Water consumption control method and system for photovoltaic panel cleaning robot

Abstract

The present application relates to photovoltaic power generation operation and maintenance technical field, especially in kind of photovoltaic panel cleaning robot water consumption control method and system, including synchronous acquisition current to be cleaned photovoltaic panel area's pollution characteristic image, real-time environmental parameter and robot body operating parameter;Pollution characteristic image is carried out feature recognition, obtains the pollution type and pollution grade of to be cleaned photovoltaic panel area;Real-time environmental parameter is carried out evaporation evaluation, obtains the evaporation intensity index of to be cleaned photovoltaic panel area;With pollution type parameter as index dimension, in the preset water consumption benchmark mapping table, index is determined basic unit water consumption;The basic unit water consumption, pollution grade and robot body operating parameter are input to the preset real-time cleaning optimization model, obtains segmented cleaning control parameter instruction, and in accordance with this control cleaning robot executes cleaning operation.Through the present application, can be based on real-time pollution characteristic and environmental parameter, carries out water consumption dynamic calculation and multi-objective optimization control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power generation operation and maintenance, and particularly relates to a water consumption control method and system for a photovoltaic panel cleaning robot. BACKGROUND

[0002] Photovoltaic panel cleaning is an important operation and maintenance link to ensure its power generation efficiency. At present, automatic cleaning robots have been widely used in photovoltaic power generation operation and maintenance, and their water consumption control is mostly dependent on preset fixed programs or control logic based on simple feedback, such as timing or stroke. However, the pollution condition of the surface of the photovoltaic panel and the on-site environmental conditions are complex and changeable. The existing control method is difficult to calculate the optimal cleaning parameters that can minimize water consumption under the premise of ensuring cleaning cleanliness according to the real-time changes of multi-dimensional factors, resulting in the use of redundant water strategy in actual operation to ensure cleaning effect, and the optimal balance between cleaning effect and water resource consumption cannot be achieved under changing working conditions.

[0003] Therefore, there is an urgent need for a method that can dynamically calculate and multi-objective optimization control water consumption based on real-time pollution characteristics and environmental parameters to solve the technical problem of accurately controlling and minimizing cleaning water consumption under the premise of ensuring the preset cleaning standard. SUMMARY

[0004] The present application provides a photovoltaic panel cleaning robot water consumption control method and system, which can effectively solve the problems in the background art.

[0005] In order to achieve the above purpose, the technical solution adopted by the present application is:

[0006] A photovoltaic panel cleaning robot water consumption control method, comprising:

[0007] synchronously collecting pollution characteristic images of a current photovoltaic panel area to be cleaned, real-time environmental parameters and robot body operation parameters;

[0008] performing feature recognition on the pollution characteristic images to obtain the pollution type and pollution level of the photovoltaic panel area to be cleaned;

[0009] performing evaporation evaluation on the real-time environmental parameters to obtain the evaporation intensity index of the photovoltaic panel area to be cleaned;

[0010] using the pollution type, pollution level and evaporation intensity index as index dimensions, indexing in a pre-set water consumption benchmark mapping table to determine the basic unit water consumption for the current photovoltaic panel area to be cleaned;

[0011] The base unit water consumption, the pollution level and the robot body operation parameters are input into a preset real-time cleaning optimization model to obtain segmented cleaning control parameter instructions, and the cleaning robot is controlled to perform cleaning work according to the segmented cleaning control parameter instructions.

[0012] Further, the robot body operation parameters include a current walking speed, a current brush disc pressure and a current water pump pressure.

[0013] The real-time environmental parameters include a photovoltaic panel surface temperature, an illumination intensity, an environmental temperature and a wind speed.

[0014] Further, the pollution types include dust uniform coverage, sand dust accumulation and bird droplet pollution.

[0015] Further, the pollution feature image is subjected to feature recognition to obtain a pollution type and a pollution level of a photovoltaic panel area to be cleaned, including:

[0016] The pollution feature image is analyzed by using a pre-trained convolutional neural network model to output a classification result of the pollution type and generate an image mask for identifying a pollution area.

[0017] The pollution level is comprehensively determined according to a pixel proportion covered by the image mask, an average gray value in the area and a texture feature parameter.

[0018] The pollution level is quantified into multiple discrete levels, and the pollution level is higher when the pollution coverage proportion is higher, the gray value difference from the background is larger or the texture is coarser.

[0019] Further, the construction method of the water consumption benchmark mapping table includes:

[0020] In a standard test environment, for each combination of the pollution type and the pollution level and different intervals of the evaporation intensity index, the minimum unit water consumption is determined by an orthogonal experiment method when a preset base cleanliness is guaranteed.

[0021] The pollution type, the pollution level and the evaporation intensity index interval are taken as input features, and the determined minimum unit water consumption is taken as an output label, and a regression model is trained, and an output result of the regression model is discretized to form the water consumption benchmark mapping table.

[0022] Further, the segmented cleaning control parameter instructions include a segmented cleaning path walking speed, a segmented water flow and a segmented brush head pressure.

[0023] Further, the real-time cleaning optimization model is a nonlinear programming model with constraints, and an objective function thereof is expressed as minimization of total water consumption, and the constraint conditions include:

[0024] The cleanliness prediction value of each cleaning segment is greater than or equal to a preset threshold, and the cleanliness prediction value is calculated by a preset prediction sub-model based on the pollution level, water flow, brush head pressure and travel speed of the segment.

[0025] Further, the preset prediction sub-model is a machine learning model trained based on historical cleaning data, and the input is the pollution level, planned water flow, brush head pressure and travel speed of the cleaning segment, and the output is the cleanliness prediction value of the cleaning segment.

