters.

[0067] Environmental monitoring module: Integrates ultrasonic wind speed sensor, temperature and humidity sensor and laser dust sensor, installed inside the windproof cover on the top of the robot, used to monitor wind speed, temperature, relative humidity and dust density on the surface of photovoltaic panels in the working environment; adopts polling acquisition method, wind speed and temperature and humidity data are updated every 1 second, and dust density data is collected every 3 seconds. When the wind speed exceeds 10m / s or the dust density exceeds 1mg / m³, the sampling interval is shortened to 0.5 seconds.

[0068] Path planning module: Equipped with an industrial camera and CAD model parsing unit for identifying panel edges and seams; built-in coordinate system establishment unit to generate cleaning paths with the upper left corner of the panel as the origin, with path densification within 5cm on both sides of the seams; the path includes straight segments and turning segments; also equipped with a backup node adjustment unit to adjust the backup node settings of path planning based on good signals; also includes an obstacle detection unit and an intervention processing unit for detecting temporary obstacles and generating obstacle avoidance paths, as well as processing intervention signals.

[0069] The cleaning execution module includes two independently lifting cleaning brushes, a high-pressure airflow jet unit, and four symmetrically arranged propellers. It features a parameter adjustment unit that dynamically adjusts cleaning parameters based on "allowable" and "difficult" signals. Specifically, it executes cleaning with standard parameters when an "allowable" signal is received, and adjusts cleaning pressure, brush speed, and other parameters to adapt to changes in equipment status when a "difficult" signal is received. It also includes a built-in cleaning timing control unit. The module performs an initial flight dust removal; after takeoff, during flight to the starting point of the current photovoltaic panel, it adjusts the propeller angle downwards to loosen surface dust. After removing the stains, the system first initiates a high-pressure airflow pre-blowing, then starts the brush assembly cleaning. After cleaning, a second airflow blew is performed after a 0.3s delay. Then, a second flight dust blowing is performed. After detaching from the photovoltaic panel, the propeller angle is adjusted again during flight to blow away residual dust. The propeller supports forward and reverse rotation. During flight, forward rotation provides lift, and reverse rotation generates 0.2~0.5MPa downward pressure when adsorbing and fixing. The pressure sensor provides feedback and dynamic adjustment to achieve coordinated control of "flight pre-blowing → airflow pre-blowing + brush assembly cleaning + 0.3s delay secondary airflow blew → flight secondary dust blowing".

[0070] Summary and Analysis Module: Configures a data receiving unit, a normalization processing unit, and a signal generation unit; the data receiving unit receives cleaning path coverage, stain removal rate, equipment energy consumption, number of abnormal downtimes, and historical fault / recovery parameters; the normalization processing unit normalizes fault parameters and recovery parameters to obtain a data recovery effect coefficient, and normalizes resource parameters and performance parameters to obtain an operability coefficient; the signal generation unit generates feasible signals, difficult signals, excellent signals, or intervention signals by comparing with corresponding thresholds; it also includes a backup strategy optimization unit, which calculates a compensation backup interval time based on the data recovery effect coefficient, data recovery effect threshold, and database backup time interval, to replace the original data backup interval time.

[0071] Image acquisition module: configured to acquire near-infrared image data of the area to be inspected on the photovoltaic panel for stain detection and cleaning effect verification.

[0072] Control unit: Communicates with each module, generates control commands based on the signal types generated by the summary and analysis modules, and controls the robot's cleaning operations and parameter adjustments.

[0073] Dual-motor drive assembly: Ensures stable operation of the robot. It uses a non-contact angle sensor to collect the displacement difference between the upper and lower motors in real time, and the fuzzy controller generates an adjustment signal based on the displacement difference to control the motor speed.

[0074] Communication module: Enables data interaction and resume transmission after network outage, ensuring real-time communication and data synchronization between the robot and the central control system.

[0075] In photovoltaic power plants with complex terrain, the robot first acquires initial data and environmental parameters through data acquisition and environmental monitoring modules. The path planning module generates a cleaning path based on the photovoltaic power plant's CAD (Computer-Aided Design) model and real-time image data, optimizing the path by considering terrain undulations and obstacle distribution. The cleaning execution module dynamically adjusts cleaning parameters based on stain detection results and environmental parameters, employing a collaborative cleaning strategy to ensure cleaning effectiveness. The summary and analysis module monitors and analyzes cleaning results and equipment data in real time, generating control signals to guide the robot's operation. A dual-motor drive assembly ensures smooth robot operation and precise steering. The communication module enables data interaction and remote control between the robot and the central control system. The application of this system significantly improves the cleaning efficiency and operational stability of photovoltaic power plants with complex terrain.

[0076] Example 3: In another preferred embodiment, based on the above embodiment 1, this embodiment provides a photovoltaic panel cleaning robot control system, including a data acquisition module, an environmental monitoring module, a path planning module, a cleaning execution module, and a summary and analysis module; The data acquisition module is used to collect the robot's tilt angle, contact pressure with the photovoltaic panel surface, and motion acceleration data, as well as the robot's historical fault parameters and recovery parameters; The environmental monitoring module is used to monitor the wind speed, ambient temperature, relative humidity, and dust density on the surface of the photovoltaic panels in the working environment. The path planning module generates a cleaning path based on the CAD model of a small-capacity distributed photovoltaic power station and the real-time acquisition of panel edge and gap features by machine vision. The cleaning execution module is used to clean small-capacity distributed photovoltaic power stations; The summary and analysis module is used to summarize and analyze the cleaning results.

[0077] The data acquisition module operation specifically includes the following: A six-axis gyroscope, a high-precision tilt sensor, and a pressure sensor are installed in a distributed layout in the contact area between the robot chassis and the adsorption surface. The robot's tilt angle, contact pressure with the photovoltaic panel, and motion acceleration are collected in real time. At the same time, the robot's historical fault parameters and recovery parameters are classified and stored.

[0078] Simultaneously, positioning devices are installed at key locations on the robot's body. These devices provide second-level and meter-level positioning accuracy for the cleaning robot and have a built-in backup power supply during power outages, enabling continuous positioning signals for 30-60 minutes to ensure rapid retrieval after the device becomes unreachable. The second-level and meter-level positioning capability of the device helps calibrate the robot's movement trajectory, preventing missed areas due to path deviation. The continuous positioning function after a power outage reduces the risk of device loss and provides location support for maintenance.

[0079] Meanwhile, a camera device capable of capturing high-definition images is installed on the bottom of the robot. This camera device has far-infrared capabilities and can assist in monitoring the surface condition of the photovoltaic panels during flight and cleaning. The camera device captures real-time images of the photovoltaic panel surface during flight, and combined with data from the laser dust sensor, it assesses the thickness and distribution of dust coverage. It visually presents the differences in dust coverage on the photovoltaic panel surface, providing a visual basis for subsequent path densification. The far-infrared function can identify defects such as hot spots and abnormal heating on the photovoltaic panels during cleaning, and simultaneously record the location of defects (combined with data from the positioning device), realizing the dual functions of cleaning and detection.

