Photovoltaic panel automatic cleaning system and method based on unmanned aerial vehicle

Abstract

The invention discloses a photovoltaic panel automatic cleaning system based on an unmanned aerial vehicle and a method thereof, and relates to the technical field of solar photovoltaic. The system comprises an unmanned aerial vehicle, a pollutant cleaning device, a pollutant detection device and a remote control center, comprising the steps of planning a flight path of an unmanned aerial vehicle and obtaining a photovoltaic panel image to determine the type, position and cleaning time of pollutants; a cleaning mode is selected to carry out cleaning operation, and secondary detection is carried out on the cleaned photovoltaic panel; and recording each cleaning task and process and establishing a self-adaptive model. The unmanned aerial vehicle is used for monitoring the photovoltaic panel in real time, image analysis and pollutant identification are carried out through the remote control center, the cleaning device is used for cleaning the photovoltaic panel, and the pollutant detection device is used for carrying out secondary detection after cleaning; according to the cleaning device, multiple cleaning modes are combined, flexible selection is achieved according to the types of pollutants, the cleaning effect of the photovoltaic panel is guaranteed, the cleaning efficiency is improved, and meanwhile the energy-saving effect is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of solar photovoltaics, and in particular relates to an automatic cleaning system and method for photovoltaic panels based on an unmanned aerial vehicle. Background Art

[0002] Photovoltaic panels are devices that use the photoelectric effect to convert solar energy into electrical energy. They play an important role in the field of renewable energy and are widely used in household, commercial and industrial electricity production. Since solar photovoltaic panels are usually installed in outdoor environments, environmental factors have a great impact on their overall performance, especially the long-term accumulation of dust, leaves, bird droppings, etc. on photovoltaic panels, which seriously affects the power generation efficiency. However, manual cleaning of photovoltaic panels is inefficient and poses safety risks. Therefore, regular cleaning of photovoltaic panels is an important means to maintain the performance and power generation efficiency of photovoltaic panels.

[0003] In the prior art, cleaning stains on photovoltaic panels is often done by combining drones with cleaning devices. However, due to the limited size of drones, the cleaning devices they carry are usually not too heavy and cannot have a high water carrying capacity, which will result in the cleaning efficiency of drones on photovoltaic panels being limited to a certain extent and the cleaning effect cannot be guaranteed. It will also increase the cleaning cost and energy consumption; and the stain recognition ability of the drone itself will also affect the cleaning effect and efficiency. Therefore, based on the cleaning of photovoltaic panels by drones, it has become an urgent problem to improve the cleaning device carried by drones and improve the ability to recognize stains / dirt on photovoltaic panels. Summary of the invention

[0004] The object of the present invention is to provide a photovoltaic panel automatic cleaning system and method based on a drone, which can be achieved through the following technical solutions: In a first aspect, an embodiment of the present application provides a photovoltaic panel automatic cleaning system based on a drone, comprising: a drone, a pollutant cleaning device carried by the drone, a pollutant detection device carried by the drone, and a remote control center; The remote control center is used to obtain the operation data of the photovoltaic panels and the geographical information and meteorological data of the area where the photovoltaic panels are located, and use them to plan the flight path of the drone; The drone is used to monitor the photovoltaic panels in real time according to the planned flight path to obtain images of the photovoltaic panels; The remote control center is also used to analyze and identify the photovoltaic panel image to determine the type and location of the contaminant; and is also used to determine the cleaning time of the photovoltaic panel; The pollutant cleaning device is used to select a cleaning method and a cleaning device according to the type of the pollutant and the cleaning time of the photovoltaic panel and perform a cleaning operation according to the location of the pollutant; The pollutant detection device is used to perform a secondary inspection on the cleaned photovoltaic panel and determine the cleaning status of the photovoltaic panel. If pollutants still exist on the cleaned photovoltaic panel, a cleaning method corresponding to the type of pollutant is selected again to perform a cleaning operation. Otherwise, a cleaning completion signal is output; The remote control center is also used to record the cleaning tasks and cleaning processes of each photovoltaic panel cleaning performed by the drone according to the cleaning completion signal, and to establish an adaptive model based on the cleaning tasks and cleaning processes. Wherein, the photovoltaic panel image includes a first panel image and a second panel image; The adaptive model is used to adaptively adjust the relevant parameters of the pollutant cleaning device after each round of drone cleaning operation; Determining the type and location of contaminants at the remote control center includes: Pre-building a pollutant detection model in the remote control center; Inputting the photovoltaic panel image into the pollutant detection model for identification to obtain an identification result specifically includes: The first panel image is input into the pollutant detection model for recognition to obtain a first recognition result, specifically: Inputting the first panel image into the pollutant detection model to obtain an original standard image corresponding to the location of the first panel image; The first panel image is compared with the original standard image for similarity, and if the comparison result exceeds a set similarity threshold, the first panel image currently being compared is determined to be a suspected pollutant image; Performing edge detection on the suspected pollutant image to obtain the pollutant contour; Performing target detection based on the pollutant profile to obtain the type of the pollutant and outputting a first recognition result; The second panel image is input into the pollutant detection model for recognition to obtain a second recognition result, specifically: The second panel image is input into the pollutant detection model for identification. If the second panel image matches the light spot shadow image in the model, it is determined as a non-pollutant image, otherwise it is determined as a pollutant image; Further identification of pollutant images is performed, specifically: Calculate the pollutant attachment rate of the pollutant image. If the pollutant attachment rate is greater than the set standard pollutant attachment probability, the pollutant in the pollutant image is determined to be a non-dust particle pollutant; otherwise, it is determined to be a dust particle pollutant. The non-dust pollutant images determined to be non-dust particle pollutants are further identified as follows: Calculate the dirty color depth value of the non-dust pollutant image. If the dirty color depth value is greater than the set color depth reference value, the non-dust pollutant image is determined to be a heavy pollutant image, otherwise it is a light pollutant image. outputting a second recognition result; The first identification result and the second identification result are combined to determine the type and location of the contaminant.

