An Optimization Method for Photovoltaic Panel Cleaning Strategy

Abstract

This invention relates to the field of photovoltaic power generation operation and maintenance technology, and particularly to a method for optimizing photovoltaic panel cleaning strategies. The method includes: responding to a cleaning command or pollution warning, controlling a cleaning robot to perform initial cleaning operations and collecting initial operation spectrum data; extracting features from the initial operation spectrum data to obtain residual pollution characteristics, cleaning response characteristics, and environmental corrosion factors for each photovoltaic panel unit; assessing the cleaning urgency of each photovoltaic panel unit; constructing a priority cleaning queue; adjusting the queue length and execution order to generate an executable cleaning task package; matching the optimal combination of composite cleaning actions from a preset multi-mode cleaning strategy library and optimizing and determining a refined control parameter set; and executing the executable cleaning task package using the refined control parameter set. This invention enables cleaning strategies to be highly adaptable to different pollution types and levels in real-world scenarios, improving cleaning effectiveness while optimizing resource consumption.

Description

Technical Field

[0001] This invention relates to the technical field of photovoltaic power generation operation and maintenance, and in particular to an optimization method for photovoltaic panel cleaning strategies. Background Technology

[0002] During operation, photovoltaic (PV) panels accumulate pollutants such as dust, sand, bird droppings, and industrial oil. These pollutants reduce the photoelectric conversion efficiency of the PV panels, affecting the power generation of the power plant. To maintain the efficient operation of PV panels, regular cleaning is necessary. Currently, the main method for cleaning PV panels is through automated cleaning equipment, such as track-mounted or mobile cleaning robots, which typically operate using fixed-cycle, fixed-route, or fixed-cleaning programs.

[0003] Existing photovoltaic panel cleaning strategies cannot identify the differences in the degree of contamination between different panels or areas in a photovoltaic array. This can lead to over-cleaning of slightly contaminated areas, wasting water and electricity and causing equipment damage, while under-cleaning of stubborn or heavily contaminated areas may result in poor results. Summary of the Invention

[0004] This invention provides an optimized method for cleaning photovoltaic panels, which can effectively solve the problems in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for optimizing photovoltaic panel cleaning strategies includes:

[0007] In response to cleaning instructions or pollution warnings, the cleaning robot is controlled to perform an initial cleaning operation on the target photovoltaic array along a preset track based on preset reference parameters, and the first operation spectrum data is collected.

[0008] Feature extraction was performed on the initial operation spectrum data to obtain the residual pollution characteristics, cleaning response characteristics, and environmental corrosion factors of each photovoltaic panel unit;

[0009] Based on the residual pollution characteristics, cleaning response characteristics, and environmental corrosion factors, the cleaning urgency of each photovoltaic panel unit is assessed; and all photovoltaic panel units are ranked according to the cleaning urgency to construct a priority cleaning queue.

[0010] Obtain the real-time available resource status of the cleaning system, adapt it to the priority cleaning queue, adjust the queue length and execution order, and generate an executable cleaning task package.

[0011] For each task in the executable cleaning task package, based on its corresponding residual contamination characteristics and cleaning response characteristics, the optimal combination of composite cleaning actions is matched from the preset multi-mode cleaning strategy library, and the corresponding set of refined control parameters is optimized and determined.

[0012] The cleaning robot is driven to execute the executable cleaning task package according to the refined control parameter set.

[0013] Furthermore, the initial operation spectrum data includes at least the pre-operation contamination image, operation trajectory, real-time cleaning parameters, and post-operation surface condition image.

[0014] Furthermore, the extraction of the cleaning response features includes:

[0015] By comparing and analyzing the pollution images of the same photovoltaic panel unit before and after the operation, the stubborn pollution areas and their morphology that were not effectively removed under the preset benchmark parameters can be identified.

[0016] Analyze the spatiotemporal relationship between the robot's operational trajectory and the stubbornly polluted area to generate an effect attenuation coefficient;

[0017] The environmental corrosion factors include the potential corrosivity level of the pollution type on the board surface; the pollution types include dust, grease, biological excrement and chemical residues.

[0018] Furthermore, the resource status includes the base station's clean water storage capacity, the remaining sewage tank capacity, and the robot's battery life.

[0019] Adapting the real-time available resource status to the priority cleaning queue, adjusting the queue length and execution order, includes:

[0020] Photovoltaic panel units whose dynamic cleaning urgency value exceeds the first threshold in the priority cleaning queue are marked as critical targets, ensuring that they are included in the executable cleaning task package;

[0021] Calculate the maximum cleaning load that can be supported based on the real-time available resource status;

[0022] If the total load of the priority cleaning queue exceeds the maximum cleaning load, tasks from non-critical targets will be temporarily suspended from low to high based on the dynamic cleaning urgency value, until the total load of the priority cleaning queue matches the maximum cleaning load.

[0023] Furthermore, the combined cleaning action sequence includes the timing arrangement of at least two actions among dry brushing, wet washing, spraying, and scraping;

[0024] The refined control parameter set includes the brush head rotation speed, travel speed, water spray pressure, water flow temperature, detergent ratio, and the attitude angle of the robotic arm end effector for different contaminated areas.

[0025] The process of optimizing and determining the set of refined control parameters adopts a parameter optimization algorithm based on physical model simulation, with the optimization objective being to minimize water and electricity consumption and operation time while meeting the cleaning effect threshold.