[0080] By employing a distributed layout of six-axis gyroscopes, high-precision tilt sensors, and pressure sensors in the contact area between the robot chassis and the adsorption surface, the robot's status can be perceived comprehensively, in real-time, and with precision. The six-axis gyroscope accurately measures the robot's rotation angle and angular velocity in three-dimensional space, providing crucial data for posture adjustment. This ensures the robot maintains the correct posture at all times in complex photovoltaic panel environments, preventing the risk of slippage due to posture loss. The high-precision tilt sensor monitors the tilt angle between the robot and the photovoltaic panel surface in real time, ensuring that the robot maintains a suitable contact angle with the photovoltaic panel throughout the cleaning process, improving cleaning effectiveness and efficiency. The pressure sensor accurately measures the contact pressure between the robot and the photovoltaic panel, ensuring stable adsorption force and stable operation on photovoltaic panels with different tilt angles, greatly enhancing the robot's stability and reliability under various working conditions.

[0081] Real-time acquisition of raw data on the robot's tilt angle, contact pressure, and motion acceleration creates a real-time profile of the robot's operational status, providing first-hand data for dynamically adjusting cleaning strategies and optimizing control parameters. Simultaneously, the categorized storage of historical fault and recovery parameters is of profound significance. On one hand, in-depth analysis of historical fault parameters can uncover potential fault modes and patterns, enabling early warning and proactive prevention of faults, significantly reducing equipment failure rates and downtime, and improving equipment availability. On the other hand, storing recovery parameters provides quantitative data support for evaluating equipment maintenance effectiveness and improving maintenance strategies, contributing to precise maintenance and intelligent operation, and continuously improving the robot's overall performance and lifespan.

[0082] The environmental monitoring module operation specifically includes the following: An integrated ultrasonic wind speed sensor, temperature and humidity sensor, and laser dust sensor are installed inside the windproof cover on the top of the robot. The system uses a polling method to collect data. Wind speed and temperature / humidity data are updated every 1 second, and dust density data is collected every 3 seconds. The system analyzes the rate of change of environmental parameters in real time. When the wind speed is greater than 10 m / s or the dust density is greater than 1 mg / m³, high-frequency monitoring is triggered, and the sampling interval is shortened to 0.5 seconds.

[0083] Integrating an ultrasonic anemometer, a temperature and humidity sensor, and a laser dust sensor, all housed within a top windproof cover, this system comprehensively captures key parameters of the operating environment. Multi-dimensional monitoring of wind speed, temperature, humidity, and dust density allows the system to maintain a real-time overview of the environment. The polling-based data acquisition method, combined with dynamically adjusted sampling intervals, ensures stable data updates under normal conditions while enabling timely detection of sudden environmental changes through high-frequency monitoring (0.5-second intervals) in special circumstances such as wind speeds exceeding 10 m / s or dust densities exceeding 1 mg / m³. For example, in strong winds, it can provide early warnings of robot slippage and fall risks, facilitating the system to trigger emergency braking and other protective measures, significantly improving operational safety.

[0084] By analyzing the rate of change of environmental parameters in real time and adjusting the monitoring frequency, the system can dynamically optimize cleaning strategies based on the environment. When dust density is high, key cleaning areas can be planned in advance or cleaning intensity can be increased. Temperature and humidity data provide a basis for determining whether the photovoltaic panel surface is prone to condensation or icing, avoiding operation in unsuitable environments that could affect cleaning results or damage equipment. This dynamic response to the environment makes robotic cleaning operations more targeted, effectively improving cleaning efficiency and quality while reducing unnecessary energy consumption and equipment wear and tear.

[0085] The path planning module operation specifically includes the following: It integrates CAD models of small-capacity distributed photovoltaic power stations with real-time images of the panels captured by industrial cameras to identify borders and splicing seams; Establish a coordinate system with the upper left corner of the plate as the origin, avoid the metal bracket area of ​​the frame, and densify the path within 5cm on both sides of the splicing seam. The generated path includes straight segments and turning segments. When a good signal is received, the backup node settings of the path planning are adjusted according to the compensation backup interval.

[0086] When a temporary obstacle is detected, local path replanning is performed, and the obstacle bypass path maintains a safe distance of ≥3cm from the edge of the obstacle; If an intervention signal is received during the route planning process, the current planning will be paused and a manual intervention instruction will be awaited.

[0087] By integrating CAD models of small-capacity distributed photovoltaic (PV) power plants with real-time images from industrial cameras, the robot can accurately identify edges and seams. Based on this, a coordinate system is established to plan the path, avoiding areas that do not require cleaning, such as metal supports on the edges. Simultaneously, the path is densified within a 5cm radius on both sides of the seams to ensure thorough cleaning of critical areas prone to dust accumulation, significantly improving cleaning coverage. The rational planning of straight and turning segments makes the robot's movement smoother, reducing ineffective paths and ensuring the uniformity and accuracy of cleaning. This effectively avoids problems such as missed areas and repeated cleaning, guaranteeing the cleanliness of the PV panels.

[0088] By combining the flight trajectory of the takeoff and landing flight control submodule, a complete path is planned from the standby position to the target photovoltaic panel, then to the next photovoltaic panel, and back to the standby position, clearly defining the two dust-blowing nodes during the flight process. The synchronized planning of the flight path and the dust-blowing nodes allows for the early loosening of dust during robot movement (first dust blowing) and the secondary removal of residue after cleaning (second dust blowing), improving overall cleaning efficiency.

[0089] The robot can quickly replan local paths to overcome temporary obstacles, maintaining a safe distance of ≥3cm between the obstacle and its edge to prevent collisions and ensure equipment safety without affecting overall cleaning progress. Adjusting backup node settings based on favorable signals optimizes data backup strategies and ensures path recovery. Pausing planning upon receiving intervention signals and awaiting manual instructions allows for human judgment in complex situations, ensuring the rationality of path planning. This enables the robot to adapt to changing working environments, significantly improving operational efficiency and safety.

[0090] In a preferred embodiment, the cleaning execution module specifically includes the following: It includes two sets of independently lifting sweeping brushes and a high-pressure air jet unit, with the sweeping brushes and the high-pressure air jet unit arranged symmetrically. The parameters are dynamically adjusted according to the type of stain. When a "can do" signal is received, cleaning is performed according to the normal parameters; when a "difficult to do" signal is received, the parameters are adjusted to adapt to the change in equipment status. When a stain is detected, the airflow is first activated for pre-blowing, then the brush assembly is activated for cleaning. After cleaning is completed, a second airflow is activated after a 0.3-second delay to ensure that no stain remains.