[0026] The cleanliness prediction value is positively correlated with the water flow and the brush head pressure, and is negatively correlated with the travel speed and the pollution level, and the negative impact of the pollution level on the cleaning effect shows an exponential decay trend.

[0027] Further, the real-time cleaning optimization model adopts a particle swarm optimization algorithm.

[0028] In the iterative solution process of the particle swarm optimization algorithm, the search step of the particle swarm optimization algorithm is adjusted according to the gap between the cleanliness prediction value corresponding to the current candidate solution and the preset threshold.

[0029] When the cleanliness prediction value approaches the preset threshold, the search step is reduced for fine search.

[0030] On the other hand, the application also provides a water consumption control system for a photovoltaic panel cleaning robot, comprising:

[0031] A data acquisition module is configured to synchronously acquire pollution feature images, real-time environmental parameters and robot body operation parameters of a current photovoltaic panel area to be cleaned.

[0032] A pollution feature recognition module is configured to perform feature recognition on the pollution feature images to obtain a pollution type and a pollution level of the photovoltaic panel area to be cleaned.

[0033] An environmental parameter processing module is configured to perform evaporation evaluation on the real-time environmental parameters to obtain an evaporation intensity index of the photovoltaic panel area to be cleaned.

[0034] A reference water consumption determination module is configured to index the pollution type, the pollution level and the evaporation intensity index as index dimensions in a preconfigured water consumption reference mapping table to determine a basic unit water consumption for the current photovoltaic panel area to be cleaned.

[0035] An optimization control module is configured to input the basic unit water consumption, the pollution level and the robot body operation parameters into a preset real-time cleaning optimization model to obtain segment cleaning control parameter instructions, and control the cleaning robot to perform cleaning work according to the segment cleaning control parameter instructions.

[0036] The technical scheme of the present application can realize the following technical effects:

[0037] The present application couples the feature recognition result of the pollution type and the pollution level with the environmental evaporation intensity evaluation result to serve as a composite index for searching the water use benchmark mapping table, ensures that the determined basic unit water consumption reflects the inherent difficulty of pollutant removal and the water loss risk caused by the environment, makes the water benchmark value fit the actual physical process, and solves the adaptability problem caused by setting the water consumption according to the pollution degree or fixed experience value;

[0038] The basic unit water consumption and the real-time collected robot body running parameters are input into the real-time cleaning optimization model, the water consumption minimization is taken as the core target, the cleaning degree prediction value is taken as the key constraint, the benchmark water consumption is used to provide the optimization starting point and boundary that meet the current working condition, the response state of the cleaning mechanism is compensated and adjusted according to the real-time running parameters, and the segmented cleaning control parameter instruction that can adapt to the differentiated pollution and evaporation state of each region under the global water consumption constraint is solved;

[0039] The feature recognition and evaporation evaluation steps provide initial decision basis for water consumption calculation; the water use benchmark mapping table step efficiently converts multi-source features into operable water use benchmarks based on prior knowledge; and the real-time cleaning optimization model step integrates the initial decision, real-time state and multi-objective constraint to generate execution instructions, so that the cleaning robot can autonomously maintain the optimal balance between the preset cleaning standard and water resource consumption efficiency under complex and variable working conditions.

[0040] The above description is only a summary of the technical scheme of the present application, in order to more clearly understand the technical means of the present application, the specific embodiments of the present application can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical scheme in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments described in the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0042] Figure 1 The flowchart of the water consumption control method for the photovoltaic panel cleaning robot in the embodiments of the present application is shown in the figure.

[0043] Figure 2 The logic block diagram of the water consumption control system for the photovoltaic panel cleaning robot in the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application.

[0045] 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 application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in this description, the singular forms "a", "an" and "the" include plural references unless the context clearly dictates otherwise.

[0046] As shown in Figure 1 A water consumption control method for a photovoltaic panel cleaning robot according to the present application specifically comprises the following steps:

[0047] Step S1, synchronously collecting a pollution feature image of a current photovoltaic panel area to be cleaned, real-time environment parameters and robot body operation parameters;

[0048] Step S2, performing feature recognition on the pollution feature image to obtain a pollution type and a pollution level of the photovoltaic panel area to be cleaned;

[0049] Step S3, performing evaporation evaluation on the real-time environment parameters to obtain an evaporation intensity index of the photovoltaic panel area to be cleaned;

[0050] Step S4, indexing in a preset water consumption benchmark mapping table with the pollution type, the pollution level and the evaporation intensity index as index dimensions to determine a basic unit water consumption for the current photovoltaic panel area to be cleaned;

[0051] Step S5, inputting the basic unit water consumption, the pollution level and the robot body operation parameters into a preset real-time cleaning optimization model to obtain segmented cleaning control parameter instructions, and controlling the cleaning robot to perform cleaning work according to the segmented cleaning control parameter instructions.

[0052] In the embodiment, by coupling the feature recognition result of the pollution type and the pollution level with the environmental evaporation intensity evaluation result as a composite index for retrieving the water benchmark mapping table, it is ensured that the determined basic unit water consumption reflects the inherent difficulty of pollution removal and the water loss risk caused by the environment, so that the water benchmark value is in line with the actual physical process, and the adaptability problem caused by setting the water consumption according to the pollution level or fixed experience value is solved; by inputting the basic unit water consumption and the real-time collected robot body running parameters into the real-time cleaning optimization model, taking water consumption minimization as the core target and taking the cleanliness prediction value as the key constraint, the benchmark water consumption is used to provide an optimization starting point and boundary that meets the current working condition, and the response state of the cleaning mechanism is compensated and adjusted according to the real-time running parameters, so that the segmented cleaning control parameter instruction that can adapt to the differentiated pollution and evaporation state of each region under the global water consumption constraint is solved; the feature recognition and evaporation evaluation steps provide initial decision basis for water consumption calculation; the water benchmark mapping table step efficiently converts multi-source features into operable water benchmarks based on prior knowledge; and the real-time cleaning optimization model step integrates the initial decision, real-time state and multi-objective constraint to generate an execution instruction, so that the cleaning robot can autonomously maintain the optimal balance between the preset cleaning standard and water resource consumption efficiency under complex and variable working conditions.