[0026] Furthermore, the construction of the multi-mode cleaning strategy library is based on historical data; when matching the optimal combination of composite cleaning actions, cases with similar pollution characteristics and cleaning response characteristics in the historical data are first retrieved from the multi-mode cleaning strategy library, and then a reward value is assigned according to the final performance data achieved by the strategy adopted in the case, thereby driving strategy selection.

[0027] Furthermore, the urgency of cleaning each photovoltaic panel unit was assessed, including:

[0028] Pre-defined assessment weights were assigned to residual contamination characteristics, cleaning response characteristics, and environmental corrosion factors;

[0029] The indicators of each feature are quantized into quantized values ​​within a unified numerical range;

[0030] A weighted summation algorithm is used to calculate the cleaning urgency value for each photovoltaic panel unit.

[0031] Furthermore, the quantization value of the cleaning response characteristic is directly adopted using the effect attenuation coefficient;

[0032] The quantitative value of the environmental corrosion factor is obtained by mapping according to the preset corrosion level corresponding to the pollution type.

[0033] Further, the priority cleaning queue is constructed, including:

[0034] The photovoltaic panel units are sorted from highest to lowest cleaning urgency value to form an initial queue;

[0035] If multiple photovoltaic panel units have the same cleaning urgency value, they are then sorted by the corrosivity level of the environmental corrosion factors, with higher corrosion levels appearing first.

[0036] Furthermore, the preset benchmark parameters are stored in a parameter library according to the photovoltaic panel material and installation scenario;

[0037] After the initial cleaning operation is triggered, the cleaning robot obtains the corresponding baseline parameters from the edge computing node and performs a pre-start verification of the actuator. After confirming that the parameter error is within the preset range, the operation is started.

[0038] The technical solution of this invention achieves the following technical effects: By combining the initial cleaning operation based on preset benchmark parameters with the collection and analysis of first-operation spectrum data, not only can the initial contamination state of the photovoltaic panel surface be obtained, but more importantly, by analyzing the residual contamination characteristics and cleaning response characteristics after the first cleaning, stubborn contamination areas and their characteristics that are difficult to remove under standard cleaning procedures can be identified, enabling a preliminary diagnosis of contamination differences and cleaning equipment performance; By combining the cleaning urgency ranking based on multi-dimensional feature evaluation with the dynamic adaptation of real-time available resource status, the urgency of efficiency loss caused by contamination and the current hydropower reserves can be considered simultaneously when planning cleaning tasks. Physical constraints such as equipment battery life are considered. The resulting execution task packages ensure that, under limited resource conditions, the cleaning tasks with the highest overall urgency are prioritized, achieving a real-time balance between cleaning demand and system supply capacity. The multi-mode cleaning strategy matching based on specific pollution characteristics is combined with corresponding refined control parameter optimization. Instead of simply calling fixed programs, it selects the most suitable composite cleaning action sequence from the strategy library based on the specific residual pollution characteristics and cleaning response characteristics of each task unit, and further optimizes the detailed control parameters of each action. This allows the cleaning strategy to be highly adaptable to actual scenarios with different types and degrees of pollution, thereby improving cleaning effectiveness while optimizing resource consumption.

[0039] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating the optimization method for photovoltaic panel cleaning strategy of the present invention. Detailed Implementation

[0042] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0044] like Figure 1 As shown, the photovoltaic panel cleaning strategy optimization method of the present invention specifically includes the following steps:

[0045] Step S1: In response to a cleaning command or pollution warning, control the cleaning robot to perform an initial cleaning operation on the target photovoltaic array along a preset track based on preset reference parameters, and collect the first operation spectrum data;

[0046] Step S2: Extract features from the initial operation spectrum data to obtain the residual pollution characteristics, cleaning response characteristics, and environmental corrosion factors of each photovoltaic panel unit;

[0047] Step S3: Based on residual contamination characteristics, cleaning response characteristics, and environmental corrosion factors, assess the cleaning urgency of each photovoltaic panel unit; and rank all photovoltaic panel units according to cleaning urgency to construct a priority cleaning queue.

[0048] Step S4: Obtain the real-time available resource status of the cleaning system, adapt it to the priority cleaning queue, adjust the queue length and execution order, and generate an executable cleaning task package.

[0049] Step S5: For each task in the executable cleaning task package, based on its corresponding residual contamination characteristics and cleaning response characteristics, match the optimal combination of composite cleaning actions from the preset multi-mode cleaning strategy library, and optimize and determine the corresponding set of refined control parameters.

[0050] Step S6: Drive the cleaning robot to execute the executable cleaning task package according to the refined control parameter set.

[0051] In this embodiment, by combining the initial cleaning operation based on preset benchmark parameters with the collection and analysis of first-operation spectrum data, not only can the initial contamination state of the photovoltaic panel surface be obtained, but more importantly, by analyzing the residual contamination characteristics and cleaning response characteristics after the first cleaning, stubborn contamination areas and their characteristics that are difficult to remove under standard cleaning procedures can be identified, achieving a preliminary diagnosis of contamination differences and cleaning equipment performance. Furthermore, by combining the cleaning urgency ranking based on multi-dimensional feature evaluation with dynamic adaptation to real-time available resource status, the planning of cleaning tasks can simultaneously consider the urgency of efficiency losses caused by contamination and current water and electricity reserves, equipment operating time, and other factors. The system employs a theoretical constraint mechanism; the resulting execution task package ensures that, under limited resource conditions, the cleaning tasks with the highest overall urgency are prioritized, achieving a real-time balance between cleaning demand and system supply capacity. It combines multi-mode cleaning strategy matching tailored to specific contamination characteristics with corresponding refined control parameter optimization. Instead of simply calling fixed programs, it selects the most suitable composite cleaning action sequence from the strategy library based on the specific residual contamination characteristics and cleaning response characteristics of each task unit, and further optimizes the detailed control parameters of each action. This allows the cleaning strategy to be highly adaptable to actual scenarios with different types and degrees of contamination, thereby improving cleaning effectiveness while optimizing resource consumption.

