Photovoltaic panel cleanliness detection method and system

Abstract

The present invention relates to the field of intelligent detection technology, and specifically discloses a photovoltaic panel cleanliness detection method and system, the method comprising: determining the overall cleanliness of a photovoltaic panel based on the brightness value of each remote sensing image pixel in a multispectral remote sensing image of the photovoltaic panel in a red band and a near-infrared band, a spectral curve, and an overall cleanliness quantification model; obtaining a numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period; obtaining a verification result of the overall cleanliness quantification model based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods; obtaining the overall cleanliness of the photovoltaic panel to be tested at the latest moment based on the local cleanliness and verification results of corresponding position points of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested; combining the multispectral remote sensing image of the photovoltaic panel with a neural network model to achieve accurate detection of the cleanliness of the photovoltaic panel.

Description

Technical Field

[0001] The present invention relates to the field of intelligent detection technology, and in particular to a photovoltaic panel cleanliness detection method and system. Background Art

[0002] At present, the surface of photovoltaic panels will be affected by dust, grease, pollutants, etc., which will cause the surface of photovoltaic panels to become unclean. These dirt will affect the reflection and absorption of sunlight on the surface of photovoltaic panels, and then affect the conductivity and photoelectric conversion efficiency of photovoltaic panels. Generally speaking, cleaned photovoltaic panels can improve the photoelectric conversion efficiency to a certain extent. In addition, the dirt on photovoltaic panels may also cause short circuits or open circuits in some areas of the photovoltaic panel surface, which will affect the power generation of photovoltaic panels. The cleanliness of photovoltaic panels will also affect the life of photovoltaic panels. If the surface of photovoltaic panels is in a dirty state for a long time, it will cause the material of photovoltaic panels to age and break, and may even cause safety problems such as the glass of photovoltaic panels to break. Therefore, it is very important to determine the cleanliness of photovoltaic panels in time so as to maintain the cleanliness of photovoltaic panels on time.

[0003] However, among the existing intelligent detection methods for the cleanliness of photovoltaic panels, the more common one is to use image analysis technology to analyze the monitoring images of photovoltaic panels. However, since the working environment of photovoltaic panels is outdoors, their appearance in the monitoring images will be greatly affected by weather factors and natural light images. Therefore, if the monitoring images of photovoltaic panels are analyzed using image analysis technology to detect the cleanliness of photovoltaic panels, it is necessary to simultaneously obtain the weather data of the location of the photovoltaic panels and the reflection of light by the photovoltaic panels, etc., in order to shield the visual errors caused by them in the monitoring images of the photovoltaic panels. For example, the Chinese invention patent with the publication number "CN117895899A" and the patent name "A photovoltaic panel cleanliness detection method and system" discloses a photovoltaic panel cleanliness detection method and system, which involves the field of cleanliness detection technology. The method includes: dividing the photovoltaic power station into m areas, each area includes at least one photovoltaic panel, setting a fixed shooting point, collecting images of the photovoltaic panels in each area, and obtaining the photovoltaic panel images of each area; pre-processing the obtained photovoltaic panel images to obtain the cleanliness feature information in the photovoltaic panel images; obtaining weather data of all areas, measuring the reflectivity of the photovoltaic panels in all areas; calculating the weather change data and reflectivity change data of all areas; constructing a cleanliness detection model based on the weather change data, reflectivity change data and cleanliness feature information; based on the cleanliness detection model, detecting the cleanliness of the photovoltaic panel to obtain the cleanliness detection result. Analyze the weather changes and the reflection of the photovoltaic panel surface to further reduce the errors existing in the cleanliness detection of photovoltaic panels using machine vision. Although this patent uses the photovoltaic panel image to obtain its cleanliness detection result, it needs to obtain the weather data of all areas and the reflectivity of the photovoltaic panels in all areas, that is, the detection procedure is complicated and the detection cost is high.