[0053] In some embodiments of the application, the robot body running parameters include the current walking speed, the current brush disc pressure and the current water pump pressure; the real-time environmental parameters include the photovoltaic panel surface temperature, the light intensity, the environmental temperature and the wind speed; to realize data collection of the current to-be-cleaned photovoltaic panel area, the following steps are specifically included:

[0054] Step S11, a high-definition industrial camera is fixed at the central position of the front end beam of the cleaning robot, the lens is vertically directed to the photovoltaic panel surface, the focal length is adapted to the working height of the robot, and it is ensured that the shooting range covers a single area of the current cleaning path; integrated sensing units are symmetrically arranged on both sides of the camera, a photovoltaic panel surface temperature sensor, an environmental temperature sensor, a light intensity sensor and a micro wind speed sensor are built-in each unit, the sensing unit probe is exposed and maintains a fixed distance from the photovoltaic panel surface, so as to avoid contact with the panel body or pollutants during cleaning operation; the walking motor, the brush disc driving motor and the water pump of the robot are respectively configured with an incremental encoder, a pressure sensor and a pressure transmitter, and the running parameters are collected in real time;

[0055] Step S12, the photovoltaic panel to be cleaned is divided into continuous rectangular collection sections according to the single cleaning width and length of the robot in advance, and the boundary coordinates of each section are stored in the robot control system; when the robot travels to the starting position of a collection section, the control system triggers a synchronous collection signal, which is sent to the high-definition industrial camera, the integrated sensing unit and the detection elements of each motor / pump body at the same time, triggering the camera to shoot the pollution feature image, the sensing unit to detect the environmental parameters and the detection elements to collect the operating parameters;

[0056] Step S13, when the collection signal is triggered, the control system synchronously records the positioning coordinates of the current robot, which are provided by the robot self-positioning module and used as the area identifier of the current collection section; the pollution feature image shot by the high-definition industrial camera is embedded with the positioning coordinates and a collection timestamp, the environmental parameters detected by the integrated sensing unit and the operating parameters collected by the detection elements are all associated with the same positioning coordinates and timestamp, and all the data are stored in the robot local storage unit as indexed by the positioning coordinates and the timestamp, forming a multi-source parameter set of a single collection section.

[0057] In this embodiment, the design of the preset collection section, the synchronous trigger signal and the coordinate association index solves the problem of mismatch of multi-source data collection in time and space; the synchronous triggering ensures that the pollution image, the environmental parameters and the operating parameters correspond to the same physical time, eliminating the timing error caused by the response delay of the sensor or the difference in sampling period; the strict association of the collected data with the accurate real-time positioning coordinates of the robot avoids the incorrect combination of perception information at different spatial positions; all the data are stored with a unified spatiotemporal coordinate index, constructing a single collection section multi-source parameter set with strictly consistent spatiotemporal attributes, and improving the accuracy of pollution assessment, evaporation calculation and control parameter optimization.

[0058] In some embodiments of the present application, the pollution feature image is analyzed by using a pre-trained convolutional neural network model, a classification result of the pollution type is output, and an image mask identifying the pollution area is generated, wherein the pollution type includes uniform dust coverage, sand dust accumulation and bird droplet pollution; the pollution level is comprehensively determined according to the pixel proportion covered by the image mask, the average gray value in the area and the texture feature parameter; wherein the pollution level is quantized into multiple discrete levels, the higher the pollution coverage proportion, the greater the difference between the gray value and the background or the rougher the texture, and the higher the determined pollution level. The specific implementation is as follows:

[0059] Step S21, a pre-trained convolutional neural network model is constructed, specifically including: collecting image samples of dust uniform coverage, sand accumulation, bird droplet pollution and clean photovoltaic panels, the samples need to cover the surface state of photovoltaic panels under different light intensities and environmental temperatures; double-label labeling is performed on each sample image, one is pollution type label, including dust uniform coverage, sand accumulation, bird droplet pollution and cleanliness, and the other is pollution area boundary coordinate label marked in the form of pixel point coordinate group;

[0060] A convolutional neural network model is constructed based on the ResNet50 architecture, and a pollution-specific feature extraction layer is added between the convolutional layer and the fully connected layer of the model; special convolution kernels are set for the features of various pollution types, for example, the texture of dust uniform coverage is continuously distributed, 1x5 horizontal convolution kernel and 5x1 vertical convolution kernel are configured; the edge profile of sand accumulation is irregular, 3x3 edge detection convolution kernel is configured; bird droplet pollution is in the form of isolated block, 5x5 point feature convolution kernel is configured;

[0061] Model training uses cross-validation, the labeled samples are divided into training set, validation set and test set, Adam optimizer is selected, the initial learning rate is set to 0.001, when the validation set accuracy rate does not improve for 5 consecutive epochs, the learning rate is decayed to 1 / 10 of the original; after training, the pollution feature image obtained in step S1 is input, the model outputs two results: one is the pollution type classification result, and the other is the binary image mask, that is, the pixel value of the pollution area is set to 1, and the pixel value of the clean background area is set to 0.