[0091] Two independently raised and lowered cleaning brushes and a high-pressure airflow jet unit are symmetrically arranged to fully cover the photovoltaic panel surface, avoiding blind spots. Parameters are dynamically adjusted according to the type of stain, employing appropriate cleaning strategies for different stains. For example, increased cleaning force is used for stubborn stains, while standard parameters are used for ordinary dust, achieving precise cleaning. First, airflow loosens the stains, then the brushes perform deep cleaning, and finally a second blowing removes any residue, ensuring no stains remain. This significantly improves the thoroughness of cleaning, maintains the cleanliness of the photovoltaic panel surface, and thus preserves its high photoelectric conversion efficiency.

[0092] When a "workable" signal is received, the system executes according to standard parameters; when a "difficult" signal is received, the parameters are adjusted to adapt to changes in equipment status. This flexible parameter adjustment mechanism allows the cleaning execution module to optimize operation based on the real-time status of the equipment, preventing a decrease in cleaning effectiveness or equipment damage due to fluctuations in equipment status. The independently lifting cleaning brush design allows for flexible height adjustment based on the flatness of the photovoltaic panel surface, reducing hard friction with the panel. The symmetrical layout of the high-pressure airflow jet unit ensures more balanced force, reducing wear and tear during equipment operation.

[0093] The cleaning execution module also includes a take-off and landing flight control submodule. Four sets of propellers are symmetrically installed around the robot's body to drive its take-off and landing. When the robot takes off from its standby position, the propellers rotate clockwise to provide upward lift, enabling the robot to take off and land vertically. When it flies to a position 10-20cm above the target photovoltaic panel, the propellers switch to counter-clockwise rotation, generating downward pressure (i.e., suction force). With real-time feedback from pressure sensors, the counter-clockwise rotation speed is adjusted to stabilize the adhesion pressure between the robot and the photovoltaic panel surface at 0.2-0.5MPa, achieving adhesion and fixation (replacing the traditional mechanical adsorption module). Two dust removal operations are performed according to the preset nodes of the path planning module. First dust removal: The robot takes off from the standby position or the standby point of the previous photovoltaic panel. During the flight to the starting end of the current photovoltaic panel, the propeller keeps rotating forward (lift mode) and adjusts the blade angle to make the airflow blow downward to sweep the surface of the photovoltaic panel and loosen the surface dust. Second cleaning: After the current photovoltaic panel is cleaned, the robot's propellers switch to forward rotation to disengage from the adsorption and fly to the starting point of the next photovoltaic panel or return to the standby position. During this process, the propeller angle is adjusted again to blow downwards to remove the dust particles remaining after cleaning.

[0094] After the robot is attached and fixed (propeller reverses for attachment), the cleaning process begins: first, a pre-blowing is performed by the high-pressure air jet unit (to clean residual dust after the first flight dust removal), then the cleaning brush is activated to clean along the planned path. After cleaning, a second airflow is performed 0.3 seconds later to ensure no stains remain. After the second blow-off, the propeller switches to forward rotation to detach from the attachment and prepares for the second flight dust removal. During the cleaning process, the bottom camera continuously captures images of the cleaned area and provides real-time feedback on the stain removal status. If the image shows that some stains have not been removed (removal rate <90%), a second cleaning of that area is triggered.

[0095] The summary and analysis module operation specifically includes the following: It receives real-time data on cleaning path coverage, stain removal rate, equipment energy consumption, and number of abnormal shutdowns, while also receiving historical fault parameters and recovery parameters of the robot from the data acquisition module. The fault parameters and recovery parameters are normalized to obtain the data recovery effect coefficient. The data recovery effect coefficient is compared with the data recovery effect threshold to generate a feasible signal or a difficult signal. After obtaining the difficult-to-operate signal, the resource parameters and performance parameters are normalized to obtain the operability coefficient, which is then compared with the operability threshold to generate the good signal or intervention signal.

[0096] After obtaining a good signal, the data backup interval time is obtained. The data recovery effect coefficient, data recovery effect threshold and database backup time interval are comprehensively calculated to obtain the compensation backup interval time, which is used to replace the data backup interval time and optimize the data backup strategy.

[0097] The system receives core data such as cleaning path coverage and stain removal rate in real time, and performs quantitative analysis by combining historical fault and recovery parameters. Through normalization processing, it generates data recovery effect coefficients and operability coefficients, providing objective basis for generating feasible / difficult-to-operate signals and excellent / intervention-required signals. This data-driven dynamic judgment can accurately identify the equipment's operating status. When a difficult-to-operate signal is triggered, resource and performance parameters can be used to determine whether manual intervention is necessary, avoiding inefficient or equipment-damaged operations caused by blind operation. Simultaneously, a data backup strategy optimized based on excellent signals (compensating for backup interval time) reduces redundant storage while ensuring data security, further improving system operating efficiency.

[0098] By integrating real-time data such as equipment energy consumption and abnormal downtime frequency with historical fault parameters, the system comprehensively assesses equipment health status and identifies potential fault risks in advance. If the data recovery effectiveness coefficient continues to decline, the system can predict a decrease in equipment recovery capability and proactively prevent the fault from escalating by adjusting backup strategies or triggering maintenance prompts. Furthermore, linking data recovery effectiveness with backup strategies allows for dynamic optimization of data storage logic when equipment status fluctuates, ensuring that critical operational data is not lost.

[0099] In addition, standby positions and monitoring systems are also set up.

[0100] The standby position is located in the wind-sheltered area at the edge of the photovoltaic power station and has the following functions: The standby position has a built-in wireless charging coil or charging interface. The robot will automatically dock and charge after returning. The charging efficiency is ≥80%. When fully charged, the robot can continuously clean 50-80 photovoltaic panels. Integrating 4G / 5G and LoRa communication modules, it serves as a relay node between the robot and the remote monitoring center, amplifying the communication signal (coverage radius ≥500m) and preventing signal interruption for the robot in the edge areas of the site; The standby position is equipped with an automatically closing rainproof canopy and insulation layer. When the environmental monitoring module detects heavy rain (rainfall > 10 mm / h), heavy snow (snow thickness > 5 cm), or high temperature (ambient temperature > 40℃), it sends a command to guide the robot back to the standby position, the rainproof canopy closes automatically, and the insulation layer is activated (maintaining an internal temperature of 5-15℃ in low temperatures) to protect the equipment from damage caused by extreme environments.