[0005] Preferably, planning the flight path of the drone includes the following: Establish a digital twin model based on the operating data of photovoltaic panels, and use it to monitor the operating status and performance changes of photovoltaic panels; Determine the abnormal time and location of the photovoltaic panel when the photovoltaic panel has abnormal operation according to the geographical information and meteorological data of the area where the photovoltaic panel is located; Construct a dynamic spatiotemporal map of the entire photovoltaic panel area based on the abnormal time and location of the photovoltaic panels; The point where the drone takes off relative to the photovoltaic panel is used as the origin to construct a position coordinate system in the dynamic space-time map; The abnormal position of the photovoltaic panel is used as the target position of the UAV flight, the optimal path of the UAV flight is calculated in the position coordinate system, and the optimal path is used as the final flight path of the UAV. Preferably, determining the abnormal position of the photovoltaic panel includes the following contents: Obtain historical photovoltaic data and structural data of photovoltaic panels; The photovoltaic panels are divided into multiple photovoltaic areas and several photovoltaic modules based on historical photovoltaic data, structural data and geographic information; Select a photovoltaic module and detect the power generation value of this photovoltaic module under the current light intensity; Calculate the difference between the power generation value and the power generation standard value, and record it as the power generation difference; The remote control center determines the difference between the power generation amount and the preset abnormal threshold value. If the power generation amount difference is greater than the preset abnormal threshold value, a power generation abnormality signal is generated and the photovoltaic module is marked as an abnormal module; otherwise, the next photovoltaic module is selected to detect the power generation value; Record the location information of each abnormal module on the photovoltaic panel and construct an abnormal location set based on it; Wherein, the abnormal position of the photovoltaic panel is selected and determined in the abnormal position set; The historical photovoltaic data includes historical light intensity, historical temperature and humidity, historical current data, historical voltage data and historical power generation efficiency; the structural data includes series structure, parallel structure, size data and arrangement method.

[0006] Preferably, acquiring the photovoltaic panel image includes: The UAV determines the abnormal module on the photovoltaic panel according to the abnormal position of the photovoltaic panel, and collects images of the abnormal module to obtain the first panel image. Preferably, acquiring the photovoltaic panel image further comprises: The drone collects images of the entire area of ​​the photovoltaic panel and cuts the collected images to obtain a number of cut images of equal size; Extracting color features of the cut image and performing feature screening to obtain a feature image; Performing pollutant identification on the characteristic image and selecting an image containing pollutants as a second panel image; Among them, feature screening is performed to obtain feature images, specifically: The color feature is compared with the standard feature when the photovoltaic panel is not contaminated and the characteristic color difference is obtained. If the characteristic color difference is greater than the set color difference threshold, the cutting image represented by this color feature is selected as the characteristic image, otherwise the next round of feature comparison is performed.

[0007] Preferably, the pollutant detection model is pre-built, including: Obtain model training data, including image samples of photovoltaic panels and labels corresponding to the image samples; Inputting the model training data into a deep learning model to extract features of image samples; Make predictions based on the extracted features to output prediction results; Generate loss values ​​based on prediction results and labels; Training the deep learning model according to the loss value until an end condition of the model training is met, and determining the deep learning model that has finished training as the pollutant detection model; The image samples include images of photovoltaic panels with pollutants, images of photovoltaic panels without pollutants, and images of photovoltaic panels suspected of being polluted; The labels corresponding to the image samples include whether there is a pollutant, the type of the pollutant, the shape of the pollutant, the size of the pollutant, and the location of the pollutant. Preferably, the cleaning methods include blowing cleaning, water washing cleaning and blowing-water washing cleaning.

[0008] Preferably, the cleaning time of the photovoltaic panel is determined as follows: Based on the operation data of the photovoltaic panel, the power generation efficiency and current power generation efficiency at each image acquisition time point are obtained, and the power generation evaluation coefficient of the photovoltaic panel is analyzed; Based on the geographic information and meteorological data of the area where the photovoltaic panels are located, the current light intensity and current irradiance at the current time point are obtained, and the environmental assessment coefficient of the photovoltaic panels is analyzed; Analyze the power generation assessment coefficient and environmental assessment coefficient to determine the cleaning time of photovoltaic panels; The contamination area is calculated based on the type of contamination and used to determine the duration of PV panel cleaning. Preferably, establishing the adaptive model comprises: Inspect the surface of the photovoltaic panel that has completed the cleaning task to obtain cleanliness parameters; assessing the initial surface cleanliness of the photovoltaic panel according to the cleanliness parameter; extracting cleaning interference parameters related to the initial surface cleanliness; Setting a defect cleaning assessment factor and an environmental load cleaning assessment factor of the photovoltaic panel according to the cleaning interference parameter; The surface cleanliness of the photovoltaic panel is evaluated again based on the defect cleaning assessment factor and the environmental load cleaning assessment factor, and the final comprehensive surface cleanliness is obtained by statistics; Conduct qualification evaluation on the cleaning process of completing the cleaning task and obtain the task evaluation results; Establishing the adaptive model based on the comprehensive surface cleanliness and the task evaluation result; The cleanliness parameters include the surface image, the ambient temperature of each monitoring period, the scanned temperature distribution diagram, and the reflectivity in each area of ​​the photovoltaic panel; the clean interference parameters include structural interference parameters, chromatic interference parameters, and environmental interference parameters.

[0009] In a second aspect, an embodiment of the present application provides a method for automatically cleaning a photovoltaic panel based on a drone, which is applied to the automatic cleaning system for a photovoltaic panel as described above, and includes the following steps: Obtain the operating data of the photovoltaic panels and the geographical information and meteorological data of the area where the photovoltaic panels are located, and use them to plan the flight path of the drone; Monitoring the photovoltaic panel in real time according to the planned flight path to obtain a photovoltaic panel image; the photovoltaic panel image includes a first panel image and a second panel image; Analyzing and identifying the photovoltaic panel image to determine the type and location of contaminants; and also used to determine when to clean the photovoltaic panel; Selecting a cleaning method and a cleaning device according to the type of the pollutant and the cleaning time of the photovoltaic panel and performing a cleaning operation according to the location of the pollutant; Perform secondary inspection on the cleaned photovoltaic panel and judge the cleaning status of the photovoltaic panel. If there are still pollutants on the cleaned photovoltaic panel, select the cleaning method corresponding to the type of pollutants to perform the cleaning operation again, otherwise output a cleaning completion signal; Recording the cleaning tasks and cleaning processes of each photovoltaic panel cleaning using the drone according to the cleaning completion signal, and establishing an adaptive model based on the cleaning tasks and cleaning processes; Wherein, the photovoltaic panel image includes a first panel image and a second panel image; The adaptive model is used to adaptively adjust the relevant parameters of the pollutant cleaning device after each round of drone cleaning operation; Determining the type and location of contaminants at the remote control center includes: Pre-building a pollutant detection model in the remote control center; Inputting the photovoltaic panel image into the pollutant detection model for identification to obtain an identification result specifically includes: The first panel image is input into the pollutant detection model for recognition to obtain a first recognition result, specifically: Inputting the first panel image into the pollutant detection model to obtain an original standard image corresponding to the location of the first panel image; The first panel image is compared with the original standard image for similarity, and if the comparison result exceeds a set similarity threshold, the first panel image currently being compared is determined to be a suspected pollutant image; Performing edge detection on the suspected pollutant image to obtain the pollutant contour; Performing target detection based on the pollutant profile to obtain the type of the pollutant and outputting a first recognition result; The second panel image is input into the pollutant detection model for recognition to obtain a second recognition result, specifically: The second panel image is input into the pollutant detection model for identification. If the second panel image matches the light spot shadow image in the model, it is determined as a non-pollutant image, otherwise it is determined as a pollutant image; Further identification of pollutant images is performed, specifically: Calculate the pollutant attachment rate of the pollutant image. If the pollutant attachment rate is greater than the set standard pollutant attachment probability, the pollutant in the pollutant image is determined to be a non-dust particle pollutant; otherwise, it is determined to be a dust particle pollutant. The non-dust pollutant images determined to be non-dust particle pollutants are further identified as follows: Calculate the dirty color depth value of the non-dust pollutant image. If the dirty color depth value is greater than the set color depth reference value, the non-dust pollutant image is determined to be a heavy pollutant image, otherwise it is a light pollutant image. outputting a second recognition result; The first identification result and the second identification result are combined to determine the type and location of the contaminant.