[0062] Step S22, performing morphological closing operation optimization on the binary image mask output by the model; specifically, a 3x3 rectangular structural element is used to fill the small holes in the pollution area to dilate the mask, and then erosion operation is performed to eliminate isolated noise points on the edge of the mask, ensuring that the pollution area profile is complete and free of redundant noise; by traversing the optimized image mask pixel by pixel, the total number of pollution pixels with pixel value of 1 is counted, and the ratio of the total number of pollution pixels to the total number of image pixels is calculated to obtain the pollution coverage ratio.

[0063] Step S23, based on the optimized image mask, a pollution region sub-image is segmented from the original pollution feature image, only the original image region corresponding to the mask pixel value of 1 is retained, and the sub-image is converted into an 8-bit grayscale image; a grayscale image of a clean photovoltaic panel of the same model under standard light conditions is collected, and the average grayscale value thereof is calculated as a standard grayscale value; the average grayscale value of the pollution region sub-image is calculated, and the absolute difference with the standard grayscale value is calculated to obtain a grayscale difference value for reflecting pollution thickness related information; a texture feature parameter of the pollution region sub-image is calculated by using a gray level co-occurrence matrix method, specifically, the distance parameter of the gray level co-occurrence matrix is set to 1 pixel, the angle parameter is 0°, 45°, 90° and 135°, and the gray level is 256 levels, and the entropy value for reflecting the complexity of the texture, the contrast for reflecting the light and dark difference of the texture, and the energy for reflecting the uniformity of the texture are calculated under the four angles respectively, and the average value of each parameter under the four angles is taken as the final texture feature parameter.

[0064] Step S24, a plurality of discrete pollution levels are preset, and the threshold value interval corresponding to each level is determined by orthogonal experiment; under the standard test environment, the minimum water consumption required for cleaning from clean to the preset standard is determined for different combinations of pollution coverage ratio, grayscale difference value and texture feature parameter, the cleaning difficulty level is divided according to the water consumption, and then the parameter threshold value interval corresponding to each level is determined reversely; the comprehensive score is calculated by using the weighted summation method, and the weight distribution is that the pollution coverage ratio accounts for the highest proportion, and the grayscale difference value and the texture feature comprehensive value account for the same proportion; the calculation method of the texture feature comprehensive value is that the entropy value, the contrast and the energy are normalized to the same interval respectively, and the average value of the sum of the three is taken; the pollution level is determined according to the threshold value interval into which the comprehensive score falls, the higher the score, the higher the pollution level, and the greater the cleaning difficulty.

[0065] In the embodiment, the convolutional neural network model is additionally provided with a pollution exclusive feature extraction layer and a special convolution kernel, which specifically identifies the shape and texture difference of the three types of pollution, reduces the classification confusion of the general model, and improves the matching degree of the classification result and the actual pollution type; after the image mask is optimized by morphological closing operation, the pollution region and the background are accurately isolated, the pollution coverage ratio, the grayscale difference value and the texture feature parameter are all calculated based on the pollution region, and the interference of clean background pixels is excluded; the pollution level is calculated by weighting the pollution coverage ratio, the grayscale difference value and the texture feature parameter, a single parameter cannot completely represent the influence of pollution on cleaning, and multi-dimensional parameter integration can fully match the actual cleaning demand.

[0066] In a specific implementation, as an embodiment, the evaporation of real-time environmental parameters is evaluated to obtain an evaporation intensity index of the photovoltaic panel area to be cleaned. The evaporation rate of water on the surface of the photovoltaic panel is a key environmental interference factor that determines the effectiveness of the cleaning water. The temperature rise of the photovoltaic panel surface under illumination, the specific surface roughness and the influence of the inclination angle on the airflow near the wall, and the change of the pollution layer to the surface thermodynamic properties all make the evaporation process significantly different from open water or general wet surface. If the evaluation model fails to include the above key influencing factors, the evaporation intensity index obtained cannot accurately reflect the real retention risk of the cleaning liquid on the actual working surface, resulting in inaccurate water control. To solve the above problems, the embodiment constructs a structured calculation framework including driving items, transmission items and surface state correction items. Based on the classical evaporation theory, the framework introduces special parameters and correction functions for photovoltaic application scenarios, converts the collected multi-point environmental parameters into a single quantitative index representing the current instantaneous evaporation potential, and is implemented as follows:

[0067] Step S31, instantaneous measurement values of the photovoltaic panel surface temperature, environmental temperature, wind speed and illumination intensity are extracted from the synchronously collected data frames. At the same time, the inherent parameters related to the photovoltaic panel model and installation location are called from the pre-stored database, including the absorption rate of the photovoltaic panel surface to the solar radiation, the installation inclination angle of the photovoltaic panel, and the aerodynamic roughness parameter obtained by wind tunnel experiment or field calibration for the airflow characteristics on the upper surface of the photovoltaic panel array.

[0068] Step S32, the evaporation core thermodynamic driving potential is calculated. The driving potential is represented by the difference in saturated water vapor pressure. Specifically, the corresponding saturated water vapor pressure values are calculated by the saturated water vapor pressure calculation formula according to the photovoltaic panel surface temperature measurement value and the environmental temperature measurement value respectively. The difference between the two saturated water vapor pressure values is the core thermodynamic driving potential for the evaporation of water from the panel surface to the air.

[0069] Step S33, the net radiation energy input value for water vaporization is calculated. This calculation mainly considers the solar radiation energy. First, the solar incident angle is calculated according to the installation inclination angle of the photovoltaic panel, the current geographical location and the time. The measured illumination intensity is multiplied by the panel surface radiation absorption rate, and then multiplied by the cosine value of the incident angle to obtain the absorbed shortwave radiation energy on the panel surface. Then, based on the Stefan-Boltzmann law, the net exchange value of longwave radiation between the panel surface and the atmosphere is calculated according to the surface temperature and the environmental temperature. The net radiation energy input value acting on the panel surface and the water film is obtained by subtracting the longwave radiation loss from the absorbed shortwave radiation energy.