[0101] The monitoring system is deployed in a remote operations and maintenance center and is linked with the communication gateway of the standby unit. Its specific functions include: The real-time display of each robot's position (positioning device data), cleaning progress, remaining battery power, and fault alarm status (such as propeller failure, abnormal suction pressure) is achieved through a visual interface. Alarm information is presented with sound and light prompts and is automatically recorded. It supports selective viewing of real-time or historical images captured by the bottom cameras of each robot (searchable by time / board number), and can switch between "high-definition visible light mode" and "far-infrared thermal imaging mode" to intuitively view the cleaning effect and board defects; The intelligent system automatically assigns robots to perform cleaning tasks based on the number of photovoltaic panels at the site, dust density (data from the environmental monitoring module), and robot status (remaining power and fault status), thus avoiding resource idleness or overload. When an intervention signal is received or a special situation needs to be handled, maintenance personnel can remotely and manually control the robot's movement (such as adjusting the flight trajectory or initiating an emergency stop) through the monitoring system, which supports two control methods: joystick or coordinate input. It supports remotely sending firmware upgrade packages to the robot without interrupting the current cleaning task during the upgrade process (upgrade in stages in the background). After the upgrade is completed, it will automatically restart and resume the job, ensuring continuous optimization of system functions.

[0102] Example 4: In another preferred embodiment, based on Embodiment 1 above, this embodiment provides a control system and method for an intelligent photovoltaic panel cleaning robot, applied to a small-capacity distributed photovoltaic power station with a capacity of 500kW. This power station includes 120 photovoltaic panels with dimensions of 1.6m × 1m, a panel splicing seam width of 5mm, and a frame metal support width of 8cm. Temporary obstacles (such as maintenance tools and cables) are randomly distributed on the site, and two standby positions are set up (30m apart, each covering 60 photovoltaic panels). The operating environment has a wind speed range of 0~15m / s, a relative humidity of 30%~85%, and a normal dust density of 0.3~0.8mg / m³ on the photovoltaic panel surface, which can reach 1.2mg / m³ in extreme weather conditions. The standby positions are equipped with wireless charging devices and rainproof and heat-insulating covers, supporting the robot's vertical take-off and landing and status maintenance.

[0103] I. System Hardware Configuration (a) Data Acquisition Module Sensor selection: MPU6050 six-axis gyroscope, SCA100T high-precision tilt sensor (measurement range ±15°, accuracy 0.01°), and FSH800-Plus pressure sensor (measurement range 0~100N, accuracy 0.1N, suitable for adsorption pressure detection). Installation method: Three sensors are distributed and installed in the front, middle and rear contact areas of the robot chassis adsorption surface, with a spacing of 20cm, and communicate with the control unit via I2C bus; an additional pressure sensor is installed in the center of the chassis to monitor the adsorption pressure. Storage unit: SD (Secure Digital) card (capacity 32GB) is used to classify and store historical fault parameters (such as motor misalignment, abnormal propeller speed, insufficient adsorption pressure) and recovery parameters (such as voltage adjustment, recovery time, propeller speed correction value). The storage format is CSV (Comma-Separated Values) file, and it is managed by date partition.

[0104] (II) Environmental Monitoring Module Sensor integration: The ultrasonic wind speed sensor is FS100 (measurement range 0~20m / s, accuracy 0.1m / s), the temperature and humidity sensor is SHT30 (temperature range -40℃~85℃, humidity range 0~100%RH), and the laser dust sensor is PMS5003 (measurement range 0~10mg / m³, accuracy 0.01mg / m³). Installation location: The sensor is uniformly fixed inside the windproof cover on the top of the robot. The opening diameter of the windproof cover is 2cm to avoid direct impact of airflow on the sensor. The wind speed sensor is facing the direction of the robot's flight to ensure the accuracy of data collection. The acquisition circuit uses an STM32F103 microcontroller as the acquisition controller. The polling period is set according to claim 3. The high-frequency mode triggering condition is implemented through a hardware comparator. At the same time, an extreme environment (rainstorm, high temperature) detection circuit is added to link the standby position return instruction.

[0105] (III) Path Planning Module Industrial camera: Baslerac A1300-200uc (resolution 1280×960, frame rate 20fps) was selected and mounted on the front of the robot. The lens focal length was 8mm and the shooting angle was 45° with the normal of the board (target shooting angle). An additional auxiliary camera was installed on the top of the robot for flight path navigation. CAD Model Parsing Unit: The PCL (Point Cloud Library) is used to parse the CAD model of the site (in .dwg format) and extract the coordinate information of the plate frame, splicing seam and standby position; Coordinate system parameters: A local rectangular coordinate system is established with the top left corner vertex of each photovoltaic panel as the origin (X-axis along the horizontal axis of the panel, Y-axis along the vertical axis of the panel, unit mm); at the same time, a global coordinate system of the site is established to clarify the relative position of the standby position and each photovoltaic panel, and to generate a complete path of "standby position → target photovoltaic panel → next photovoltaic panel → standby position".

[0106] (iv) Takeoff and landing flight control and cleaning execution module Take-off and landing flight components: 4 sets of symmetrically arranged propellers (model 1045 carbon fiber propellers, diameter 10cm), installed around the robot body (spaced 40cm apart), the drive motor is a 2208 brushless motor (rated voltage 24V, speed adjustable from 0 to 3000rpm); equipped with an ESC (model ESC30A) to achieve precise speed control and support forward and reverse switching; Cleaning brushes: Two sets of independently lifting nylon brushes (15cm in diameter, 3cm in bristle length) are symmetrically installed on both sides of the robot (30cm apart), with a lifting stroke of 0~5cm. The drive motor is a 24V DC brushless motor (adjustable speed from 0 to 300rpm). High-pressure airflow jet unit: a miniature air pump (output pressure 0.3MPa, airflow speed 25m / s) with a nozzle diameter of 2mm is used, and the nozzles are symmetrically arranged on the inside of the sweeping brush (5cm away from the brush group). Timing and pressure control: The timer integrated into the STM32H743ZIT6 control unit realizes the timing trigger of "flight pre-blowing → airflow pre-blowing → brush group cleaning → secondary airflow purging → flight secondary dust blowing"; the adsorption pressure is adjusted by the feedback of the pressure sensor, and the propeller reverse rotation speed is linearly related to the pressure (1500 rpm corresponds to 0.2 MPa pressure, 2500 rpm corresponds to 0.5 MPa pressure).

[0107] (v) Summary and Analysis Module and Control Unit Control unit: STM32H743ZIT6 microcontroller is selected, with built-in fuzzy controller (membership function adopts triangular function, quantization factor Ke=0.5, Kec=0.3), and a new propeller speed-pressure regulation algorithm is added; Normalization: Linear normalization formula is used. (1) The data recovery effectiveness coefficient ranges from 0 to 1, with a threshold set at 0.7; the operability coefficient threshold is set at 0.6. Backup strategy optimization: The initial data backup interval is 30 seconds, and the compensation backup interval is calculated as follows: (2) in, For the initial interval, The data recovery effectiveness coefficient. This is the threshold for data recovery effectiveness.