[0010] The beneficial effects of the present invention are: (1) The present invention utilizes a camera device carried by a drone to monitor photovoltaic panels in real time, and performs image analysis and pollutant identification through a remote control center. At the same time, a cleaning device carried by the drone is used to clean the photovoltaic panels, and a pollutant detection device is used to perform re-detection after cleaning. In addition, the cleaning device carried by the drone is improved in the present invention, and a variety of cleaning methods such as air blowing cleaning, water washing cleaning, and a combination of air blowing and water washing cleaning are adopted. Flexible selection can be made according to the type of pollutants, thereby ensuring the cleaning effect of the drone on the photovoltaic panels and improving the cleaning efficiency. At the same time, an adaptive model is established to automatically adjust subsequent cleaning parameters to achieve energy-saving effects.

[0011] (2) The present invention first obtains the operation data of the photovoltaic panels and the geographical information and meteorological data of the area where the photovoltaic panels are located through the remote control center, and plans the flight path of the drone based on this data; then the drone monitors the photovoltaic panels in real time according to the planned flight path to obtain the photovoltaic panel images; then the remote control center analyzes and identifies the photovoltaic panel images to determine the type and location of the pollutants, and also determines the cleaning time of the photovoltaic panels; then the pollutant cleaning device carried by the drone selects the corresponding cleaning method and cleaning equipment according to the type of pollutants and the cleaning time of the photovoltaic panels, and performs the cleaning operation based on the location of the pollutants; then the pollutant detection device carried by the drone performs a secondary detection on the cleaned photovoltaic panels and determines the current cleaning status of the photovoltaic panels. If the pollutants are not clean, the cleaning method and cleaning equipment are selected according to the type of pollutants and the cleaning time of the photovoltaic panels. If there are still pollutants on the photovoltaic panel after cleaning, the cleaning method corresponding to the pollutant type is selected again for cleaning operation. If there are no pollutants on the photovoltaic panel after cleaning, a cleaning completion signal indicating the completion of the current or this round of cleaning task is output; finally, the remote control center records the cleaning tasks and cleaning processes each time the photovoltaic panel is cleaned by the drone according to the above cleaning completion signal, and establishes an adaptive model based on this; it should be noted that the adaptive model is used to adaptively adjust the cleaning parameters of the above pollutant cleaning device after each round of drone cleaning operation, so as to make the best cleaning decision for each round of photovoltaic panel cleaning, and also enable the drone to achieve adaptive adjustment, so as to achieve the purpose of saving energy while ensuring the cleaning effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] For better understanding and implementation, the technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0013] Figure 1 A schematic diagram of the structure of an automatic photovoltaic panel cleaning system based on a drone provided in an embodiment of the present application; Figure 2 A flowchart of the steps for planning the flight path of a drone provided in an embodiment of the present application; Figure 3A flowchart of the steps for determining the type and location of pollutants provided in an embodiment of the present application; Figure 4 A flowchart of the steps of an automatic cleaning method of photovoltaic panels based on a drone provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, exemplary embodiments will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are only examples of methods and systems consistent with some aspects of the present application as detailed in the attached claims.

[0015] The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in this article refers to any or all possible combinations of one or more associated listed items.

[0016] The specific implementation methods, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0017] Example 1 See also Figure 1 , an embodiment of the present application provides a photovoltaic panel automatic cleaning system based on a drone, including a drone, a pollutant cleaning device carried by the drone, a pollutant detection device carried by the drone, and a remote control center; The remote control center is used to obtain the operation data of the photovoltaic panels and the geographical information and meteorological data of the area where the photovoltaic panels are located, and use them to plan the flight path of the drone; The drone is used to monitor the photovoltaic panel in real time according to the planned flight path to obtain a photovoltaic panel image; the photovoltaic panel image includes a first panel image and a second panel image; The remote control center is also used to analyze and identify the photovoltaic panel image to determine the type and location of the contaminant; and is also used to determine the cleaning time of the photovoltaic panel; The pollutant cleaning device is used to select a cleaning method and a cleaning device according to the type of the pollutant and the cleaning time of the photovoltaic panel and perform a cleaning operation according to the location of the pollutant; The pollutant detection device is used to perform a secondary inspection on the cleaned photovoltaic panel and determine the cleaning status of the photovoltaic panel. If pollutants still exist on the cleaned photovoltaic panel, a cleaning method corresponding to the type of pollutant is selected again to perform a cleaning operation. Otherwise, a cleaning completion signal is output; The remote control center is also used to record the cleaning tasks and cleaning processes of each photovoltaic panel cleaning performed by the drone according to the cleaning completion signal, and to establish an adaptive model based on the cleaning tasks and cleaning processes. The adaptive model is used to adaptively adjust the relevant parameters of the pollutant cleaning device after each round of drone cleaning operation.

[0018] Specifically, in the prior art, the cleaning efficiency of photovoltaic panels by drones is limited to a certain extent and the cleaning effect cannot be guaranteed, which will also increase the cleaning cost and energy consumption; and the stain recognition ability of the drone itself will also affect the cleaning effect and efficiency. Therefore, in order to ensure the effect and efficiency of using drones to clean photovoltaic panels, this application improves the cleaning device carried by drones and improves the ability to recognize pollutants (stains / dirt) on photovoltaic panels. At the same time, a pollutant detection device is added to the drone to ensure the cleaning effect of the photovoltaic panels. The specific contents are as follows: First, the operation data of the photovoltaic panels and the geographical information and meteorological data of the area where the photovoltaic panels are located are obtained through the remote control center, and the flight path of the drone is planned based on this data; then the drone is used to monitor the photovoltaic panels in real time according to the planned flight path to obtain the photovoltaic panel images; then the remote control center analyzes and identifies the photovoltaic panel images to determine the type and location of the pollutants, and also determines the cleaning time of the photovoltaic panels; then the pollutant cleaning device carried by the drone is used to select the corresponding cleaning method and cleaning equipment according to the type of pollutants and the cleaning time of the photovoltaic panels, and the cleaning operation is performed based on the location of the pollutants; then the pollutant detection device carried by the drone is used to perform a secondary detection on the cleaned photovoltaic panels and judge the current cleaning status of the photovoltaic panels. If the cleaning is If there are still pollutants on the photovoltaic panel, the cleaning method corresponding to the pollutant type is selected again for cleaning operation. If there are no pollutants on the photovoltaic panel after cleaning, a cleaning completion signal indicating the completion of the current or this round of cleaning task is output; finally, the remote control center records the cleaning tasks and cleaning processes of each photovoltaic panel cleaning by the drone according to the above cleaning completion signal, and establishes an adaptive model based on this; it should be noted that the adaptive model is used to adaptively adjust the cleaning parameters of the above pollutant cleaning device after each round of drone cleaning operation, so as to make the best cleaning decision for each round of photovoltaic panel cleaning, and also enable the drone to achieve adaptive adjustment, so as to achieve the purpose of saving energy while ensuring the cleaning effect.