[0070] Step S34, calculating the water vapor mass transfer efficiency under the action of wind speed; the calculation introduces a wind speed correction function based on the logarithmic law; the wind speed correction function takes the measured wind speed as the input, combines the specific aerodynamic roughness parameter of the photovoltaic panel, and calculates the transfer coefficient representing the diffusion ability of water vapor from the panel surface to the atmosphere, to ensure that the air retention effect is compensated for at low wind speed, and the growth of the transfer efficiency is smoothly limited in accordance with the fluid boundary layer law at high wind speed.

[0071] Step S35, merging the above-mentioned net radiation energy input value and the water vapor mass transfer efficiency value according to the physical weight of their respective contributions to evaporation, to obtain a preliminary evaporation rate estimation value; the merging process follows the principles of energy and mass conservation, and the weight coefficient is determined by the local slope of the temperature and saturation vapor pressure curve and the dry and wet table constant; then, a dynamic correction factor is introduced, which is obtained from a pre-set query table according to the current regional pollution level determined in real time in step S2, and the query table is established through experiments, reflecting the comprehensive influence of different pollution types and thicknesses on surface temperature, hydrophilicity and thermal insulation, so as to dynamically adjust the evaporation rate estimation value and obtain the corrected evaporation rate.

[0072] Step S36, dividing the corrected evaporation rate by a standard evaporation rate reference value calculated by the model under predefined standard environmental conditions, including standard temperature, standard humidity, standard wind speed, standard illumination and clean surface state; the division operation produces a dimensionless ratio, which is the final evaporation intensity index; the evaporation intensity index greater than 1 indicates that the current environmental evaporation capacity is higher than the standard condition, and less than 1 indicates that it is lower than the standard condition.

[0073] In the embodiment, by establishing a structured physical hybrid model, the thermodynamic, radiation and aerodynamic factors affecting the evaporation process are integrated; by introducing special parameters for the surface characteristics of the photovoltaic panel and the wind speed correction function based on the logarithmic law, the model is closer to the real microclimate conditions of the photovoltaic panel surface; by incorporating the dynamic correction factor based on the real-time pollution level, the evaporation evaluation can respond to the dynamic changes of the surface properties during the cleaning operation; finally, the standardized evaporation intensity index is output, so that the water consumption control strategy can be adaptively adjusted according to the quantitative evaluation of the environmental evaporation potential.

[0074] In some embodiments of the present application, the basic unit water consumption for the current photovoltaic panel area to be cleaned is determined by offline construction of a water benchmark knowledge base and real-time operation-oriented water consumption determination logic;

[0075] Specifically, the construction of the water benchmark knowledge base is completed through laboratory calibration of the system, as follows:

[0076] In a standard laboratory, using the same type of photovoltaic panel samples as the target to be cleaned, three types of standard pollution samples are prepared: dust evenly covered, sand accumulation, and bird droplet pollution; Each type of pollution is divided into multiple discrete pollution levels according to its coverage density, particle size or attachment area, and the division standard and step S2 level determination logic remain the same;

[0077] Place the pollution samples in an experimental chamber that can accurately control the environmental conditions; The experimental chamber can independently adjust the temperature, humidity, wind speed and light intensity; By adjusting the above parameters, multiple discrete evaporation intensity levels from very low evaporation to very strong evaporation are simulated, which correspond to the evaporation intensity index intervals output by step S3;

[0078] For each set of determined parameter combinations (pollution type i, pollution level j, evaporation intensity level k), perform a cleaning orthogonal experiment; Use a high-precision flow meter and nozzle to clean the sample with a specific amount of water per unit area; After cleaning, use standard instruments to measure the surface cleanliness; By iteratively adjusting the amount of water per unit area, find the minimum amount of water that can first reach the preset qualified threshold; This minimum amount of water is the baseline water consumption under the given conditions;

[0079] Collect the minimum water consumption data for all (pollution type i, pollution level j, evaporation intensity level k) combinations to form a three-dimensional baseline database; Use this three-dimensional baseline database as a training set to train a gradient boosting decision tree regression model with pollution type, pollution level, and evaporation intensity index as input features and baseline water consumption as output target, to learn the complex nonlinear mapping relationship and interaction effects between the three input dimensions and water consumption; Finally, perform dense grid reasoning on the entire input space (3 types of pollution, multiple pollution levels, and multiple evaporation intensity indicators) using the model, and archive the predicted water consumption for each grid point to generate a high-resolution preset water consumption baseline mapping table for quick queries.

[0080] Further, in real-time operations, the basic unit water consumption is queried and determined, when the pollution type, pollution level and evaporation intensity index of the current area are obtained, first try to perform an accurate matching query in the preset water consumption baseline mapping table; The query key value is (pollution type, discrete level to which the pollution level belongs, discrete interval to which the evaporation intensity index belongs);

[0081] If the accurate match is successful, directly read the corresponding water consumption as the basic unit water consumption;

[0082] If the query key value is between the discrete grid of the mapping table, that is, the transition working condition is not pre-stored, the system calls the regression model for real-time inference; the continuous pollution type value and pollution level value are input into the regression model together with the evaporation intensity index value, the model outputs the continuous basic unit water consumption prediction value, and the prediction value is used as the basic unit water consumption under the current working condition.

[0083] The embodiment overcomes the problem of insufficient precision of the traditional discrete look-up table method when the parameters fall between the grid by using the regression model to construct a continuous mapping relationship and combining real-time interpolation inference, realizes continuous and smooth water consumption benchmark output in the full parameter range through model calculation, and improves the fineness of water consumption setting; the regression model is used as the bottom logic, so that the system can accurately fit and interpolate the non-linear and interactive effects of the continuously changing input features, and the output basic unit water consumption is more suitable for the actual demand under complex working conditions than the method based on simple linear interpolation or nearest neighbor query.