[0004] Therefore, the present invention proposes a photovoltaic panel cleanliness detection method and system. Summary of the invention

[0005] The present invention provides a photovoltaic panel cleanliness detection method and system, which are used to combine the multispectral remote sensing image of the photovoltaic panel with the neural network model to realize accurate detection of the cleanliness of the photovoltaic panel. There is no need to set up additional detection equipment to obtain detection data (such as weather change data and reflectivity data). Only the existing photovoltaic panel power generation efficiency data needs to be applied to the model verification process, that is, only the multispectral remote sensing image of the photovoltaic panel and the overall cleanliness quantification model are needed to obtain accurate photovoltaic panel cleanliness detection results, which simplifies the detection procedure, reduces the detection cost, reduces the data source items required for the detection process, and reduces the probability of detection result errors due to errors in the data source items.

[0006] The present invention provides a photovoltaic panel cleanliness detection method, comprising:

[0007] S1: Based on the brightness value of each remote sensing image pixel in the red band and near-infrared band of the photovoltaic panel at each historical moment in the historical period and the corresponding spectral curve, the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel is calculated;

[0008] S2: Based on the local cleanliness and overall cleanliness quantification model of all remote sensing image pixels in the multispectral remote sensing image at the corresponding position points in the corresponding photovoltaic panel, the overall cleanliness of the photovoltaic panel is determined;

[0009] S3: Align the overall cleanliness of the photovoltaic panel and the power generation efficiency of the photovoltaic panel at all the same moments in each sub-historical period within the historical period and extract the numerical relationship, so as to obtain the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period;

[0010] S4: The overall cleanliness quantification model is constructed based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panels in all sub-historical periods within the historical period, and the verification results are obtained;

[0011] S5: Based on the local cleanliness and verification results of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is obtained.

[0012] Photovoltaic panel cleanliness detection method, S1: Based on the brightness value of each remote sensing image pixel in the red band and the near-infrared band of the photovoltaic panel in the multispectral remote sensing image at each historical moment in the historical period and the corresponding spectral curve, the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel is calculated, including:

[0013] S101: Obtain the brightness value of each remote sensing image pixel in the red band and the near infrared band of the photovoltaic panel in the multispectral remote sensing image at each historical moment in the historical period and the corresponding spectral curve;

[0014] S102: based on the spectral curve of each remote sensing image pixel in the multispectral remote sensing image of the photovoltaic panel at each historical moment in the historical period, calculating the reflectivity balance coefficient in the near infrared band and the reflectivity balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image;

[0015] S103: Based on the brightness values ​​of each remote sensing image pixel in the red band and the near-infrared band of the photovoltaic panel in the multispectral remote sensing image at each historical moment in the historical period and the reflectivity balance coefficient of the corresponding remote sensing image pixel position in the near-infrared band and the reflectivity balance coefficient in the red band, calculate the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel.

[0016] Preferably, the photovoltaic panel cleanliness detection method, S102: based on the spectral curve of each remote sensing image pixel in the multispectral remote sensing image of the photovoltaic panel at each historical moment in the historical period, calculate the reflectivity balance coefficient in the near infrared band and the reflectivity balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image, including:

[0017] The spectral curves of the remote sensing image pixels at the same position in the multispectral remote sensing images of the photovoltaic panel at all historical moments in the historical period are time-series aligned to obtain the alignment curve combination of each remote sensing image pixel position;

[0018] Determine the average of all ordinate values ​​corresponding to each abscissa value in the alignment curve combination of each remote sensing image pixel position as the first average reflectivity of each abscissa value in the alignment curve combination of each remote sensing image pixel position;

[0019] The average value of all first average reflectances in the alignment curve combination of all horizontal coordinate values ​​belonging to the near-infrared band at each remote sensing image pixel position is regarded as the second average reflectance of the corresponding remote sensing image pixel position in the near-infrared band. At the same time, the average value of all first average reflectances in the alignment curve combination of all horizontal coordinate values ​​belonging to the red band at each remote sensing image pixel position is regarded as the second average reflectance of the corresponding remote sensing image pixel position in the red band.

[0020] Based on all the second average reflectances of all remote sensing image pixel positions in the near-infrared band and all the second average reflectances in the red band, the reflectance balance coefficient in the near-infrared band and the reflectance balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image are calculated.