[0019] like Figure 2 As shown, in one embodiment provided in the present application, planning the flight path of the drone includes the following contents: Establish a digital twin model based on the operating data of photovoltaic panels, and use it to monitor the operating status and performance changes of photovoltaic panels; Determine the abnormal time and location of the photovoltaic panel when the photovoltaic panel has abnormal operation according to the geographical information and meteorological data of the area where the photovoltaic panel is located; Construct a dynamic spatiotemporal map of the entire photovoltaic panel area based on the abnormal time and location of the photovoltaic panels; The point where the drone takes off relative to the photovoltaic panel is used as the origin to construct a position coordinate system in the dynamic space-time map; The abnormal position of the photovoltaic panel is used as the target position of the UAV flight, the optimal path of the UAV flight is calculated in the position coordinate system, and the optimal path is used as the final flight path of the UAV.

[0020] Specifically, this embodiment monitors and optimizes the performance of photovoltaic panels in real time by establishing a digital twin model, thereby improving the power generation efficiency and performance of photovoltaic panels; it also constructs a dynamic spatiotemporal map of photovoltaic panels through the monitored abnormal data (time data and location data), thereby providing spatial information for the path planning of drones, and reflecting dynamic changes, providing more comprehensive data support for the calculation of the optimal path; and this embodiment also uses the abnormal time of photovoltaic panels as a time reference factor for selecting the pollutant cleaning method, thereby providing data support for the subsequent selection of a suitable cleaning method in the pollutant cleaning device. It should be noted that the abnormal time of photovoltaic panels can reflect the seasonal changes in the area where the photovoltaic panels are located, and can also reflect the time changes in the cleaning of photovoltaic panels. For example, if the seasons in the area where the photovoltaic panels are located are spring and summer, it is easy to accumulate dust and feces, while in autumn and winter, it is easy to accumulate snow and fallen leaves; on the other hand, light, temperature, humidity and wind speed will affect the cleaning effect of photovoltaic panels, so the abnormal time of photovoltaic panels can be used as one of the reference factors for selecting the cleaning method, so that the selection of the cleaning method is more accurate, reasonable and reliable.

[0021] It should be noted that geographic information includes longitude and latitude, altitude, terrain and orientation, as well as tilt angle, etc. Specifically, longitude and latitude: the precise geographic coordinates of photovoltaic panels, which are helpful in calculating sunlight conditions and angles; altitude: affects air density and temperature, which in turn affects the power generation efficiency of photovoltaic panels; terrain and orientation: depending on the installation location and direction of photovoltaic panels (south, east, west, etc.), the sunlight conditions and power generation efficiency in different regions may be different; tilt angle: the tilt angle of photovoltaic panels has a direct impact on light intensity and power generation; meteorological data include average annual sunshine hours, light intensity, precipitation, temperature and humidity, number of cloudy and sunny days, duration of different seasons, etc.; specifically, average annual sunshine hours: sunshine time in a region, which is very important for the assessment of photovoltaic power generation potential; light intensity: including changes in light intensity in different seasons and under different weather conditions; precipitation: affects the cleanliness and corrosion of photovoltaic panels; temperature and humidity: ambient temperature and humidity will affect the power generation efficiency of photovoltaic panels.

[0022] It should be noted that the above operating data include but are not limited to data related to the performance of photovoltaic panels, such as voltage, current, power generation efficiency and power generation.

[0023] In an embodiment provided in the present application, determining the abnormal position of a photovoltaic panel includes the following contents: Obtain historical photovoltaic data and structural data of photovoltaic panels; Based on historical photovoltaic data, structural data and geographic information, the photovoltaic panel is divided into multiple photovoltaic areas and several photovoltaic modules; wherein multiple photovoltaic areas represent multiple larger blocks on the photovoltaic panel, and each photovoltaic area contains multiple photovoltaic modules, which can be understood as: multiple photovoltaic modules constitute a photovoltaic area, and multiple photovoltaic areas constitute the entire photovoltaic panel; Select a photovoltaic module and detect the power generation value of this photovoltaic module under the current light intensity; Calculate the difference between the power generation value and the power generation standard value, and record it as the power generation difference; The remote control center determines the difference between the power generation amount and the preset abnormal threshold value. If the power generation amount difference is greater than the preset abnormal threshold value, a power generation abnormality signal is generated and the photovoltaic module is marked as an abnormal module; otherwise, the next photovoltaic module is selected to detect the power generation value; Record the location information of each abnormal module on the photovoltaic panel and construct an abnormal location set based on it; Wherein, the abnormal position of the photovoltaic panel is selected and determined in the abnormal position set; The historical photovoltaic data includes historical light intensity, historical temperature and humidity, historical current data, historical voltage data and historical power generation efficiency; the structural data includes series structure, parallel structure, size data and arrangement mode. The above arrangement modes include linear arrangement and grid arrangement.

[0024] In an embodiment provided in the present application, acquiring the photovoltaic panel image includes: The UAV determines the abnormal module on the photovoltaic panel according to the abnormal position of the photovoltaic panel, and collects images of the abnormal module to obtain the first panel image.

[0025] In an embodiment provided in the present application, acquiring the photovoltaic panel image further includes: The drone collects images of the entire area of ​​the photovoltaic panel and cuts the collected images to obtain a number of cut images of equal size; Extracting color features of the cut image and performing feature screening to obtain a feature image; Performing pollutant identification on the characteristic image and selecting an image containing pollutants as a second panel image; Among them, feature screening is performed to obtain feature images, specifically: The color feature is compared with the standard feature when the photovoltaic panel is not contaminated and the characteristic color difference is obtained. If the characteristic color difference is greater than the set color difference threshold, the cutting image represented by this color feature is selected as the characteristic image, otherwise the next round of feature comparison is performed.