[0108] (vi) Dual-motor drive assembly and communication module Dual motors: 24V DC geared motors (reduction ratio 1:50) are selected, and the non-contact angle sensor is AS5600 (12-bit resolution), which is installed on the motor output shaft; Communication module: The EC200S 4G module is selected, which supports TCP / IP protocol. It automatically switches to local storage when the network is disconnected, and the offline data storage time can reach 15 minutes. It automatically resumes the data transmission after the network is restored. A short-range LoRa communication module (communication distance ≤50m) with the standby position is added to ensure the stability of take-off and landing command transmission.

[0109] III. Control Method Execution Flow (a) Data collection and environmental monitoring The sensors collect data in real time: tilt angle sampling frequency 10Hz, adhesion pressure (including adsorption pressure) sampling frequency 10Hz, and motion acceleration sampling frequency 20Hz; the adsorption pressure data is uploaded to the control unit in real time for dynamic adjustment of the propeller speed. Environmental parameter acquisition: Wind speed, temperature and humidity are updated every 1 second, and dust density is collected every 3 seconds; when a wind speed of 10.2 m / s or a dust density of 1.05 mg / m³ is detected, high-frequency monitoring is triggered (sampling interval 0.5 s), and an environmental warning signal is sent to the control unit at the same time; when heavy rain (humidity > 95%) or high temperature (> 60℃) is detected, a standby position return command is triggered.

[0110] (ii) Path planning Feature recognition: An industrial camera captures images of the board body, and the Canny edge detection algorithm is used to identify the borders and seams (recognition accuracy ±1mm); a top auxiliary camera acquires images of the flight path, and the flight path is generated by combining the standby position coordinates analyzed from the CAD model. Path generation: A longitudinal straight cleaning path (Y-axis direction, step size 5cm) is generated with the upper left corner (0,0) of the panel as the origin. The path is densified with a step size of 2cm within a 5cm range on both sides of the splicing seam (X-axis direction ±5cm). The turning section adopts an arc transition (radius 10cm). The flight path is planned as "standby position 1 → photovoltaic panel 1-60 → standby position 1" and "standby position 2 → photovoltaic panel 61~120 → standby position 2". Two dust blowing nodes are defined: the first dust blowing is performed when the plane flies to the starting end of the photovoltaic panel (50cm away from the panel). After leaving the panel, the plane flies to the front of the next panel (50cm away) to perform the second dust blowing. Obstacle avoidance and intervention: When an obstacle is detected (image recognition area ≥ 10cm²), an obstacle avoidance path is generated (3.5cm away from the edge of the obstacle); when a manual intervention signal is received (sent by the 4G module), the planning is paused and the "waiting for intervention" status is displayed; the motion trajectory is calibrated in real time through the positioning calibration points in the flight path (one is set every 10m).

[0111] (III) Stain detection and cleaning execution Image acquisition: The background image is acquired when the near-infrared light source (wavelength 850nm, power 5W) is off, and the target image is acquired when it is on (exposure time 10ms). Transmittance index calculation: Pixel grayscale value range 0~255, transmittance index = target image pixel grayscale value - background image corresponding pixel grayscale value, no stains represent an index set to 230; Stain assessment: If the light transmittance deviation is >20 (light transmittance <210), it is a stubborn stain. Activate the wet brush unit and spray cleaning fluid (5ml / min); if the deviation is ≤20, it is loose dust. Activate the dry brush unit. Cleaning execution: Step 5.1: Start the two sets of cleaning brushes, the high-pressure air jet unit, and the four sets of propellers; Step 5.2: First flight dust removal: The propeller rotates forward (2200 rpm) and takes off from the standby position. When it flies to the starting end of the current photovoltaic panel (50 cm away), the propeller angle is adjusted downward (30° with the horizontal direction) and blows for 1.5 seconds to remove more than 60% of the surface dust. Step 5.3: Adsorption and fixation: Fly to a position 15cm above the plate, switch the propeller to reverse (initial speed 1800rpm), and adjust the speed through feedback from the central pressure sensor to stabilize the adsorption pressure at 0.3MPa (error ±0.02MPa). Step 5.4: Pre-blow with high-pressure airflow for 1 second, start the brush group (normal speed 200 rpm) for cleaning; when the "difficult to move signal" is received, the brush group speed is increased to 250 rpm, and the adsorption pressure is increased by 0.05 MPa; Step 5.5: After cleaning is completed, delay for 0.3 seconds, then start secondary airflow purging for 0.5 seconds; Step 5.6: Detachment and Secondary Dust Blowing: Switch the propeller to forward rotation (2000 rpm) to detach from the plate. When flying to the next plate (50 cm away), adjust the propeller angle to blow downwards for 1 second to remove residual dust.

[0112] (iv) Summary, Analysis and Synchronization Control Signal generation: The summary and analysis module receives data (path coverage 100%, stain removal rate 99.1%, dust pre-removal rate 65%, energy consumption 15Wh / m², number of abnormal shutdowns 0), combines it with historical fault parameters (no fault records), normalizes it to obtain the data recovery effect coefficient 0.85≥0.7, and generates a "workable signal". Backup optimization: Compensate backup interval time = 30s × (0.85 / 0.7) ≈ 36s, replacing the initial interval time; Dual-motor synchronization: Angle sensors collect displacement differences in real time. When a displacement difference of 0.5° is detected, the fuzzy controller generates an adjustment signal to increase the speed of the driven motor by 5%, and the motor returns to stable operation within 1.0s (meeting the requirement of ≤1.2s). Propeller control: When the adsorption pressure fluctuates beyond ±0.03MPa, the control unit adjusts the reverse rotation speed of the propeller in real time (adjustment step size 50rpm), and restores stability within 100ms.

[0113] This embodiment involves a 7-day test at the aforementioned photovoltaic power station, and the results are as follows: Key cleaning metrics: 100% path coverage, 99.1% stain removal rate (97.2% for stubborn stains, 100% for loose dust), and 65% loose dust removal rate via pre-blowing. Operational stability: The average recovery time for dual-motor misalignment is 0.9s (maximum 1.1s), the adsorption pressure stability error is ±0.02MPa, and the propeller forward / reverse switching response time is ≤80ms; Environmental adaptability: The operating efficiency remains at 88% when the wind speed is 12m / s and the dust density is 1.1mg / m³; the standby position return response time is ≤3s under extreme environments. Other indicators: 0 abnormal shutdowns, 99.9% data transmission success rate, 72 photovoltaic panels cleaned continuously on full charge (including flight energy consumption), and an average flight time of 2.5 seconds per panel for dust removal.

[0114] In summary, this embodiment fully realizes the technical solution of the latest claims by adding hardware configurations such as propeller take-off and landing, aerial soot blowing, and adsorption pressure control, and optimizes the process. It verifies the efficiency and stability of the "flight + cleaning" collaborative mode and adapts to the operational needs of small-capacity distributed photovoltaic power stations.