[0052] In some embodiments of the present invention, existing automatic cleaning equipment suffers from problems such as insufficient triggering accuracy, inadequate track positioning accuracy, and limited data acquisition dimensions and synchronization when performing initial cleaning operations. Cleaning triggering relies solely on fixed cycles or single sensor signals, easily leading to false triggering or missed triggering. Track positioning lacks an effective calibration mechanism, resulting in insufficient consistency of the work trajectory and variations in initial cleaning conditions among different photovoltaic panel units. To address these issues, this invention combines the modular layout characteristics of distributed photovoltaic panel arrays, the localized concentration patterns of contamination distribution, and the track operation constraints of the cleaning robot. By clarifying the quantitative judgment criteria for triggering conditions, improving the standardization of track positioning and operation parameters, and designing a multi-device collaborative data acquisition and synchronization mechanism, the initial cleaning operation acquires spectral data while completing preliminary cleaning. The specific implementation is as follows:

[0053] Step S11: The cleaning command supports two triggering modes: local and remote. Local triggering is manually issued through the touch terminal of the photovoltaic power station operation and maintenance platform, while remote triggering is sent through the Internet of Things terminal. The command includes the target photovoltaic array number, operation start time and preset benchmark parameter identifier. The command transmission adopts an industrial-grade communication protocol to ensure real-time performance.

[0054] Pollution early warnings are triggered through multi-sensor fusion. Reflectivity sensors and dust concentration sensors are deployed at the edge of each working unit of the photovoltaic array. The reflectivity sensors are installed at a preset height above the photovoltaic panel surface at a preset angle, collecting data at a fixed frequency. The preset reflectivity threshold is set according to the photovoltaic panel material. When the average reflectivity value collected multiple times is lower than the threshold, a primary warning is triggered. The dust concentration sensor collects particulate matter concentration at a fixed frequency. When the concentration of particulate matter of a specific size reaches a preset value and remains so for a preset duration, a secondary warning is triggered. A high-definition camera is deployed in the central area of ​​the photovoltaic array to capture panoramic images at fixed intervals. The pollution area ratio is analyzed through edge computing nodes. When the ratio reaches a preset proportion, a high-level warning is triggered. Different warning levels correspond to different preset benchmark parameters, with high-level warnings corresponding to higher cleaning intensity parameters.

[0055] Step S12: The track adopts a modular structure design, with standard sections and corner sections designed according to the spacing of the photovoltaic panel array. It is fixed to the photovoltaic bracket through a detachable connection method to control the installation error within the preset range; displacement measuring components are set on the inner side of the track along the length direction to ensure the accuracy of the travel distance measurement.

[0056] The cleaning robot is equipped with a laser positioning sensor and a displacement reading component. The laser positioning sensor is used to identify the positioning marks at both ends of the track. The positioning marks include the work unit number and calibration information. The displacement reading component works with the track displacement measuring component to collect the robot's movement displacement in real time and sample at a fixed frequency. The robot control system integrates the laser positioning data and displacement data, and uses a filtering algorithm to correct positioning errors, ensuring the repeatability of the work trajectory and achieving consistency of the initial cleaning work trajectory for different batches.

[0057] Step S13: Preset benchmark parameters are stored in the robot control system parameter library according to the photovoltaic panel material and installation scenario. The parameter library supports remote updates. The parameters include brush head rotation speed, robot travel speed, water spray pressure, single operation time and cleaning agent ratio. The brush head rotation speed is set according to the hardness grade of the photovoltaic panel material, the travel speed is adaptively adjusted according to the array area, the water spray pressure is adjusted according to the pollution warning level, and the cleaning agent ratio defaults to clean water cleaning. When a high warning occurs, it is automatically adjusted to a preset ratio of neutral cleaning agent.

[0058] After the initial cleaning operation is triggered, the robot obtains the reference parameters of the corresponding target array from the edge computing node through the industrial Ethernet. After the parameters are sent out, the robot control system performs a pre-start verification of the brush head motor, water pump, and drive motor. After confirming that the error of each parameter is within the preset range, the operation is started. If the verification fails, a fault signal is fed back and the operation is suspended.

[0059] Step S14: Pre-operation pollution image acquisition is achieved through a dual-camera assembly on the robot, including a panoramic camera and a close-up camera. The panoramic camera is installed at a preset height on the robot, perpendicular to the surface of the photovoltaic panel, and is used to capture panoramic images of the work unit. The close-up camera is controlled by an adjustable gimbal and captures partial images of the panel surface at fixed intervals along the track travel direction, covering the entire photovoltaic panel surface. The camera is equipped with an ambient light sensor, which automatically turns on the supplementary light when the light intensity is lower than a preset value. The brightness of the supplementary light is adaptively adjusted according to the ambient light intensity to ensure consistent image brightness. During pre-operation image acquisition, the robot moves to the starting position of the work unit and calibrates the shooting angle using a laser positioning sensor.