[0026] Specifically, in this embodiment, the entire area of ​​the photovoltaic panel is imaged by a drone, and the images within the range are cut and divided into several cut images of equal size. Since these cut images are only of equal size, some of them may contain pollutants, while others may not. Therefore, this embodiment extracts features from the cut images, and screens the feature images by comparing the color difference between the features with the set color difference threshold, and uses the screened feature images as the second panel image. At this time, the screened images are all images containing suspected pollutants (this embodiment is only a preliminary judgment, and perhaps the screened images contain shadows cast by obstructions on the photovoltaic panel, not pollutants). It should be noted that the method used in this embodiment is different from the means used to obtain the first panel image. The two use different means, so the first panel image and the second panel image are used as a distinction to provide data support for the subsequent analysis of the photovoltaic panel image and the identification of pollutants. It can also be understood that the present application obtains the above photovoltaic panel images by two different means to improve the accuracy and reliability of subsequent image analysis and pollutant identification.

[0027] like Figure 3 As shown, in one embodiment provided in the present application, determining the type and location of the pollutant at the remote control center includes: Pre-building a pollutant detection model in the remote control center; The photovoltaic panel image is input into the pollutant detection model for identification to obtain an identification result, specifically: Inputting the first panel image into the pollutant detection model for recognition to obtain a first recognition result; inputting the second panel image into the pollutant detection model for identification to obtain a second identification result; The first identification result and the second identification result are combined to determine the type and location of the contaminant.

[0028] Specifically, the present application identifies the first panel image and the second panel image respectively and obtains identification results, and combines the two identification results to jointly determine the type and location of the pollutant, so as to ensure the accuracy and reliability of image recognition.

[0029] In one embodiment provided in the present application, the pollutant detection model is pre-built, including: Obtain model training data, including image samples of photovoltaic panels and labels corresponding to the image samples; Inputting the model training data into a deep learning model to extract features of image samples; Make predictions based on the extracted features to output prediction results; Generate loss values ​​based on prediction results and labels; Training the deep learning model according to the loss value until an end condition of the model training is met, and determining the deep learning model that has finished training as the pollutant detection model; The image samples include images of photovoltaic panels with pollutants, images of photovoltaic panels without pollutants, and images of photovoltaic panels suspected of being polluted; The labels corresponding to the image samples include whether there is a pollutant, the type of the pollutant, the shape of the pollutant, the size of the pollutant, and the location of the pollutant.

[0030] Specifically, the pollutant detection model of this embodiment is obtained by training with a deep learning model. In this embodiment, the model training data is first obtained, including but not limited to the image samples of the photovoltaic panel and the labels corresponding to the image samples, and the labels may include whether there are pollutants, the type of pollutants, the shape of the pollutants, the size of the pollutants, and the location of the pollutants. The image samples may include images of photovoltaic panels with pollutants, images of photovoltaic panels without pollutants, and images of photovoltaic panels with suspected pollutants; then the model training data is input into the deep learning model for feature extraction, and prediction is performed based on the extracted features to output the prediction results; then a loss value is generated based on the prediction results and the labels, and the deep learning model is trained based on the loss value; the deep learning model is continuously trained through the above steps until the end condition is met, and the model that ends the training at this time is determined as the pollutant detection model. This embodiment enables this pollutant detection model to have the ability to predict the probability and location of various types of pollutants on the photovoltaic panel through the above steps. This embodiment trains and establishes a pollutant detection model in advance, so that a lot of time and resources can be saved when analyzing and identifying images in the remote control center, thereby ensuring the efficiency of image recognition while also improving the efficiency of automatic cleaning of photovoltaic panels.

[0031] In an embodiment provided in the present application, performing image recognition on the first panel image includes: Inputting the first panel image into the pollutant detection model to obtain an original standard image corresponding to the location of the first panel image; The first panel image is compared with the original standard image for similarity, and if the comparison result exceeds a set similarity threshold, the first panel image currently being compared is determined to be a suspected pollutant image; Performing edge detection on the suspected pollutant image to obtain the pollutant contour; Target detection is performed based on the pollutant profile to obtain the type of the pollutant and output a first recognition result.

[0032] In an embodiment provided in the present application, performing image recognition on the second panel image includes: The second panel image is input into the pollutant detection model for identification. If the second panel image matches the light spot shadow image in the model, it is determined as a non-pollutant image, otherwise it is determined as a pollutant image; Further identification of pollutant images is performed, specifically: Calculate the pollutant attachment rate of the pollutant image. If the pollutant attachment rate is greater than the set standard pollutant attachment probability, the pollutant in the pollutant image is determined to be a non-dust particle pollutant; otherwise, it is determined to be a dust particle pollutant. The non-dust pollutant images determined to be non-dust particle pollutants are further identified as follows: Calculate the dirty color depth value of the non-dust pollutant image. If the dirty color depth value is greater than the set color depth reference value, the non-dust pollutant image is determined to be a heavy pollutant image, otherwise it is a light pollutant image. The second recognition result is output.

[0033] Specifically, this embodiment firstly identifies and excludes the image of the light spot shadow in the second panel image, and then further identifies the pollutant image, and determines whether the pollutant is a dust particle by judging the attachment rate of the pollutant in the image. If not, it means that the pollutant is a non-dust particle pollutant such as a liquid or a solid (the solid here refers to a stubborn object condensed after bird droppings or liquid evaporation), and then judges the degree of pollution of this non-dust particle pollutant again. If it is light pollution, the number of subsequent cleanings does not need to be too many. If it is heavy pollution, it will take many repeated cleanings to completely clean it. It should be noted that the difference between dust particle pollutants and non-dust particle pollutants can be distinguished by the cleaning method used. If it is a dust particle pollutant, the cleaning method adopts the wind blowing method, otherwise it adopts the water washing method. If it is a combination of the two (that is, the dust particle pollutant is on the surface, and the non-dust particle pollutant is on the bottom), the wind blowing and water washing methods are combined.

[0034] In an embodiment provided in the present application, the cleaning time of the photovoltaic panel is determined as follows: Based on the operation data of the photovoltaic panel, the power generation efficiency and current power generation efficiency at each image acquisition time point are obtained, and the power generation evaluation coefficient of the photovoltaic panel is analyzed; Based on the geographic information and meteorological data of the area where the photovoltaic panels are located, the current light intensity and current irradiance at the current time point are obtained, and the environmental assessment coefficient of the photovoltaic panels is analyzed; Analyze the power generation assessment coefficient and environmental assessment coefficient to determine the cleaning time of photovoltaic panels; The contamination area is calculated based on the type of contamination and used to determine the duration of PV panel cleaning.

[0035] Specifically, this embodiment determines the power generation evaluation coefficient of the photovoltaic panel by analyzing the current power generation efficiency, and also derives the environmental assessment coefficient of the photovoltaic panel through the current light intensity and the current irradiance, and uses the above-mentioned power generation evaluation coefficient and environmental assessment coefficient to derive the optimal time for cleaning the photovoltaic panel, and determines the time required for cleaning by calculating the area of ​​the pollutants.