[0021] Preferably, the photovoltaic panel cleanliness detection method calculates the reflectivity balance coefficient in the near infrared band and the reflectivity balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image based on all second average reflectivities in the near infrared band and all second average reflectivities in the red band of all remote sensing image pixel positions, including:

[0022] The average of all the second average reflectances of all the remote sensing image pixel positions in the near-infrared band is used as the reference reflectance of the near-infrared band. At the same time, the average of all the second average reflectances of all the remote sensing image pixel positions in the red band is used as the reference reflectance of the red band.

[0023] The ratio of the difference between the second average reflectivity of each remote sensing image pixel in the near-infrared band and the reference reflectivity of the near-infrared band to the second average emissivity of the corresponding remote sensing image pixel in the near-infrared band is taken as the reflectivity balance coefficient of the corresponding remote sensing image pixel in the near-infrared band. At the same time, the ratio of the difference between the second average reflectivity of each remote sensing image pixel in the red band and the reference reflectivity of the red band to the second average emissivity of the corresponding remote sensing image pixel in the red band is taken as the reflectivity balance coefficient of the corresponding remote sensing image pixel in the red band.

[0024] Preferably, the photovoltaic panel cleanliness detection method, S103: based on the brightness value of each remote sensing image pixel in the multispectral remote sensing image of the photovoltaic panel at each historical moment in the historical period in the red band and the near-infrared band and the reflectivity balance coefficient of the corresponding remote sensing image pixel position in the near-infrared band and the reflectivity balance coefficient in the red band, calculate the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel, including:

[0025]

[0026] Where Cleanliness is the local cleanliness of a single remote sensing image pixel in a multispectral remote sensing image at the corresponding position in the corresponding photovoltaic panel, NIR is the brightness value of a single remote sensing image pixel in a multispectral remote sensing image in the near infrared band, R is the brightness value of a single remote sensing image pixel in a multispectral remote sensing image in the red band, and Norm Cleanliness It is the ratio of the brightness value in the near-infrared band to the brightness value in the red band of the remote sensing image pixel at the same position as the currently calculated remote sensing image pixel in the standard dust-free photovoltaic panel under standard lighting conditions. α1 is the reflectivity balance coefficient of the remote sensing image pixel position corresponding to the currently calculated remote sensing image pixel in the near-infrared band. α2 is the reflectivity balance coefficient of the remote sensing image pixel position corresponding to the currently calculated remote sensing image pixel in the red band.

[0027] Preferably, the photovoltaic panel cleanliness detection method, S2: based on the local cleanliness and overall cleanliness quantification model of all remote sensing image pixels in the multispectral remote sensing image at the corresponding position points in the corresponding photovoltaic panel, determines the overall cleanliness of the photovoltaic panel, including:

[0028] The local cleanliness of all remote sensing image pixels in the multispectral remote sensing image at the corresponding position points in the corresponding photovoltaic panel is input into the overall cleanliness quantification model to obtain the overall cleanliness of the photovoltaic panel.

[0029] Preferably, the photovoltaic panel cleanliness detection method, S4: based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period, the overall cleanliness quantification model is obtained to obtain a verification result, including:

[0030] Determine whether the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period meets the consistency judgment principle. If so, the overall cleanliness quantification model meets the consistency judgment principle as a verification result;

[0031] Otherwise, the overall cleanliness quantification model will be regarded as not meeting the consistency judgment principle as a verification result.

[0032] Preferably, the photovoltaic panel cleanliness detection method, S5: based on the local cleanliness and verification results of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position point in the corresponding photovoltaic panel, obtain the overall cleanliness of the photovoltaic panel to be tested at the latest moment, including:

[0033] When the verification result shows that the overall cleanliness quantification model satisfies the consistency judgment principle, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is determined based on the local cleanliness of all remote sensing image pixels at the corresponding position points in the corresponding photovoltaic panel in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment and the overall cleanliness quantification model;

[0034] When the verification result shows that the overall cleanliness quantification model does not meet the consistency judgment principle, a new overall cleanliness quantification model is obtained based on the photovoltaic panel power generation efficiency of a large number of new photovoltaic panels at each moment in the new historical period and the local cleanliness of all remote sensing image pixels in the corresponding multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel, as well as the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period;

[0035] Based on the local cleanliness of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel and a new overall cleanliness quantification model, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is determined.