[0060] The operation trajectory is acquired through the combination of the rotation speed acquisition component and the displacement measurement component integrated into the robot drive wheel. The movement speed, position coordinates and attitude angle are collected in real time, and the data is sampled at a fixed frequency to generate three-dimensional operation trajectory data, accurately marking the correspondence between the cleaning action and the position on the board surface.

[0061] Real-time cleaning parameter acquisition is achieved through sensors deployed in key parts of the robot. Pressure and flow sensors are deployed in the clean water pipeline to collect water spray pressure and clean water consumption, respectively; speed sensors are deployed in the brush head motor to collect the brush head rotation speed; and current sensors are deployed in the drive motor to indirectly reflect the travel resistance. All parameters are collected at a fixed frequency, and the data includes parameter name, value and unit, and is synchronized with the trajectory data in time.

[0062] The surface condition image acquisition after the operation is completed is carried out. The robot returns to the starting position of the operation unit along the original track and is calibrated with the same positioning point as before the operation by the laser positioning sensor to ensure that the shooting angle and height of the panoramic camera are consistent with those before the operation. The local close-up camera takes local images at the shooting interval and position before the operation, and the lighting conditions are consistent with those before the operation. After the acquisition is completed, the consistency of the perspective of the images before and after the operation is verified by the image comparison algorithm. If the matching degree reaches the preset standard, the acquisition is deemed valid; otherwise, the acquisition is repeated.

[0063] Step S15: The robot has a built-in industrial-grade storage component and adopts a standardized file system. Folders are created according to the target array number, operation date and timestamp to classify and store images before operation, operation trajectory, real-time cleaning parameters and images after operation. Image files and parameter files adopt a standardized format.

[0064] After the task is completed, the robot establishes a connection with the edge computing node through the high-speed communication module, uploads all collected data using the file transfer protocol, and verifies the integrity of the data through a data verification algorithm after the upload is completed. If the verification fails, the data is re-uploaded. The edge computing node categorizes and archives the data according to the photovoltaic panel unit number and the task batch, and associates it with the distributed database.

[0065] In this embodiment, multi-sensor fusion and hierarchical early warning design improve the accuracy of pollution early warning, avoiding resource waste caused by false triggers and pollution accumulation caused by missed triggers; modular track design and robot multi-sensor fusion positioning ensure the repeatability of the operation trajectory and keep the initial cleaning conditions of different photovoltaic panel units consistent; dual-camera acquisition combined with adaptive lighting and angle calibration ensures the consistency of brightness and viewing angle of images before and after operation, comprehensively covering the pollution status of the panel surface and accurately capturing details of local stubborn pollution; time-synchronized acquisition of operation trajectory, real-time cleaning parameters and pollution images establishes the correlation between the three, ensuring the accuracy of cleaning response feature extraction; standardized data storage format and remote synchronization verification mechanism ensure the integrity and compatibility of the collected data; the archiving design of edge computing nodes supports rapid data retrieval in subsequent steps, avoiding process interruption due to data loss or format incompatibility.

[0066] In some embodiments of the present invention, for step S2, by integrating image, trajectory, and parameter data, a correlation logic between the pollution state and the cleaning action is established, clarifying the correspondence between pollution type and corrosivity, and achieving accurate extraction of residual pollution characteristics, cleaning response characteristics, and environmental corrosion factors. The specific implementation is as follows:

[0067] Step S21: Preprocess the spectrum data of the first operation. The pollution image before operation and the surface state image after operation are processed by grayscale conversion, noise reduction and image registration to eliminate the interference caused by the shooting environment. The operation trajectory data is sorted by timestamp and associated and aligned with the real-time cleaning parameters to form a three-dimensional data chain of time and position parameters. All data are classified according to the photovoltaic panel unit number to ensure that the various types of data of a single photovoltaic panel correspond one-to-one.

[0068] Step S22: Extract the residual pollution features of each photovoltaic panel unit. Based on the registered pre-operation pollution image and post-operation surface state image, the image difference algorithm is used to obtain the residual pollution area. Through threshold segmentation and morphological processing, the area, contour, gray-scale mean and distribution density of the residual pollution are extracted. The proportion of the residual pollution area to the total area of ​​the photovoltaic panel unit is calculated. The gray-scale mean reflects the thickness of the residual pollution. The distribution density is characterized by the degree of pixel clustering in the residual area. A set of residual pollution features is formed by combining these features.

[0069] Step S23: Extract the cleaning response characteristics of each photovoltaic panel unit, and execute the following process:

[0070] By comparing and analyzing the pollution images of the same photovoltaic panel unit before and after the operation, the correspondence between the pollution area before the operation and the residual pollution area after the operation is located by the feature point matching algorithm. Areas that existed before the operation and did not disappear after the operation are selected as stubborn pollution areas. The contour of the stubborn pollution area is extracted by the edge detection algorithm to obtain its shape, size and location information, thus completing the identification of the stubborn pollution area and its morphology.