[0115] In a preferred embodiment, step 3 further includes a path adjustment step: when a temporary obstacle is detected, local path replanning is performed, maintaining a safe distance of ≥3cm between the obstacle bypass path and the obstacle edge; if an intervention signal is received during path planning, the current planning is paused and awaits manual intervention instructions; these settings ensure that the robot can flexibly respond to emergencies in complex environments, avoid collision damage, and ensure personnel safety. After a manual intervention instruction is issued, the system responds quickly, reassesses the environmental information, and continues to complete the remaining path planning, improving overall operational efficiency and stability.

[0116] In a preferred embodiment, step 6 further includes a backup strategy optimization step: obtaining the current data backup interval, comprehensively calculating the data recovery effectiveness coefficient, data recovery effectiveness threshold, and database backup time interval to obtain a compensating backup interval; replacing the original data backup interval with the compensating backup interval to optimize the data backup strategy; these settings ensure that the backup interval is automatically shortened when the data recovery effect is unsatisfactory, thereby improving data security. Simultaneously, regularly evaluating changes in the data recovery effectiveness coefficient and dynamically adjusting the compensating backup interval ensures that the backup strategy always aligns with actual needs.

[0117] In a preferred embodiment, the data acquisition module includes a distributed layout of six-axis gyroscopes, high-precision tilt sensors, and pressure sensors, installed in the contact area between the robot chassis and the adsorption surface. A data storage unit is configured to classify and store historical fault parameters and recovery parameters of the robot. This setup allows for real-time monitoring of the robot's operating posture and adsorption status, accurately capturing abnormal data. The data storage unit employs encryption and compression technology, ensuring data security while saving storage space, facilitating subsequent in-depth analysis of fault patterns, and providing a reliable basis for optimizing robot performance.

[0118] In the preferred embodiment, the environmental monitoring module integrates an ultrasonic anemometer, a temperature and humidity sensor, and a laser dust sensor, all housed within a windproof shield on the top of the robot. A polling-based data acquisition method is used, updating wind speed and temperature / humidity data every 1 second, and dust density data every 3 seconds. When the wind speed exceeds 10 m / s or the dust density exceeds 1 mg / m³, the sampling interval is shortened to 0.5 seconds. These settings ensure the robot can quickly respond to environmental changes in complex environments, improving the real-time performance and accuracy of data acquisition. Simultaneously, the windproof shield design effectively protects the sensors from severe weather interference, extending equipment lifespan, reducing maintenance costs, and enhancing the overall stability and reliability of the monitoring system.

[0119] In the preferred embodiment, the path planning module is equipped with an industrial camera and a CAD model analysis unit to identify the board frame and seams; it has a built-in coordinate system establishment unit to generate a cleaning path with the upper left corner of the board as the origin, and the path is densified within a 5cm range on both sides of the seam. The path includes straight segments and turning segments; a backup node adjustment unit is configured to adjust the backup node settings of the path planning according to the good signal. The above settings ensure that the cleaning path accurately covers the entire area of ​​the board, especially enhancing the cleaning effect at the seams and avoiding missed or repeated cleaning; at the same time, by dynamically adjusting the backup nodes, the fault tolerance and adaptability of the path planning are improved, ensuring the stable operation of the equipment under complex working conditions.

[0120] In a preferred embodiment, the path planning module further includes an obstacle detection unit and an intervention processing unit. When the obstacle detection unit detects a temporary obstacle, it generates an obstacle avoidance path that maintains a safe distance of ≥3cm from the edge of the obstacle. When the intervention processing unit receives an intervention signal, it pauses the current planning and triggers a manual intervention command receiving mechanism. These settings ensure that the robot can flexibly respond to emergencies in complex environments, automatically avoid obstacles to ensure operational safety, and respond promptly to human intervention, thereby improving the overall reliability and controllability of the operation and meeting diverse scenario requirements.

[0121] In the preferred embodiment, the cleaning execution module includes two independently raised and lowered cleaning brushes and a high-pressure airflow jet unit, arranged symmetrically. It is equipped with a parameter adjustment unit that can dynamically adjust cleaning parameters based on feasible and difficult-to-clean signals. A built-in cleaning timing control unit enables coordinated control of "airflow pre-blowing → brush cleaning → 0.3s delayed secondary airflow blowing." These features effectively improve cleaning efficiency and quality, and can flexibly adapt to different levels of stains and floor materials. The two sets of independently raised and lowered cleaning brushes can reach deep into crevices, while the high-pressure airflow jet provides powerful stain removal. The coordinated control process avoids residue and ensures effective cleaning.

[0122] In the preferred embodiment, the summary and analysis module is configured with a data receiving unit, a normalization processing unit, and a signal generation unit. The data receiving unit receives cleaning path coverage, stain removal rate, equipment energy consumption, number of abnormal downtimes, and historical fault / recovery parameters. The normalization processing unit normalizes the fault and recovery parameters to obtain a data recovery effect coefficient, and normalizes the resource and performance parameters to obtain an operability coefficient. The signal generation unit generates a feasible signal, a difficult signal, a good signal, or an intervention signal by comparing with corresponding thresholds. These settings allow for accurate assessment of the cleaning equipment's operating status. A feasible signal indicates smooth equipment operation, a difficult signal indicates potential problems, a good signal indicates excellent overall performance, and an intervention signal requires immediate inspection and maintenance, providing a scientific basis for equipment management.

[0123] In a preferred embodiment, the summary and analysis module further includes a backup strategy optimization unit. This unit calculates a compensation backup interval based on the data recovery effectiveness coefficient, the data recovery effectiveness threshold, and the database backup time interval, replacing the original data backup interval. This setting dynamically adjusts the backup frequency; when the data recovery effectiveness fails to reach the threshold, the compensation backup interval is shortened to ensure data security; conversely, the interval is extended to reduce storage resource consumption. Through real-time monitoring and intelligent adjustment, the optimal configuration of the backup strategy is achieved.

[0124] In summary, this invention provides a control method and system for an intelligent photovoltaic panel cleaning robot, effectively solving key technical challenges in the cleaning process of small-capacity distributed photovoltaic power plants. Addressing the problems of existing photovoltaic panel cleaning equipment when adapted to small-capacity distributed scenarios, such as loss of posture control, poor environmental adaptability, inefficient path planning, limited cleaning strategies, and lack of data management, this invention achieves a systemic breakthrough through multi-dimensional technological innovation.