[0036] It should be noted that due to the influence of environmental factors such as weather and season, the cleaning of photovoltaic panels needs to take into account the above-mentioned environmental influences. For example, cleaning photovoltaic panels under high temperature conditions in summer will have a great impact on the operating performance and water washing cleaning of the drone. The water sprayed by the water washing method will evaporate faster due to the high temperature, which greatly prolongs the cleaning duration. Therefore, this embodiment comprehensively considers the influence of the above-mentioned environmental factors and selects the best cleaning time to clean the drone. This will not delay the power generation efficiency of the photovoltaic panels and can also better complete the cleaning task.

[0037] In an embodiment provided in the present application, establishing the adaptive model includes: Inspect the surface of the photovoltaic panel that has completed the cleaning task to obtain cleanliness parameters; assessing the initial surface cleanliness of the photovoltaic panel according to the cleanliness parameter; extracting cleaning interference parameters related to the initial surface cleanliness; Setting a defect cleaning assessment factor and an environmental load cleaning assessment factor of the photovoltaic panel according to the cleaning interference parameter; The surface cleanliness of the photovoltaic panel is evaluated again based on the defect cleaning assessment factor and the environmental load cleaning assessment factor, and the final comprehensive surface cleanliness is obtained by statistics; Conduct qualification evaluation on the cleaning process of completing the cleaning task and obtain the task evaluation results; Establishing the adaptive model based on the comprehensive surface cleanliness and the task evaluation result; The cleanliness parameters include the surface image, the ambient temperature of each monitoring period, the scanned temperature distribution diagram, and the reflectivity in each area of ​​the photovoltaic panel; the clean interference parameters include structural interference parameters, chromatic interference parameters, and environmental interference parameters.

[0038] Specifically, this embodiment evaluates the surface cleanliness from three dimensions: the appearance, surface temperature, and surface reflection of the photovoltaic panel, and sets the evaluation factor of the surface cleanliness in combination with the performance status, structural information, and environmental information of the photovoltaic panel, thereby analyzing the comprehensive surface cleanliness of the photovoltaic panel; and establishes an adaptive model in combination with the evaluation results of the cleaning task, and adaptively adjusts the relevant parameters of the pollutant cleaning device after each round of drone cleaning operation, so as to improve the efficiency of photovoltaic panel cleaning and ensure the reliability of each round of photovoltaic panel cleaning. On the other hand, the influence of visible and invisible pollutants on the surface of the photovoltaic panel is fully combined, and the multi-dimensional evaluation of the surface cleanliness of the photovoltaic panel is realized, and the coverage of the surface cleanliness analysis of the photovoltaic panel is expanded, thereby ensuring the reliability, authenticity, and persuasiveness of the analysis results of the surface cleanliness of the photovoltaic panel.

[0039] In an embodiment provided in the present application, the cleaning methods include air blowing cleaning, water washing cleaning and air blowing-water washing cleaning.

[0040] Specifically, in the blowing cleaning adopted in this embodiment, the drone is equipped with a blowing device (such as a hair dryer). Different from the existing water washing and vacuuming methods, the drone does not need to carry a water washing device after selecting the blowing method to clean the photovoltaic panels. It does not need to carry water and replace the blowing device of the drone with an air compressor, which can save the energy consumption and flight time of the drone and improve the cleaning efficiency. At the same time, the present application also improves the drone battery, and improves the original battery equipped with the drone into a built-in charging system to drive the operation of the air compressor. On the other hand, if it is a cleaning method combining blowing and water washing, On the basis of the above-mentioned blowing method, the water washing equipment required for the water washing method is equipped (which may include a water tank, a clean water nozzle, a detergent nozzle and a water pump, etc., and the specific water washing equipment is not limited). Although this method increases the flight burden of the drone and will relatively reduce the cleaning efficiency of the drone, from an overall perspective, it will still increase the cleaning efficiency of the photovoltaic panel and greatly improve the cleanliness of the surface of the photovoltaic panel, thereby improving the power generation efficiency of the photovoltaic panel; and in this embodiment, the above-mentioned cleaning method is selected according to the type of pollutant, and is diverse and changeable, making the cleaning of the photovoltaic panel more comprehensive.

[0041] In addition, in order to solve the problem that the drone is equipped with both a blowing device and a water washing device in the blowing-water washing cleaning method, thereby reducing the cleaning efficiency of the drone, the present application also provides another embodiment, in which the blowing device and the water washing device are respectively installed on different drones. When the blowing-water washing cleaning method is selected, the blowing drone is first used for cleaning, and then the water washing drone is dispatched to perform water washing cleaning at a predetermined position, thereby achieving the same purpose as above. The difference between this embodiment and the above embodiment is that this embodiment separates the equipment for the two different cleaning methods on different drones, that is, the blowing-water washing cleaning method is achieved by cooperating between two drones equipped with water washing equipment and blowing equipment.

[0042] This embodiment adopts different cleaning methods for different types of pollutants, and through the synergistic effect of one or more methods, the cleaning efficiency is improved while achieving the purpose of saving energy consumption.

[0043] Example 2 See also Figure 4 The embodiment of the present application provides a method for automatically cleaning a photovoltaic panel based on a drone, which is applied to the automatic cleaning system for photovoltaic panels as described above, and includes the following steps: Obtain the operating data of the photovoltaic panels and the geographical information and meteorological data of the area where the photovoltaic panels are located, and use them to plan the flight path of the drone; Monitoring the photovoltaic panel in real time according to the planned flight path to obtain a photovoltaic panel image; the photovoltaic panel image includes a first panel image and a second panel image; Analyzing and identifying the photovoltaic panel image to determine the type and location of contaminants; and also used to determine when to clean the photovoltaic panel; Selecting a cleaning method and a cleaning device according to the type of the pollutant and the cleaning time of the photovoltaic panel and performing a cleaning operation according to the location of the pollutant; Perform secondary inspection on the cleaned photovoltaic panel and judge the cleaning status of the photovoltaic panel. If there are still pollutants on the cleaned photovoltaic panel, select the cleaning method corresponding to the type of pollutants to perform the cleaning operation again, otherwise output a cleaning completion signal; Recording the cleaning tasks and cleaning processes of each photovoltaic panel cleaning using the drone according to the cleaning completion signal, and establishing an adaptive model based on the cleaning tasks and cleaning processes; Wherein, the photovoltaic panel image includes a first panel image and a second panel image; The adaptive model is used to adaptively adjust the relevant parameters of the pollutant cleaning device after each round of drone cleaning operation; Determining the type and location of contaminants at the remote control center includes: Pre-building a pollutant detection model in the remote control center; Inputting the photovoltaic panel image into the pollutant detection model for identification to obtain an identification result specifically includes: The first panel image is input into the pollutant detection model for recognition to obtain a first recognition result, specifically: Inputting the first panel image into the pollutant detection model to obtain an original standard image corresponding to the location of the first panel image; The first panel image is compared with the original standard image for similarity, and if the comparison result exceeds a set similarity threshold, the first panel image currently being compared is determined to be a suspected pollutant image; Performing edge detection on the suspected pollutant image to obtain the pollutant contour; Performing target detection based on the pollutant profile to obtain the type of the pollutant and outputting a first recognition result; The second panel image is input into the pollutant detection model for recognition to obtain a second recognition result, specifically: The second panel image is input into the pollutant detection model for identification. If the second panel image matches the light spot shadow image in the model, it is determined as a non-pollutant image, otherwise it is determined as a pollutant image; Further identification of pollutant images is performed, specifically: Calculate the pollutant attachment rate of the pollutant image. If the pollutant attachment rate is greater than the set standard pollutant attachment probability, the pollutant in the pollutant image is determined to be a non-dust particle pollutant; otherwise, it is determined to be a dust particle pollutant. The non-dust pollutant images determined to be non-dust particle pollutants are further identified as follows: Calculate the dirty color depth value of the non-dust pollutant image. If the dirty color depth value is greater than the set color depth reference value, the non-dust pollutant image is determined to be a heavy pollutant image, otherwise it is a light pollutant image. outputting a second recognition result; The first identification result and the second identification result are combined to determine the type and location of the contaminant.