[0071] The system retrieves the operational trajectory data and real-time cleaning parameters of the photovoltaic panel unit, synchronizes the location information of the stubborn contamination area with the robot's operational position according to the timestamp, and analyzes the correspondence between parameters such as the robot's operation time, brush head rotation speed, and water spray pressure in the stubborn contamination area and the stubborn contamination morphology. It calculates the ratio of the average cleaning parameters in the stubborn contamination area to the average cleaning parameters in the entire area, and combines the area attenuation ratio of the stubborn contamination area to generate an effect attenuation coefficient through weighted calculation. The larger the value of this coefficient, the worse the adaptability of the preset benchmark parameters to the removal of the corresponding type of contamination.

[0072] Step S24: Determine the environmental corrosion factor for each photovoltaic panel unit, and proceed according to the following process:

[0073] Based on the grayscale and texture features of the pre-operation contamination images and the changes in clean water consumption in real-time cleaning parameters, combined with a historical contamination type database, a classification algorithm is used to identify the contamination type and determine whether the contamination belongs to dust, grease, biological excrement, or chemical residue. Dust contamination is identified by grayscale mean and particle size distribution; grease contamination is identified by image texture smoothness and increase in clean water consumption; biological excrement contamination is identified by contour irregularity and grayscale variance; and chemical residue contamination is identified by grayscale anomalies in the residual area after operation.

[0074] Based on the chemical composition and physical characteristics of different types of pollution, a standard for classifying corrosivity levels is established. For example, dust pollution has the lowest corrosivity level due to its physical adhesion characteristics; oil pollution has a medium corrosivity level because it easily adsorbs harmful substances; biological excrement pollution has a medium to high corrosivity level because it contains acidic or alkaline substances; and chemical residue pollution has the highest corrosivity level because it contains chemical reagents. The corrosivity level corresponding to the pollution type is used as the main indicator of environmental corrosion factors, forming complete environmental corrosion factor data.

[0075] In this embodiment, the extraction of residual pollution features combines image differencing, threshold segmentation, and morphological processing, and integrates multi-dimensional indicators such as area, contour, and grayscale mean to comprehensively reflect the actual state of residual pollution. The extraction of cleaning response features is achieved through the identification of stubborn pollution areas and spatiotemporal correlation analysis. The effect attenuation coefficient directly reflects the degree of adaptation between the preset benchmark parameters and the pollution type, and the extraction logic fits the actual physical process of the cleaning operation. The determination of environmental corrosion factors is based on the accurate identification of pollution types and the scientific classification of corrosion levels, establishing a clear correlation between pollution types and plate corrosion, so that environmental corrosion factors have practical application value.

[0076] In a specific implementation, as one example, regarding step S3, by combining the characteristics of residual pollution, cleaning response characteristics, the correlation with environmental corrosion factors, the actual influencing factors of photovoltaic panel cleaning needs, and the optimization objectives of queue scheduling, a precise assessment of cleaning urgency and a reasonable queue ordering are achieved through a combination of multi-dimensional indicator weighted fusion, quantitative evaluation, and orderly sorting. The specific implementation is as follows:

[0077] Step S31: Determine the evaluation index weights for each feature. Based on historical cleaning data, photovoltaic panel material characteristics, and pollution impact patterns, and combined with expert experience, assign evaluation weights to residual pollution features, cleaning response features, and environmental corrosion factors. The weight of residual pollution features is set according to the degree of impact of residual pollution on power generation efficiency; the weight of cleaning response features is set according to the cleaning adaptation difficulty reflected by the effect decay coefficient; and the weight of environmental corrosion factors is set according to the degree of impact of corrosion level on photovoltaic panel lifespan. The weight allocation results are stored in the evaluation system parameter library and can be dynamically adjusted based on actual application scenarios.

[0078] Step S32: Quantify the evaluation indicators of each feature. Convert the indicators of residual pollution features, cleaning response features, and environmental corrosion factors into unified quantitative values ​​with a range of 0-1. The quantitative value of residual pollution features is obtained by weighting the area ratio and the grayscale mean. The area ratio is the ratio of the residual pollution area to the total area of ​​the photovoltaic panel unit. The grayscale mean is mapped to the corresponding quantitative value according to a preset interval. The two are weighted and summed to obtain the quantitative value of residual pollution features. The quantitative value of cleaning response features is directly obtained by using the effect attenuation coefficient. This coefficient has been converted into a value in the 0-1 interval through previous calculations. The larger the value, the higher the difficulty of cleaning adaptation. The quantitative value of environmental corrosion factors is obtained by mapping according to the corrosion level. For example, the lowest corrosion level corresponds to a quantitative value of 0.2, the medium level corresponds to 0.4, the medium-high level corresponds to 0.7, and the highest level corresponds to 1.0.

[0079] Step S33: Calculate the cleaning urgency of each photovoltaic panel unit. Use a weighted summation algorithm to multiply each feature quantification value by its corresponding weight and then sum them to obtain the cleaning urgency value of each photovoltaic panel unit. The urgency value ranges from 0 to 1. The larger the value, the more urgent the cleaning requirement.

[0080] Step S34: Construct a priority cleaning queue. Sort each photovoltaic panel unit from highest to lowest cleaning urgency value to form an initial queue. If multiple photovoltaic panel units have the same cleaning urgency value, sort them according to the corrosivity level of the environmental corrosion factor, with photovoltaic panel units of higher corrosivity level ranking higher. If the corrosion levels are also the same, sort them according to the area ratio of residual pollution characteristics, with larger area ratio ranking higher. Finally, a priority cleaning queue is formed. The queue information includes the photovoltaic panel unit number, cleaning urgency value, characteristic quantification value, and ranking position, and is stored in the queue management module.