[0036] Preferably, the photovoltaic panel cleanliness detection method obtains a new overall cleanliness quantification model based on the photovoltaic panel power generation efficiency of a large number of new photovoltaic panels at each moment in a new historical period and the local cleanliness of all remote sensing image pixels in the corresponding multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel, and the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period, including:

[0037] Based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods, a standard numerical relationship between the overall cleanliness and power generation efficiency is obtained;

[0038] Substitute the photovoltaic panel power generation efficiency of all new photovoltaic panels at each moment in the new historical period into the standard numerical relationship between overall cleanliness and power generation efficiency to obtain the overall cleanliness of each new photovoltaic panel at each moment in the new historical period;

[0039] The local cleanliness and the corresponding overall cleanliness of all remote sensing image pixels in the multispectral remote sensing images of all new photovoltaic panels at each moment in the new historical period are used as model input and model output for neural network model training to obtain a new overall cleanliness quantification model.

[0040] The present invention provides a photovoltaic panel cleanliness detection system, which is used to perform the photovoltaic panel cleanliness detection method described in any one of embodiments 1 to 9, comprising:

[0041] A local cleanliness calculation module is used to calculate the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at a corresponding position point in the corresponding photovoltaic panel based on the brightness value of each remote sensing image pixel in the red band and the near-infrared band and the corresponding spectral curve of each remote sensing image pixel in the multispectral remote sensing image at each historical moment in the historical period;

[0042] An overall cleanliness calculation module is used to determine the overall cleanliness of the photovoltaic panel based on the local cleanliness and overall cleanliness quantification model of all remote sensing image pixels in the multispectral remote sensing image at corresponding positions in the corresponding photovoltaic panel;

[0043] A numerical relationship extraction module is used to align the overall cleanliness of the photovoltaic panel and the photovoltaic panel power generation efficiency at all the same moments in each sub-historical period within the historical period and extract the numerical relationship, so as to obtain the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period;

[0044] A quantitative model verification module is used to obtain a verification result for the overall cleanliness quantification model based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period;

[0045] The overall cleanliness detection module is used to obtain the overall cleanliness of the photovoltaic panel to be tested at the latest moment based on the local cleanliness and verification results of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel.

[0046] The beneficial effects of the present invention compared to the prior art are as follows: the multispectral remote sensing images of photovoltaic panels are combined with the neural network model to achieve accurate detection of the cleanliness of photovoltaic panels. There is no need to set up additional detection equipment to obtain detection data (such as weather change data and reflectivity data). Only the existing photovoltaic panel power generation efficiency data needs to be applied to the model verification process, that is, only the multispectral remote sensing images of the photovoltaic panels and the overall cleanliness quantification model are needed to obtain accurate photovoltaic panel cleanliness detection results, which simplifies the detection procedure, reduces the detection cost, reduces the data source items required for the detection process, and reduces the probability of detection result errors due to errors in the data source items.

[0047] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in this application document.

[0048] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0050] Figure 1 is a flow chart of a photovoltaic panel cleanliness detection method in an embodiment of the present invention;

[0051] Figure 2 is a flowchart of a specific execution method of step S1 in an embodiment of the present invention;

[0052] Figure 3 Schematic diagram of the internal functional modules for photovoltaic panel cleanliness detection in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0054] Embodiment 1:

[0055] The present invention provides a photovoltaic panel cleanliness detection method, referring to Figure 1 ,include:

[0056] S1: Based on the brightness value of each remote sensing image pixel in the red band and near-infrared band of the photovoltaic panel at each historical moment in the historical period and the corresponding spectral curve, the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel is calculated;