[0044] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0045] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0046] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0047] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. An automatic photovoltaic panel cleaning system based on drones, characterized in that: include: Unmanned aerial vehicles, unmanned aerial vehicle-mounted pollutant cleaning devices, unmanned aerial vehicle-mounted pollutant detection devices, and remote control centers; The remote control center is used to obtain the operation data of the photovoltaic panels and the geographical information and meteorological data of the area where the photovoltaic panels are located, and use them to plan the flight path of the drone; The drone is used to monitor the photovoltaic panels in real time according to the planned flight path to obtain images of the photovoltaic panels; The remote control center is also used to analyze and identify the photovoltaic panel image to determine the type and location of the contaminant; and is also used to determine the cleaning time of the photovoltaic panel; The pollutant cleaning device is used to select a cleaning method and a cleaning device according to the type of the pollutant and the cleaning time of the photovoltaic panel and perform a cleaning operation according to the location of the pollutant; The pollutant detection device is used to perform a secondary inspection on the cleaned photovoltaic panel and determine the cleaning status of the photovoltaic panel. If pollutants still exist on the cleaned photovoltaic panel, a cleaning method corresponding to the type of pollutant is selected again to perform a cleaning operation. Otherwise, a cleaning completion signal is output; The remote control center is further used to record the cleaning tasks and cleaning processes of each photovoltaic panel cleaning performed by the drone according to the cleaning completion signal, and to establish an adaptive model based on the cleaning tasks and cleaning processes. Wherein, the photovoltaic panel image includes a first panel image and a second panel image; The adaptive model is used to adaptively adjust the relevant parameters of the pollutant cleaning device after each round of drone cleaning operation; Determining the type and location of contaminants at the remote control center includes: Pre-building a pollutant detection model in the remote control center; Inputting the photovoltaic panel image into the pollutant detection model for identification to obtain an identification result specifically includes: The first panel image is input into the pollutant detection model for recognition to obtain a first recognition result, specifically: Inputting the first panel image into the pollutant detection model to obtain an original standard image corresponding to the location of the first panel image; The first panel image is compared with the original standard image for similarity, and if the comparison result exceeds a set similarity threshold, the first panel image currently being compared is determined to be a suspected pollutant image; Performing edge detection on the suspected pollutant image to obtain the pollutant contour; Performing target detection based on the pollutant profile to obtain the type of the pollutant and outputting a first recognition result; The second panel image is input into the pollutant detection model for recognition to obtain a second recognition result, specifically: The second panel image is input into the pollutant detection model for identification. If the second panel image matches the light spot shadow image in the model, it is determined as a non-pollutant image, otherwise it is determined as a pollutant image; Further identification of pollutant images is performed, specifically: Calculate the pollutant attachment rate of the pollutant image. If the pollutant attachment rate is greater than the set standard pollutant attachment probability, the pollutant in the pollutant image is determined to be a non-dust particle pollutant; otherwise, it is determined to be a dust particle pollutant. The non-dust pollutant images determined to be non-dust particle pollutants are further identified as follows: Calculate the dirty color depth value of the non-dust pollutant image. If the dirty color depth value is greater than the set color depth reference value, the non-dust pollutant image is determined to be a heavy pollutant image, otherwise it is a light pollutant image. outputting a second recognition result; The first identification result and the second identification result are combined to determine the type and location of the contaminant.

2. The photovoltaic panel automatic cleaning system based on a drone according to claim 1, characterized in that: Plan the flight path of the drone, including the following: Establish a digital twin model based on the operating data of photovoltaic panels, and use it to monitor the operating status and performance changes of photovoltaic panels; Determine the abnormal time and location of the photovoltaic panel when the photovoltaic panel has abnormal operation according to the geographical information and meteorological data of the area where the photovoltaic panel is located; Construct a dynamic spatiotemporal map of the entire photovoltaic panel area based on the abnormal time and location of the photovoltaic panels; The point where the drone takes off relative to the photovoltaic panel is used as the origin to construct a position coordinate system in the dynamic space-time map; The abnormal position of the photovoltaic panel is used as the target position of the UAV flight, the optimal path of the UAV flight is calculated in the position coordinate system, and the optimal path is used as the final flight path of the UAV.

3. The photovoltaic panel automatic cleaning system based on a drone according to claim 2, characterized in that: Determine the abnormal location of the photovoltaic panel, including the following: Obtain historical photovoltaic data and structural data of photovoltaic panels; The photovoltaic panels are divided into multiple photovoltaic areas and several photovoltaic modules based on historical photovoltaic data, structural data and geographic information; Select a photovoltaic module and detect the power generation value of this photovoltaic module under the current light intensity; Calculate the difference between the power generation value and the power generation standard value, and record it as the power generation difference; The remote control center determines the difference between the power generation amount and the preset abnormal threshold value. If the power generation amount difference is greater than the preset abnormal threshold value, a power generation abnormality signal is generated and the photovoltaic module is marked as an abnormal module; otherwise, the next photovoltaic module is selected to detect the power generation value; Record the location information of each abnormal module on the photovoltaic panel and construct an abnormal location set based on it; Wherein, the abnormal position of the photovoltaic panel is selected and determined in the abnormal position set; The historical photovoltaic data includes historical light intensity, historical temperature and humidity, historical current data, historical voltage data and historical power generation efficiency; the structural data includes series structure, parallel structure, size data and arrangement method.

4. The photovoltaic panel automatic cleaning system based on a drone according to claim 1, characterized in that: Acquiring the photovoltaic panel image, comprising: The UAV determines the abnormal module on the photovoltaic panel according to the abnormal position of the photovoltaic panel, and collects images of the abnormal module to obtain the first panel image.