[0081] In this embodiment, the scientific allocation of weights for each characteristic indicator comprehensively considers the impact of residual pollution, cleaning adaptability, and corrosivity on cleaning needs, ensuring that the assessment results fully reflect the actual urgency of cleaning. The unified quantitative processing of each characteristic indicator eliminates the dimensional differences between different indicators, ensuring the consistency of the assessment process. The cleaning urgency value calculated by the weighted summation algorithm can objectively quantify the urgency of cleaning needs for each photovoltaic panel unit. The priority cleaning queue constructed based on the urgency value and auxiliary rules can prioritize the cleaning of photovoltaic panel units with high urgency, high corrosivity, and high residual pollution.

[0082] In a specific implementation, as one example, for step S4, based on the urgency level of the priority cleaning queue, the multi-dimensional constraints of the cleaning system resources, and the task load, key targets are locked, multi-dimensional resource fusion calculations are performed, and non-critical targets are orderly postponed to achieve precise matching of resources and queues. The specific implementation is as follows:

[0083] Step S41: Obtain the real-time available resource status of the cleaning system. Data is collected by sensors deployed on the base station and the cleaning robot: the clean water storage capacity of the base station is collected by a liquid level sensor; the remaining wastewater tank capacity is obtained by combining the liquid level sensor with volume calculation to determine the remaining available space; the robot's endurance is obtained by collecting the remaining power through the battery management system, and combined with historical operation energy consumption data to calculate the supported operation time. The collected data is updated at a fixed frequency and synchronously transmitted to the resource management module. The data includes resource type, real-time value, unit, and warning threshold.

[0084] Step S42: Mark key targets. Based on the cleaning urgency value in the priority cleaning queue, set a first threshold. This threshold is set based on historical cleaning data, the importance of photovoltaic panels, and operation and maintenance requirements. It is stored in the system parameter library and supports dynamic adjustment. Traverse the priority cleaning queue and mark photovoltaic panel units with cleaning urgency values ​​exceeding the first threshold as key targets. Key target information is stored separately, and their cleaning tasks are clearly defined as mandatory items to ensure that they are not postponed or deleted during subsequent adjustments.

[0085] Step S43: Calculate the maximum supportable cleaning load and establish a correspondence model between resources and cleaning load. For example: the clean water storage is converted into the cleanable area based on the water consumption per unit area; the sewage tank capacity is converted into the cleaning area corresponding to the sewage capacity based on the sewage generated per unit area; the robot's endurance is converted into the workable area based on the energy consumption per unit area. Take the minimum value among the three as the initial maximum cleaning load, and then combine the base station supply efficiency and the robot's round-trip time to correct the initial maximum cleaning load to obtain the final supportable maximum cleaning load.

[0086] Step S44: Adjust the queue length and execution order to generate an executable cleaning task package, including: calculating the total load of the priority cleaning queue, which is the sum of the cleaning areas of all photovoltaic panel units in the queue; if the total load does not exceed the maximum supportable cleaning load, the priority cleaning queue is directly used as the executable cleaning task package, and the task order remains unchanged; if the total load exceeds the maximum supportable cleaning load, non-critical targets in the queue are selected, and tasks are postponed in order of dynamic cleaning urgency value from low to high. For each non-critical target postponed, the remaining queue total load is recalculated until the remaining total load matches the maximum supportable cleaning load; during the adjustment process, if the total load still exceeds the standard after postponing some non-critical targets, the maximum cleaning load is recalculated based on resource complementarity, prioritizing the complete execution of critical targets, and non-critical targets are reasonably divided according to the remaining resources, retaining only the achievable tasks; the final executable cleaning task package includes task number, photovoltaic panel unit number, cleaning urgency value, task type, estimated water consumption, estimated power consumption, estimated operation time, and execution order, which is stored in the task management module and synchronized to the cleaning robot control system.

[0087] In this embodiment, the synchronous collection of multi-dimensional resource status comprehensively covers the constraints of clean water, wastewater, and battery life, avoiding operation interruptions caused by a single resource shortage; the clear marking and priority protection of key targets ensure that high-urgency cleaning needs are not sacrificed, aligning with the actual needs of photovoltaic panel operation and maintenance; the resource-load correspondence conversion model enables the scientific calculation of the maximum cleaning load, ensuring that the adjusted total queue load is accurately matched with resource support capacity, avoiding resource waste or insufficiency; and the adjustment logic of orderly postponement of non-critical targets according to urgency ensures that resources are tilted towards high-value tasks.

[0088] In some embodiments of the present invention, in order to achieve precise matching between cleaning strategies and control parameters, this embodiment combines the feature differences of executable cleaning task packages, the adaptation rules of multi-mode cleaning strategies, and the main objectives of parameter optimization. It establishes a correlation retrieval mechanism between contamination features and cleaning strategies, utilizes physical model simulation to optimize control parameters, and introduces reward values ​​to screen efficient strategies. The specific implementation is as follows:

[0089] Step S51: Construct a pre-defined multi-mode cleaning strategy library, which is organized and classified based on historical cleaning data. The historical data includes the residual pollution characteristics, cleaning response characteristics, combined cleaning actions used, corresponding control parameters, and final performance data of past cleaning tasks. The performance data includes the cleaning effect compliance rate, water and electricity consumption, and operation time. The combined cleaning action combinations are divided into different time sequences of dry brushing, wet washing, spraying, and scraping. Each combination specifies the execution order, connection time, and applicable pollution type range. Each strategy record in the strategy library is associated with the corresponding pollution characteristic range, cleaning response characteristic threshold, combined cleaning action combination, initial control parameters, and performance data. An association index of features, strategies, and performance is established to support fast retrieval by features.