[0057] S2: Based on the local cleanliness and overall cleanliness quantification model of all remote sensing image pixels in the multispectral remote sensing image at the corresponding position points in the corresponding photovoltaic panel, the overall cleanliness of the photovoltaic panel is determined;

[0058] S3: Align the overall cleanliness of the photovoltaic panel and the power generation efficiency of the photovoltaic panel at all the same moments in each sub-historical period within the historical period and extract the numerical relationship, so as to obtain the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period;

[0059] S4: The overall cleanliness quantification model is constructed based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panels in all sub-historical periods within the historical period, and the verification results are obtained;

[0060] S5: Based on the local cleanliness and verification results of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is obtained.

[0061] In this embodiment, the historical period is a period of time before a preset current moment.

[0062] In this embodiment, the historical moments are moments evenly distributed within the historical period, for example, moments within the historical period with an interval of 1 hour between adjacent moments.

[0063] In this embodiment, the remote sensing image pixel is the basic unit of the remote sensing image, which represents the ground object information contained in a pixel. This concept originates from the field of digital image processing, which divides the image into many small squares, each of which is called a pixel. Each pixel contains quantitative information about the ground object, such as intensity, color, position, etc. Through the combination of pixels, a complete remote sensing image can be formed;

[0064] A pixel can be understood as the smallest resolvable area of ​​a ground object, which contains all the information in the area. Therefore, for remote sensing images, the quality and quantity of pixels represent the accuracy and completeness of the image.

[0065] In this embodiment, the red band is a wavelength range from 620 nm to 780 nm.

[0066] In this embodiment, the near infrared band is a wavelength range of 780 to 1500 nm.

[0067] In this embodiment, the brightness values ​​of the remote sensing image pixels in the red band and the near infrared band are the DN values ​​of the remote sensing image pixels in the red band and the near infrared band.

[0068] In this embodiment, the spectral curve of the remote sensing image pixel is a curve extracted from the multispectral remote sensing image, with wavelength as the horizontal coordinate and reflectivity as the vertical coordinate, which represents the corresponding change of reflectivity of the corresponding remote sensing image pixel position with wavelength change.

[0069] In this embodiment, the remote sensing image pixel is the smallest information unit in the remote sensing image.

[0070] In this embodiment, the local cleanliness refers to the value of the cleanliness level of a single remote sensing image pixel at a corresponding position in the photovoltaic panel.

[0071] In this embodiment, the overall cleanliness quantification model is a pre-trained model that can output the overall cleanliness of the corresponding photovoltaic panel at a corresponding moment based on the local cleanliness of all remote sensing image pixels in the input multispectral remote sensing image at corresponding position points in the corresponding photovoltaic panel;

[0072] Among them, during the training process, the local cleanliness of all remote sensing image pixels in the multispectral remote sensing images of a large number of photovoltaic panels at the corresponding position points in the corresponding photovoltaic panels at a single moment and the overall cleanliness of the corresponding photovoltaic panels at the corresponding moment are manually calculated and used as model input and model output respectively to train the neural network model to obtain the model.

[0073] In this embodiment, the overall cleanliness of the photovoltaic panel represents a numerical value of the overall cleanliness of the photovoltaic panel, and this numerical value has a fixed numerical relationship with the power generation efficiency of the photovoltaic panel at a corresponding moment.

[0074] In this embodiment, the sub-historical period is a plurality of periods divided from the historical period.

[0075] In this embodiment, the photovoltaic panel power generation efficiency is the power generation efficiency of the photovoltaic panel at a corresponding moment.

[0076] In this embodiment, the photovoltaic panel to be tested is a photovoltaic panel whose overall cleanliness needs to be tested using the photovoltaic panel cleanliness testing method in this embodiment.

[0077] In this embodiment, the latest moment can be understood as the current moment, or the latest moment among all the acquisition moments of the multispectral remote sensing image of the photovoltaic panel to be tested.

[0078] In this embodiment, the method for obtaining the local cleanliness of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel is consistent with the method of obtaining "the local cleanliness of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel at the corresponding position points in the corresponding photovoltaic panel" in steps S1 and S2.