5. The photovoltaic panel automatic cleaning system based on a drone according to claim 1, characterized in that: Acquiring the photovoltaic panel image also includes: The drone collects images of the entire area of ​​the photovoltaic panel and cuts the collected images to obtain a number of cut images of equal size; Extracting color features of the cut image and performing feature screening to obtain a feature image; Performing pollutant identification on the characteristic image and selecting an image containing pollutants as a second panel image; Among them, feature screening is performed to obtain feature images, specifically: The color feature is compared with the standard feature when the photovoltaic panel is not contaminated and the characteristic color difference is obtained. If the characteristic color difference is greater than the set color difference threshold, the cutting image represented by this color feature is selected as the characteristic image, otherwise the next round of feature comparison is performed.

6. The photovoltaic panel automatic cleaning system based on a drone according to claim 1, characterized in that: The pollutant detection model is pre-built, including: Obtain model training data, including image samples of photovoltaic panels and labels corresponding to the image samples; Inputting the model training data into a deep learning model to extract features of image samples; Make predictions based on the extracted features to output prediction results; Generate loss values ​​based on prediction results and labels; Training the deep learning model according to the loss value until an end condition of the model training is met, and determining the deep learning model that has finished training as the pollutant detection model; The image samples include images of photovoltaic panels with pollutants, images of photovoltaic panels without pollutants, and images of photovoltaic panels suspected of being polluted; The labels corresponding to the image samples include whether there is a pollutant, the type of the pollutant, the shape of the pollutant, the size of the pollutant, and the location of the pollutant.

7. The photovoltaic panel automatic cleaning system based on a drone according to claim 1, characterized in that: The cleaning methods include air blowing cleaning, water washing cleaning and air blowing-water washing cleaning.

8. The photovoltaic panel automatic cleaning system based on a drone according to claim 1, characterized in that: Determine the cleaning time of the photovoltaic panels, specifically: Based on the operation data of the photovoltaic panel, the power generation efficiency and current power generation efficiency at each image acquisition time point are obtained, and the power generation evaluation coefficient of the photovoltaic panel is analyzed; Based on the geographic information and meteorological data of the area where the photovoltaic panels are located, the current light intensity and current irradiance at the current time point are obtained, and the environmental assessment coefficient of the photovoltaic panels is analyzed; Analyze the power generation assessment coefficient and environmental assessment coefficient to determine the cleaning time of photovoltaic panels; The contamination area is calculated based on the type of contamination and used to determine the duration of PV panel cleaning.

9. The photovoltaic panel automatic cleaning system based on drone according to claim 1, characterized in that: Establishing the adaptive model comprises: Inspect the surface of the photovoltaic panel that has completed the cleaning task to obtain cleanliness parameters; assessing the initial surface cleanliness of the photovoltaic panel according to the cleanliness parameter; extracting cleaning interference parameters related to the initial surface cleanliness; Setting a defect cleaning assessment factor and an environmental load cleaning assessment factor of the photovoltaic panel according to the cleaning interference parameter; The surface cleanliness of the photovoltaic panel is evaluated again based on the defect cleaning assessment factor and the environmental load cleaning assessment factor, and the final comprehensive surface cleanliness is obtained by statistics; Conduct qualification evaluation on the cleaning process of completing the cleaning task and obtain the task evaluation results; Establishing the adaptive model based on the comprehensive surface cleanliness and the task evaluation result; The cleanliness parameters include the surface image, the ambient temperature of each monitoring period, the scanned temperature distribution diagram, and the reflectivity in each area of ​​the photovoltaic panel; the clean interference parameters include structural interference parameters, chromatic interference parameters, and environmental interference parameters.

10. A method for automatically cleaning a photovoltaic panel based on a drone, applied to the automatic cleaning system for a photovoltaic panel according to any one of claims 1 to 9, characterized in that: The steps include: Obtain the operating data of the photovoltaic panels and the geographical information and meteorological data of the area where the photovoltaic panels are located, and use them to plan the flight path of the drone; Monitoring the photovoltaic panel in real time according to the planned flight path to obtain a photovoltaic panel image; the photovoltaic panel image includes a first panel image and a second panel image; Analyzing and identifying the photovoltaic panel image to determine the type and location of contaminants; and also used to determine when to clean the photovoltaic panel; Selecting a cleaning method and a cleaning device according to the type of the pollutant and the cleaning time of the photovoltaic panel and performing a cleaning operation according to the location of the pollutant; Perform secondary inspection on the cleaned photovoltaic panel and judge the cleaning status of the photovoltaic panel. If there are still pollutants on the cleaned photovoltaic panel, select the cleaning method corresponding to the type of pollutants to perform the cleaning operation again, otherwise output a cleaning completion signal; Recording the cleaning tasks and cleaning processes of each photovoltaic panel cleaning using the drone according to the cleaning completion signal, and establishing an adaptive model based on the cleaning tasks and cleaning processes; Wherein, the photovoltaic panel image includes a first panel image and a second panel image; The adaptive model is used to adaptively adjust the relevant parameters of the pollutant cleaning device after each round of drone cleaning operation; Determining the type and location of contaminants at the remote control center includes: Pre-building a pollutant detection model in the remote control center; Inputting the photovoltaic panel image into the pollutant detection model for identification to obtain an identification result specifically includes: The first panel image is input into the pollutant detection model for recognition to obtain a first recognition result, specifically: Inputting the first panel image into the pollutant detection model to obtain an original standard image corresponding to the location of the first panel image; The first panel image is compared with the original standard image for similarity, and if the comparison result exceeds a set similarity threshold, the first panel image currently being compared is determined to be a suspected pollutant image; Performing edge detection on the suspected pollutant image to obtain the pollutant contour; Performing target detection based on the pollutant profile to obtain the type of the pollutant and outputting a first recognition result; The second panel image is input into the pollutant detection model for recognition to obtain a second recognition result, specifically: The second panel image is input into the pollutant detection model for identification. If the second panel image matches the light spot shadow image in the model, it is determined as a non-pollutant image, otherwise it is determined as a pollutant image; Further identification of pollutant images is performed, specifically: Calculate the pollutant attachment rate of the pollutant image. If the pollutant attachment rate is greater than the set standard pollutant attachment probability, the pollutant in the pollutant image is determined to be a non-dust particle pollutant; otherwise, it is determined to be a dust particle pollutant. The non-dust pollutant images determined to be non-dust particle pollutants are further identified as follows: Calculate the dirty color depth value of the non-dust pollutant image. If the dirty color depth value is greater than the set color depth reference value, the non-dust pollutant image is determined to be a heavy pollutant image, otherwise it is a light pollutant image. outputting a second recognition result; The first identification result and the second identification result are combined to determine the type and location of the contaminant.

Images