[0090] Step S52: Traverse each task in the executable cleaning task package and extract its residual contamination characteristics and cleaning response characteristics; based on these characteristics, perform a similarity search in the multi-mode cleaning strategy library, using the similarity of residual contamination area ratio, grayscale mean, distribution density, and effect attenuation coefficient as the search criteria, set a similarity threshold, and filter out historical cases with similarity exceeding the threshold; assign reward values ​​to the selected cases according to their final performance data, with reward value calculation combining cleaning effect compliance rate, water and electricity consumption saving ratio, and operation time reduction ratio, the better the performance, the higher the reward value; select the composite cleaning action combination corresponding to the case with the highest reward value as the optimal combination for the current task, if there are no similarity cases, then based on the type of contamination characteristics and cleaning response characteristics, match the most suitable basic combination from the strategy library and use it as the initial combination.

[0091] Step S53: Based on the optimal combination of composite cleaning actions, a physical model simulation system is constructed. The model includes the surface characteristics of the photovoltaic panel, the physical and mechanical model of contamination adhesion, the mechanism of cleaning actions, and the resource consumption calculation model. The residual contamination characteristics, cleaning response characteristics, and initial control parameters of the current task are input into the simulation system. A cleaning effect threshold is set, which is determined based on the photovoltaic panel cleaning standards and operation and maintenance requirements. A parameter optimization algorithm is adopted, with the cleaning effect threshold as a constraint and minimizing water and electricity consumption and operation time as the optimization objectives. The brush head rotation speed, travel speed, water spray pressure, water flow temperature, detergent ratio, and the attitude angle of the robotic arm end effector are iteratively optimized. During the iteration process, the cleaning effect and resource consumption under different parameter combinations are simulated in real time, and the optimal parameter combination that meets the constraints is recorded to form a refined control parameter set. The parameter set is classified according to the contaminated area, and different contaminated areas correspond to different parameters to ensure that the parameters are adapted to the contamination distribution and morphology.

[0092] Step S54: Input the optimized set of refined control parameters into the simulation system for secondary verification to confirm that the cleaning effect meets the threshold requirements and that water and electricity consumption and operation time are within the optimal range. If the verification fails, return to the parameter optimization stage to adjust the optimization boundary and weights, and re-iterate the calculation until the parameter set meets the requirements. The refined control parameter set after successful verification is associated with and stored with the corresponding composite cleaning action combination as the basis for the execution of the current task, and is synchronously updated to the cleaning robot control system.

[0093] In this embodiment, the construction of a multi-mode cleaning strategy library and a similarity retrieval mechanism ensure accurate matching between the combination of composite cleaning actions and the residual pollution characteristics and cleaning response characteristics of the task; the reward-driven strategy selection logic selects the optimal strategy based on historical performance data, ensuring the comprehensive advantages of the selected strategy in terms of cleaning effect, resource consumption, and operation time; the application of physical model simulation and parameter optimization algorithm provides a scientific basis for determining the refined control parameter set, achieving optimization of water and electricity consumption and operation time while meeting the requirements of cleaning effect; and the setting of parameters according to the differences in pollution areas allows the control parameters to adapt to the pollution status of different areas, further improving the uniformity of cleaning effect and the rationality of resource utilization.

[0094] In a specific implementation, as one example, the cleaning robot is driven to execute an executable cleaning task package based on a refined set of control parameters, as follows:

[0095] Step S61: The verified set of refined control parameters and the corresponding executable cleaning task package are synchronously transmitted to the cleaning robot control system via industrial Ethernet. Data verification algorithms are used during the transmission process to ensure data integrity. After receiving the data, the robot control system calibrates the set of refined control parameters. For key parameters such as brush head rotation speed, travel speed, and water spray pressure, the initial readings of the robot's built-in sensors are compared with the values ​​to calculate the deviation. If the deviation exceeds the preset range, the parameter reference of the robot's execution module is automatically adjusted to ensure that the execution parameters are consistent with the refined control parameters. After calibration, a parameter calibration confirmation signal is generated and fed back to the task management module.

[0096] Step S62: The robot control system analyzes the executable cleaning task package, extracts the task execution order, the location of the photovoltaic panel unit corresponding to each task, the estimated operation time and task type, and establishes a task execution queue; the tasks are started in sequence according to the queue order, and the next task can only be started after the previous task is completed and confirmed, and cross-order execution is prohibited; for key target tasks, they are marked separately and priority is given to ensure execution resources, ensuring that they are executed completely according to the preset parameters and order, and pausing or interruption is prohibited during the execution process.

[0097] Step S63: Based on the combination of complex cleaning actions and timing requirements corresponding to the current task, the robot drives each execution module to perform actions according to the refined control parameter set. The brush head rotation speed, travel speed, water spray pressure, water flow temperature, detergent ratio, and the attitude angle of the robotic arm end effector strictly follow the differentiated settings of the parameter set. For different contaminated areas, the corresponding parameters are switched in real time. During the action connection process, a connection timing threshold is set to control the interval between the end of the previous action and the start of the next action, ensuring smooth action connection and avoiding resource waste or equipment damage caused by action superposition.