[0125] Firstly, in terms of attitude perception and dynamic adjustment, this invention integrates a six-axis gyroscope, a high-precision tilt sensor, and a pressure sensor in a distributed layout on the robot chassis for the first time. It collects tilt angle, contact pressure, and motion acceleration data in real time and accurately, and dynamically adjusts the adsorption force in combination with drive control logic. This completely solves the technical bottleneck of traditional equipment being prone to slipping and falling on vertical or steeply tilted photovoltaic panels, and significantly improves the safety and stability of high-altitude operations.

[0126] Secondly, addressing the challenge of environmental adaptability, this invention innovatively integrates an ultrasonic wind speed sensor, a temperature and humidity sensor, and a laser dust sensor into the robot's top windproof cover. Through polling and dynamic frequency adjustment mechanisms (such as shortening the sampling interval to 0.5s when the wind speed is >10m / s or the dust density is >1mg / m³), it achieves real-time monitoring and high-frequency response to extreme environments such as strong winds at high altitudes and sudden changes in temperature and humidity, filling the technological gap of existing equipment under complex climatic conditions.

[0127] In the field of path planning, this invention proposes for the first time a hybrid planning method that integrates CAD models and machine vision. By identifying the frame and splicing seams of photovoltaic power stations and establishing a coordinate system, the path is densified within a 5cm range on both sides of the splicing seams. At the same time, it supports dynamic obstacle avoidance of temporary obstacles (maintaining a safe distance of ≥3cm), which improves the cleaning coverage rate to over 99% and completely eliminates the cleaning blind spot problem caused by the fixed path of traditional equipment.

[0128] In terms of cleaning strategy, this invention creatively designs two sets of independently lifting cleaning brushes and a high-pressure airflow jet unit in a symmetrical layout. It combines a five-stage process of "flying pre-blowing → airflow pre-blowing + brush group cleaning + 0.3s delayed secondary airflow blowing → flying secondary dust blowing" and dynamically adjusts parameters to adapt to the type of stain (such as automatically increasing the cleaning intensity when the difficult signal is triggered). This ensures efficient removal of stubborn stains while avoiding damage to the coating of the board due to over-cleaning, thus achieving the dual goals of cleaning quality and equipment protection.

[0129] Finally, at the level of data-driven management and predictive maintenance, this invention introduces for the first time a quantitative analysis mechanism for data recovery effectiveness coefficient and operability coefficient. By normalizing historical fault parameters, recovery parameters, resource parameters, and performance parameters, it generates feasible / difficult-to-operate signals and excellent / intervention signals, and dynamically optimizes data backup strategies (such as compensating for backup interval time). This constructs a full lifecycle management system covering equipment status assessment, fault prediction, and operation and maintenance optimization, significantly reducing equipment downtime and maintenance costs.

[0130] Through the above-mentioned technological innovations, this invention not only solves the core pain points in the cleaning operations of small-capacity distributed photovoltaic power stations, but also provides a landmark technical solution for the intelligent and refined development of the photovoltaic equipment cleaning field.

Claims

1. A control method for an intelligent cleaning robot for photovoltaic panels, characterized in that, Includes the following steps: Step 1: Data acquisition, collecting data on robot tilt angle, pressure against photovoltaic panel, motion acceleration, historical fault parameters, and recovery parameters; Step 2: Environmental monitoring, real-time monitoring of wind speed, temperature, relative humidity and dust density on the surface of photovoltaic panels in the working environment; Step 3: Path planning, integrating the CAD model of the small-capacity distributed photovoltaic power station with the edge and gap features of the panels acquired by machine vision to generate a cleaning path; Step 4: Stain detection. The image acquisition module acquires image data of the near-infrared light source in the area to be detected in the state of being off / on, calculates the transmittance index, and segments the stain area. Step 5: Cleaning execution. The cleaning parameters are dynamically adjusted according to the type of stain and environmental parameters. The cleaning is performed in a coordinated manner of "pre-blowing by airflow + brush cleaning + secondary airflow cleaning with a delay of 0.3s → secondary airflow blowing by airflow". Step 6: Summarize and analyze the data based on cleaning path coverage, stain removal rate, equipment energy consumption, and fault data to generate feasible signals, difficult signals, excellent signals, or intervention signals. Step 7: Dual-motor synchronous control eliminates motor displacement differences in real time, ensuring stable operation; Step 8: Track and backup optimization, adjust cleaning tracks and data backup strategies according to signal type.

2. The control method for an intelligent cleaning robot for photovoltaic panels according to claim 1, characterized in that, The data collection in step 1 specifically includes: Step 1.1: Real-time acquisition of raw data on tilt angle, contact pressure, and motion acceleration using a six-axis gyroscope, a high-precision tilt sensor, and a pressure sensor distributed in the contact area between the robot chassis and the adsorption surface; Step 1.2: Categorize and store historical fault parameters and recovery parameters to establish a fault-recovery database.

3. The control method for an intelligent cleaning robot for photovoltaic panels according to claim 1, characterized in that, The environmental monitoring in step 2 specifically includes: Step 2.1: Use a polling method to collect data. Wind speed and temperature / humidity data are updated every 1 second, and dust density data is collected every 3 seconds. Step 2.2: Analyze the rate of change of environmental parameters in real time. When the wind speed is greater than 10 m / s or the dust density is greater than 1 mg / m³, high-frequency monitoring is triggered and the sampling interval is shortened to 0.5 s.

4. The control method for an intelligent cleaning robot for photovoltaic panels according to claim 1, characterized in that, The path planning in step 3 specifically includes: Step 3.1: Integrate the CAD model of the photovoltaic power station with real-time images of the panels captured by industrial cameras to identify the panel edges and splicing seams; Step 3.2: Establish a coordinate system with the upper left corner of the plate as the origin, avoid the metal bracket area of ​​the frame, and densify the path within 5cm on both sides of the splicing seam to generate a cleaning path containing straight lines and turning sections. Step 3.3: When a good signal is received, adjust the backup node settings of the path planning according to the compensation backup interval.

5. The control method for an intelligent cleaning robot for photovoltaic panels according to claim 4, characterized in that: Step 3 also includes a path adjustment step: when a temporary obstacle is detected, local path replanning is performed, and the obstacle bypass path maintains a safe distance of ≥3cm from the edge of the obstacle; if an intervention signal is received during the path planning process, the current planning is paused and a manual intervention instruction is awaited.

6. The control method for an intelligent cleaning robot for photovoltaic panels according to claim 1, characterized in that, The stain detection in step 4 specifically includes: Step 4.1: Calculate the pixel grayscale difference between the initial background image and the initial target image, and determine the transmittance index of each pixel; Step 4.2: Based on the degree of deviation between the light transmittance index and the representative index of no stains, extract the target stain area through connected components; Step 4.3: After cleaning, re-acquire images to verify the cleaning effectiveness. If the change in stain characteristic indicators is less than or equal to the preset threshold, change the cleaning method.