[0084] In some embodiments of the application, in order to achieve the purpose of minimizing total water consumption and meeting the segmented cleanliness, the embodiment realizes segmented control by combining the robot body running parameters of step S1, the segmented pollution level of step S2, the basic unit water consumption of step S4, and the collection segment division logic of step S1, and the specific implementation is as follows:

[0085] Step S51, constructing a cleanliness prediction sub-model; specifically, a training sample set composed of historical data is collected, including the measured cleanliness data under different pollution levels, water flow, brush head pressure, and travel speed combinations in the orthogonal experiment of step S4, and the recorded segmented cleaning parameters and corresponding cleanliness detection data in actual operation, and the samples cover three types of pollution and full discrete pollution levels; the input features of the samples are defined as segmented pollution level, planned water flow, brush head pressure, and travel speed, and the output feature is the measured cleanliness value;

[0086] The pollution level feature is quantitatively preprocessed, the exponential decay relationship between the pollution level and the cleanliness is fitted based on the above experimental data, and the exponential coefficient is determined, the pollution level feature value is converted through the coefficient, so that the model adapts to the exponential decay law of the pollution level on the cleaning effect; the gradient boosting regression algorithm is selected to construct the prediction sub-model, according to the identification ability of the algorithm to the interaction effect between the features, the non-linear relationship between the cleanliness and the water flow, the brush head pressure, the travel speed, and the pollution level is fitted; the samples are divided into training set, validation set, and test set in proportion, the model hyperparameters are optimized by cross-validation, the test set prediction error is controlled in a reasonable range, and the fitting goodness meets the prediction demand.

[0087] Step S52, a nonlinear programming optimization model with constraints is constructed; specifically, the model takes the minimization of total water consumption as the objective function, and the total water consumption calculation logic is closely related to the subsequent segment control parameters and the segment division criteria in the foregoing: the total water consumption is the sum of the product of the water flow rate of each segment and the corresponding segment cleaning time, the segment cleaning time is calculated from the segment area and the travel speed, and the segment area follows the segment division criteria collected in step S1, so that the calculation reference is consistent with the segment logic in the foregoing;

[0088] The constraint conditions of the nonlinear programming optimization model include:

[0089] The cleanliness constraint, the predicted value of the cleanliness of each segment is output by the aforementioned prediction sub-model, and needs to be not less than a preset qualified threshold, which is unified with the cleanliness standard threshold in the experiment in step S4;

[0090] The device operation safety constraint, the brush head pressure does not exceed the rated pressure of the robot brush disc, the travel speed is in the safe operation interval designed by the robot, and the water flow rate does not exceed the rated output of the water pump, the constraint range is determined in combination with the robot operation parameters collected in step S1, to avoid overloading and damage of the actuator;

[0091] The parameter linkage constraint, the water flow rate and the brush head pressure need to satisfy the adaptive relationship, the water flow rate is adjusted in proportion when the brush head pressure is increased, to prevent the plate body from being worn or not being cleaned completely due to insufficient water flow rate caused by excessively high pressure.

[0092] Step S53, a particle swarm optimization algorithm is used to solve the above-mentioned optimization model with constraints: the particle coding corresponds to the control parameter dimension of each segment, each particle contains three parameters of the travel speed, the water flow rate and the brush head pressure of each cleaning segment, the particle dimension is consistent with the number of subsequent cleaning segments, to ensure that the solution result can directly adapt to the segment control requirement; when initializing the particle swarm, the initial range of the particle is set based on the current operation parameters of the robot collected in step S1, and the initial range of the water flow rate is limited in combination with the basic unit water consumption determined in step S4, to reduce invalid iterations;

[0093] In the iteration solving process, the dynamic step strategy is related to the cleanliness constraint and the prediction result; specifically, first, a gap threshold interval is set, when the gap between the predicted value of the cleanliness corresponding to a candidate solution and the preset threshold exceeds the upper limit of the interval, a larger search step is used to speed up the iteration convergence speed; when the gap is within the interval, the search step is reduced for fine search, to improve the accuracy of the optimal solution; in the iteration process, each generation of particles needs to meet the three types of constraint conditions in the foregoing, and the particles violating the constraints are corrected to fall within the feasible region; the iteration termination condition is set, the deviation of the optimal solution of multiple generations is stable within the allowed range or the iteration times reach the preset upper limit, after the iteration is terminated, the current global optimal solution is output as the segment cleaning control parameter instruction, which includes the specific values of the travel speed, the water flow rate and the brush head pressure of each segment.

[0094] Step S54, generate and execute segment control parameters; specifically, on the basis of the segment collected in step S1, refine and divide the rule binding pollution characteristics and robot operation ability, within the same collection segment, according to the image mask and pollution level determination results of step S2, split the areas of different pollution levels and pollution types into independent cleaning segments; adjust the segment area according to the pollution level, the higher the pollution level, the smaller the segment area, and the segment boundary is aligned with the single cleaning width of the robot to avoid parameter adaptation chaos caused by cross-segment cleaning; after the segmentation is completed, a unique identifier is assigned to each segment, and the corresponding pollution level, area and position coordinates are associated to form a segment information table for parameter generation;

[0095] When generating parameters, first, the basic unit water consumption determined in step S4 is weighted and distributed according to the segment pollution level to obtain the initial water flow of each segment, and the higher the pollution level, the greater the weight. Then, the initial water flow, segment pollution level and real-time operation parameters of the robot in step S1 are input into the aforementioned dynamic step particle swarm optimization algorithm, and the accurate control parameters of each segment are output by the algorithm to form a control instruction set with segment identification. The instruction clearly specifies the effective start coordinates and end coordinates of each segment parameter to ensure that the parameters correspond accurately to the segment position;