[0098] In this embodiment, the parameter synchronization and calibration mechanism ensures that the robot's execution parameters are consistent with the refined control parameters, avoiding poor cleaning results or equipment damage caused by parameter deviations. Strict control over the execution sequence and priority assurance of critical tasks ensure that executable cleaning task packages are executed completely in a preset order, aligning with task priority settings and guaranteeing cleaning results for high-urgency tasks.

[0099] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for optimizing photovoltaic panel cleaning strategies, characterized in that, include: In response to a cleaning command or pollution warning, the cleaning robot is controlled to perform an initial cleaning operation on the target photovoltaic array along a preset track based on preset reference parameters, and the initial operation spectrum data is collected; the initial operation spectrum data includes at least the pollution image before operation, the operation trajectory, real-time cleaning parameters, and the surface state image after operation; Feature extraction was performed on the initial operation spectrum data to obtain the residual pollution characteristics, cleaning response characteristics, and environmental corrosion factors of each photovoltaic panel unit; The extraction of the cleaning response features includes: comparing and analyzing the contamination images of the same photovoltaic panel unit before and after the operation, identifying stubborn contamination areas and their morphology that were not effectively removed under preset benchmark parameters; analyzing the spatiotemporal relationship between the robot's operation trajectory and the stubborn contamination areas, and generating an effect attenuation coefficient; the environmental corrosion factors include the potential corrosivity level of the contamination type on the panel surface; the contamination types include dust, grease, biological excrement, and chemical residues; Based on the residual pollution characteristics, cleaning response characteristics, and environmental corrosion factors, the cleaning urgency of each photovoltaic panel unit is assessed; and all photovoltaic panel units are ranked according to the cleaning urgency to construct a priority cleaning queue; the assessment of the cleaning urgency of each photovoltaic panel unit includes: assigning preset assessment weights to the residual pollution characteristics, cleaning response characteristics, and environmental corrosion factors; quantifying the indicators of each characteristic into quantitative values ​​within a unified numerical range; and using a weighted summation algorithm to calculate the cleaning urgency value of each photovoltaic panel unit; Obtain the real-time available resource status of the cleaning system, adapt it to the priority cleaning queue, adjust the queue length and execution order, and generate an executable cleaning task package. For each task in the executable cleaning task package, based on its corresponding residual contamination characteristics and cleaning response characteristics, the optimal combination of composite cleaning actions is matched from the preset multi-mode cleaning strategy library, and the corresponding set of refined control parameters is optimized and determined. The cleaning robot is driven to execute the executable cleaning task package according to the refined control parameter set.

2. The photovoltaic panel cleaning strategy optimization method according to claim 1, characterized in that, The resource status includes the base station's clean water storage, the remaining sewage tank volume, and the robot's battery life. Adapting the real-time available resource status to the priority cleaning queue, adjusting the queue length and execution order, includes: Photovoltaic panel units whose dynamic cleaning urgency value exceeds the first threshold in the priority cleaning queue are marked as critical targets, ensuring that they are included in the executable cleaning task package; Calculate the maximum cleaning load that can be supported based on the real-time available resource status; If the total load of the priority cleaning queue exceeds the maximum cleaning load, tasks from non-critical targets will be temporarily suspended from low to high based on the dynamic cleaning urgency value, until the total load of the priority cleaning queue matches the maximum cleaning load.

3. The photovoltaic panel cleaning strategy optimization method according to claim 2, characterized in that, The combined cleaning action sequence includes the timing arrangement of at least two actions among dry brushing, wet washing, spraying, and scraping; The refined control parameter set includes the brush head rotation speed, travel speed, water spray pressure, water flow temperature, detergent ratio, and the attitude angle of the robotic arm end effector for different contaminated areas. The process of optimizing and determining the set of refined control parameters adopts a parameter optimization algorithm based on physical model simulation, with the optimization objective being to minimize water and electricity consumption and operation time while meeting the cleaning effect threshold.

4. The photovoltaic panel cleaning strategy optimization method according to claim 3, characterized in that, The construction of the multi-mode cleaning strategy library is based on historical data. When matching the optimal combination of composite cleaning actions, cases with similar pollution characteristics and cleaning response characteristics in the historical data are first retrieved from the multi-mode cleaning strategy library. Then, a reward value is assigned based on the final performance data achieved by the strategy adopted in the case, thereby driving strategy selection.

5. The photovoltaic panel cleaning strategy optimization method according to claim 1, characterized in that, The quantization value of the cleaning response characteristic is directly adopted using the effect attenuation coefficient; The quantitative value of the environmental corrosion factor is obtained by mapping according to the preset corrosion level corresponding to the pollution type.

6. The photovoltaic panel cleaning strategy optimization method according to claim 1, characterized in that, Constructing the priority cleaning queue includes: The photovoltaic panel units are sorted from highest to lowest cleaning urgency value to form an initial queue; If multiple photovoltaic panel units have the same cleaning urgency value, they are then sorted by the corrosivity level of the environmental corrosion factors, with higher corrosion levels appearing first.

7. The photovoltaic panel cleaning strategy optimization method according to claim 1, characterized in that, The preset benchmark parameters are stored in a parameter library according to the photovoltaic panel material and installation scenario; After the initial cleaning operation is triggered, the cleaning robot obtains the corresponding baseline parameters from the edge computing node and performs a pre-start verification of the actuator. After confirming that the parameter error is within the preset range, the operation is started.

Images