7. The control method for an intelligent cleaning robot for photovoltaic panels according to claim 1, characterized in that, The cleaning process in step 5 specifically includes: Step 5.1: Activate the two sets of independently lifting symmetrically arranged sweeping brushes, the high-pressure air jet unit, and the four sets of symmetrically arranged propellers; Step 5.2: First flight dust removal. After takeoff, during the flight to the starting end of the current photovoltaic panel, adjust the propeller angle to blow downwards and loosen the surface dust. Step 5.3: When a "can proceed" signal is received, perform cleaning according to the normal parameters; when a "difficult to proceed" signal is received, adjust parameters such as cleaning pressure and brush speed to adapt to changes in equipment status. Step 5.4: After identifying the stains, first start the high-pressure airflow for pre-blowing, then start the brush group for cleaning. After cleaning is completed, delay for 0.3 seconds to perform a second airflow for cleaning. Step 5.5: Second flight to blow away dust. After detaching from the photovoltaic panel, the propeller angle is adjusted again during the flight to blow away residual dust. Step 5.6: The propeller supports forward and reverse rotation. When flying, forward rotation provides lift, and when adsorbing and fixing, reverse rotation generates a downward pressure of 0.2~0.5MPa. The pressure sensor provides feedback for dynamic adjustment.

8. The control method for an intelligent cleaning robot for photovoltaic panels according to claim 1, characterized in that, The summary and analysis in step 6 specifically includes: Step 6.1: Receive data on cleaning path coverage, stain removal rate, equipment energy consumption, and number of abnormal shutdowns, and combine them with historical fault parameters and recovery parameters; Step 6.2: Normalize the fault parameters and recovery parameters to obtain the data recovery effect coefficient, compare it with the data recovery effect threshold, and generate a feasible signal or a difficult signal. Step 6.3: After obtaining the difficult-to-operate signal, normalize the resource parameters and performance parameters to obtain the operability coefficient, and compare it with the operability threshold to generate the good signal or intervention signal.

9. The control method for an intelligent cleaning robot for photovoltaic panels according to claim 8, characterized in that, Step 6 further includes a backup strategy optimization step: obtaining the current data backup interval time, comprehensively calculating the data recovery effect coefficient, data recovery effect threshold and database backup time interval to obtain the compensation backup interval time; replacing the original data backup interval time with the compensation backup interval time to optimize the data backup strategy.

10. The control method for an intelligent cleaning robot for photovoltaic panels according to claim 1, characterized in that, The dual-motor synchronous control in step 7 specifically includes: Step 7.1: Real-time acquisition of the displacement difference between the upper and lower motors using a non-contact angle sensor; Step 7.2: The fuzzy controller generates an adjustment signal based on the displacement difference to control the motor speed, so that the time for the motor to recover stable operation after misalignment is ≤1.2 seconds.

11. A photovoltaic panel cleaning robot control system, used to implement the photovoltaic panel cleaning robot control method according to any one of claims 1 to 10, characterized in that, include: The data acquisition module is used to collect data on the robot's tilt angle, pressure against the photovoltaic panel, motion acceleration, historical fault parameters, and recovery parameters. The environmental monitoring module is used to monitor the wind speed, temperature, relative humidity, and dust density on the surface of the photovoltaic panels in the working environment. The path planning module generates cleaning paths based on the CAD model of a small-capacity distributed photovoltaic power station and the panel features collected in real time by machine vision. The cleaning execution module is used to perform cleaning operations on the photovoltaic panels; The summary and analysis module is used to summarize and analyze the cleaning results and equipment data and generate control signals; The image acquisition module is configured to acquire near-infrared image data of the area to be detected on the photovoltaic panel; The control unit communicates with each module and generates control commands. Dual-motor drive assembly ensures stable robot operation; The communication module enables data interaction and resume transmission even after network outages.

12. The photovoltaic panel cleaning robot control system according to claim 11, characterized in that: The data acquisition module includes a distributed layout of six-axis gyroscopes, high-precision tilt sensors, and pressure sensors, installed in the contact area between the robot chassis and the adsorption surface; it is equipped with a data storage unit to classify and store the robot's historical fault parameters and recovery parameters.

13. A photovoltaic panel cleaning robot control system according to claim 11, characterized in that: The environmental monitoring module integrates an ultrasonic wind speed sensor, a temperature and humidity sensor, and a laser dust sensor, and is installed inside the windproof cover on the top of the robot. It adopts a polling acquisition method, with wind speed and temperature and humidity data updated every 1 second and dust density data collected every 3 seconds. When the wind speed is greater than 10 m / s or the dust density is greater than 1 mg / m³, the sampling interval is shortened to 0.5 seconds.

14. A photovoltaic panel cleaning robot control system according to claim 11, characterized in that: The path planning module is equipped with an industrial camera and a CAD model analysis unit to identify the board frame and splicing seams; it has a built-in coordinate system establishment unit to generate a cleaning path with the upper left corner of the board as the origin, and the path is densified within a 5cm range on both sides of the splicing seam. The path includes straight segments and turning segments; it is equipped with a backup node adjustment unit to adjust the backup node settings of the path planning according to the good signal; the path planning module also includes an obstacle detection unit and an intervention processing unit; when the obstacle detection unit detects a temporary obstacle, it generates an obstacle avoidance path that maintains a safe distance of ≥3cm from the edge of the obstacle; when the intervention processing unit receives an intervention signal, it pauses the current planning and triggers a manual intervention command receiving mechanism.

15. A photovoltaic panel cleaning robot control system according to claim 11, characterized in that: The system includes two sets of independent lifting sweeping brushes, a high-pressure airflow jet unit, and four sets of propellers arranged symmetrically; it is equipped with a parameter adjustment unit that can dynamically adjust sweeping parameters according to feasible and difficult signals; and it has a built-in sweeping timing control unit to achieve coordinated control of "flight pre-blowing → airflow pre-blowing + brush group sweeping + 0.3s delayed secondary airflow sweeping → flight secondary dust blowing".

16. A photovoltaic panel cleaning robot control system according to claim 11, characterized in that: The summary and analysis module is configured with a data receiving unit, a normalization processing unit, and a signal generation unit. The data receiving unit receives cleaning path coverage, stain removal rate, equipment energy consumption, number of abnormal shutdowns, and historical fault / recovery parameters. The normalization processing unit normalizes the fault parameters and recovery parameters to obtain the data recovery effect coefficient, and normalizes the resource parameters and performance parameters to obtain the operability coefficient. The signal generation unit generates feasible signals, difficult signals, excellent signals, or intervention signals by comparing with corresponding thresholds. The summary and analysis module also includes a backup strategy optimization unit. The backup strategy optimization unit calculates the compensation backup interval time based on the data recovery effect coefficient, the data recovery effect threshold, and the database backup time interval, which is used to replace the original data backup interval time.

Images