[0096] The execution process adopts a segmented synchronous control logic, and the main controller carried by the robot calls the control parameters output by the aforementioned algorithm in sequence according to the segment identification to drive the walking mechanism, brush disc driving mechanism and water pump to execute the corresponding parameters synchronously. When switching segments, the controller uses a linear transition strategy to adjust the parameters to avoid equipment impact or cleaning effect fluctuations caused by parameter mutations. During the operation process, a two-dimensional real-time detection mechanism is established, and the cleanliness detection uses a laser reflectance instrument installed on the front beam of the robot. After completing a segment cleaning, multiple-point detection is immediately performed on the segment, and the average value is taken as the actual cleanliness. The robot operation parameters are collected in real time through incremental encoders and pressure sensors, and the collection frequency is synchronized with the cleanliness detection.

[0097] When the actual cleanliness is lower than the preset threshold and the deviation exceeds the allowed range, the aforementioned optimization model is called to re-solve, the water flow of the corresponding segment is adjusted first, and the brush head pressure and travel speed are fine-tuned synchronously to perform a cleaning operation after correction;

[0098] When the brush head pressure and water flow exceed the aforementioned safety constraint range, the corresponding segment operation is immediately paused, the model is re-generated to adapt to the device performance, and the operation is resumed after verification.

[0099] In the embodiment, by indexing the pollution level feature for preprocessing, the prediction sub-model is consistent with the physical law of nonlinear growth of pollution cleaning difficulty, combined with the gradient boosting regression algorithm to capture the complex interaction effect between water flow, pressure, speed and pollution level, and to improve the accuracy of cleanliness prediction; in the optimization solving stage, the dynamic step particle swarm algorithm adaptively adjusts the search behavior according to the difference between the cleanliness prediction value and the threshold, and gives consideration to the convergence speed and the quality accuracy of the solution under the premise of constraint satisfaction; the generation and execution of the segmented control parameter divide the front-end acquisition section as a spatial reference, and embed the device safety constraints and parameter linkage rules to ensure the adaptability of the instruction to the real-time performance of the robot and the safety of the operation process; through the dual-dimensional real-time detection and feedback optimization mechanism of cleanliness and operating parameters, the local that does not meet the standard can be online corrected within a single operation, so as to maintain the stable balance between the water consumption optimization goal and the cleanliness standard constraint in the dynamic operation environment.

[0100] Based on the same inventive concept as the water consumption control method for the photovoltaic panel cleaning robot in the foregoing embodiment, the present application also provides a water consumption control system for a photovoltaic panel cleaning robot, as shown in Figure 2 The system comprises:

[0101] A data acquisition module is configured to synchronously acquire pollution feature images of a current photovoltaic panel area to be cleaned, real-time environmental parameters and robot body operating parameters.

[0102] A pollution feature identification module is configured to perform feature identification on the pollution feature images to obtain a pollution type and a pollution level of the photovoltaic panel area to be cleaned.

[0103] An environmental parameter processing module is configured to perform evaporation evaluation on the real-time environmental parameters to obtain an evaporation intensity index of the photovoltaic panel area to be cleaned.

[0104] A reference water consumption determination module is configured to index the pollution type, the pollution level and the evaporation intensity index as index dimensions in a pre-set water consumption reference mapping table to determine a basic unit water consumption for the current photovoltaic panel area to be cleaned.

[0105] An optimization control module is configured to input the basic unit water consumption, the pollution level and the robot body operating parameters into a pre-set real-time cleaning optimization model to obtain segmented cleaning control parameter instructions and control the cleaning robot to perform cleaning operations according to the instructions.

[0106] The above-mentioned system in the present application can effectively realize a water consumption control method for a photovoltaic panel cleaning robot, and the technical effects thereof are as described in the foregoing embodiment, which will not be repeated here.

[0107] Although the present application has been described in connection with certain specific features and embodiments thereof, it is to be understood that it is provided as an example to the best of the applicant's knowledge and that various modifications and combinations will occur to those skilled in the art. Accordingly, all modifications, combinations and equivalents that fall within the scope of the application are intended to be included herein. It is evident that those skilled in the art can, without departing from the scope of the application, make various changes and modifications of the application to adapt it to various usages and conditions. Thus, such changes and modifications are intended to be included within the scope of the application as defined in the appended claims.

Claims

1. A method for water quantity control for a photovoltaic panel cleaning robot, characterized in that, The method comprises the following steps: Synchronously collecting pollution feature images of the current photovoltaic panel area to be cleaned, real-time environmental parameters and robot body operation parameters; the robot body operation parameters include the current walking speed, the current brush disc pressure and the current water pump pressure; the real-time environmental parameters include the photovoltaic panel surface temperature, the light intensity, the environmental temperature and the wind speed; Performing feature recognition on the pollution feature images to obtain the pollution type and the pollution level of the photovoltaic panel area to be cleaned, which comprises the following steps: using a pre-trained convolutional neural network model to analyze the pollution feature images, outputting the classification result of the pollution type and generating an image mask for identifying the pollution area; comprehensively determining the pollution level according to the pixel proportion covered by the image mask, the average gray value in the area and the texture feature parameters; wherein the pollution level is quantized into multiple discrete levels, and the higher the pollution coverage proportion, the greater the difference between the gray value and the background or the rougher the texture, the higher the determined pollution level; Performing evaporation evaluation on the real-time environmental parameters to obtain the evaporation intensity index of the photovoltaic panel area to be cleaned, which comprises the following steps: establishing a structured physical mixing model: extracting the instantaneous measurement values of the photovoltaic panel surface temperature, the environmental temperature, the wind speed and the light intensity from the synchronously collected data frames; at the same time, calling the inherent parameters related to the photovoltaic panel model and the installation position from the pre-stored database, including the absorption rate of the photovoltaic panel surface to the solar radiation, the installation angle of the photovoltaic panel and the aerodynamic roughness parameter obtained by wind tunnel experiment or on-site calibration for the airflow characteristics of the upper surface of the photovoltaic panel array; calculating the evaporation core thermodynamic driving potential energy, which is represented by the saturated water vapor pressure difference; calculating the net radiation energy input value for water vaporization; calculating the water vapor mass transfer efficiency under the action of the wind speed; combining the net radiation energy input value and the water vapor mass transfer efficiency value according to their respective physical weights for evaporation contribution to obtain the preliminary evaporation rate estimation value; this combination process follows the principles of energy and mass conservation, and the weight coefficient is determined by the local slope of the temperature and saturated water vapor pressure curve and the dry and wet surface constant; then, a dynamic correction factor is introduced, which is obtained from a pre-set query table according to the real-time determined current regional pollution level, and the query table is established through experiments to reflect the comprehensive influence of different pollution types and thicknesses on the surface temperature, hydrophilicity and thermal insulation, so as to dynamically adjust the evaporation rate estimation value and obtain the corrected evaporation rate; the corrected evaporation rate is divided by the standard evaporation rate reference value calculated by the model under the pre-defined standard environmental conditions, including the standard temperature, the standard humidity, the standard wind speed, the standard light and the clean surface state; the division operation produces a dimensionless ratio, which is the final evaporation intensity index; the evaporation intensity index greater than 1 indicates that the current environmental evaporation capacity is higher than the standard condition, and less than 1 indicates that it is lower than the standard condition; Using the pollution type, the pollution level and the evaporation intensity index as index dimensions, indexing in the pre-set water use benchmark mapping table to determine the basic unit water consumption for the current photovoltaic panel area to be cleaned. The base unit water consumption, the pollution level and the robot body operation parameter are input into a preset real-time cleaning optimization model to obtain a segmented cleaning control parameter instruction, and the cleaning robot is controlled to perform cleaning work according to the segmented cleaning control parameter instruction; the segmented cleaning control parameter instruction includes a cleaning path segmented travel speed, a segmented water flow and a segmented brush head pressure; the real-time cleaning optimization model is a nonlinear programming model with constraints, and an objective function thereof is expressed as minimization of total water consumption, and constraint conditions include that a cleanliness prediction value of each cleaning segment needs to be greater than or equal to a preset threshold value, and the cleanliness prediction value is calculated by a preset prediction sub-model according to the pollution level, the water flow, the brush head pressure and the travel speed of the segment.

2. A water usage control method for a photovoltaic panel cleaning robot according to claim 1, characterized in that, The pollution types include dust uniform coverage, sand dust accumulation and bird droplet point pollution.

3. A water usage control method for a photovoltaic panel cleaning robot according to claim 1, characterized in that, The method for constructing the water consumption benchmark mapping table comprises: In a standard test environment, for each combination of the pollution types and the pollution levels, and different intervals corresponding to the evaporation intensity indicators, the minimum unit water consumption is determined by an orthogonal experiment method while ensuring that a preset base cleanliness meets a standard; The pollution types, the pollution levels and the evaporation intensity indicator intervals are taken as input features, and the determined minimum unit water consumption is taken as an output label, and a regression model is trained, and an output result of the regression model is discretized to form the water consumption benchmark mapping table.

4. A water usage control method for a photovoltaic panel cleaning robot according to claim 3, characterized in that, The preset prediction sub-model is a machine learning model trained based on historical cleaning data, and the input of the model is the pollution level, the planned applied water flow, the brush head pressure and the travel speed of a cleaning segment, and the output of the model is the cleanliness prediction value of the cleaning segment; The cleanliness prediction value is positively correlated with the water flow and the brush head pressure, and is negatively correlated with the travel speed and the pollution level, and the negative influence of the pollution level on the cleaning effect presents an exponential decay trend.

5. A water usage control method for a photovoltaic panel cleaning robot according to claim 4, characterized in that, The real-time cleaning optimization model adopts a particle swarm optimization algorithm; In an iterative solving process of the particle swarm optimization algorithm, a search step of the particle swarm optimization algorithm is adjusted according to a gap between the cleanliness prediction value corresponding to a current candidate solution and the preset threshold value; When the cleanliness prediction value approaches the preset threshold value, the search step is reduced for fine search.

6. A water consumption control system for a photovoltaic panel cleaning robot, the system being applied to a water consumption control method for a photovoltaic panel cleaning robot according to claim 1, characterized in that, The system comprises: A data acquisition module is configured to synchronously acquire pollution feature images, real-time environmental parameters and robot body operation parameters of a current photovoltaic panel area to be cleaned; A pollution feature identification module is configured to perform feature identification on the pollution feature images to obtain pollution types and pollution levels of the photovoltaic panel area to be cleaned; An environmental parameter processing module is configured to perform evaporation evaluation on the real-time environmental parameters to obtain an evaporation intensity indicator of the photovoltaic panel area to be cleaned; A benchmark water consumption determination module is configured to index, in a preset water consumption benchmark mapping table, the pollution types, the pollution levels and the evaporation intensity indicator as index dimensions to determine a base unit water consumption for the current photovoltaic panel area to be cleaned. An optimization control module is configured to input the basic unit water consumption, the pollution level and the robot body operation parameter into a preset real-time cleaning optimization model, obtain segmented cleaning control parameter instructions, and control the cleaning robot to perform cleaning work according to the segmented cleaning control parameter instructions.

Images