[0079] The beneficial effects of the above technology are: combining the multispectral remote sensing images of photovoltaic panels with the neural network model to achieve accurate detection of the cleanliness of photovoltaic panels. There is no need to set up additional detection equipment to obtain detection data (such as weather change data and reflectivity data). Only the existing photovoltaic panel power generation efficiency data is required to be applied to the model verification process, that is, only the multispectral remote sensing images of the photovoltaic panels and the overall cleanliness quantification model are needed to obtain accurate photovoltaic panel cleanliness detection results, which simplifies the detection procedure, reduces the detection cost, reduces the data source items required for the detection process, and reduces the probability of detection result errors due to errors in the data source items.

[0080] Embodiment 2:

[0081] On the basis of Example 1, a photovoltaic panel cleanliness detection method, S1: based on the brightness value of each remote sensing image pixel in the multispectral remote sensing image of the photovoltaic panel at each historical moment in the historical period in the red band and the near-infrared band and the corresponding spectral curve, calculate the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel, reference Figure 2 ,include:

[0082] S101: Obtain the brightness value of each remote sensing image pixel in the red band and the near infrared band of the photovoltaic panel in the multispectral remote sensing image at each historical moment in the historical period and the corresponding spectral curve;

[0083] S102: based on the spectral curve of each remote sensing image pixel in the multispectral remote sensing image of the photovoltaic panel at each historical moment in the historical period, calculating the reflectivity balance coefficient in the near infrared band and the reflectivity balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image;

[0084] S103: Based on the brightness values ​​of each remote sensing image pixel in the red band and the near-infrared band of the photovoltaic panel in the multispectral remote sensing image at each historical moment in the historical period and the reflectivity balance coefficient of the corresponding remote sensing image pixel position in the near-infrared band and the reflectivity balance coefficient in the red band, calculate the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel.

[0085] In this embodiment, the remote sensing image pixel position is the position of the remote sensing image pixel in the multispectral remote sensing image.

[0086] In this embodiment, the difference in reflectivity of different remote sensing image pixels in the near infrared band (or red band) in the multispectral remote sensing image causes an error between the brightness value (i.e., DN value) of the remote sensing image pixel corresponding to the near infrared band (or red band) and the actual appearance characteristics of the photovoltaic panel at that position, and the reflectivity balance coefficient of the remote sensing image pixel position in the near infrared band (or red band) is a coefficient used to balance the error;

[0087] Satellites are the main platform for multispectral remote sensing. These satellites are equipped with various sensors and imaging equipment, which can sense the electromagnetic radiation information of multiple bands such as visible light, near infrared, thermal infrared, microwave, etc. on the earth's surface. After image processing and analysis, these information forms our multispectral remote sensing images. In addition, multispectral remote sensing images can also be obtained using image data obtained from other platforms such as ground base stations, aircraft, and drones. These image data can also be converted into multispectral remote sensing images through certain image processing and analysis methods.

[0088] In this embodiment, the application system of the photovoltaic panel cleanliness detection method is a device or system specifically used to detect the cleanliness of photovoltaic panels. The system is usually composed of sensors, controllers, data centers and other parts.

[0089] The cleanliness sensor is the core component of the system. It can determine the cleanliness of the photovoltaic panel by detecting the intensity of scattered light after the light hits the surface of the photovoltaic panel. Specifically, the sensor collects light data on the surface of the photovoltaic panel and then analyzes this data to determine the cleanliness of the photovoltaic panel.

[0090] The controller is responsible for receiving and processing the data from the sensor. The controller communicates with the computer to further process and analyze the data collected by the sensor to obtain the cleanliness results of the photovoltaic panel.

[0091] The beneficial effect of the above technology is: by obtaining the brightness value of each remote sensing image pixel in the red band and the near-infrared band in the multispectral remote sensing image based on the photovoltaic panel at each historical moment in the historical period and the reflectivity balance coefficient of the corresponding remote sensing image pixel position in the near-infrared band and the reflectivity balance coefficient in the red band, the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel can be calculated.

[0092] Embodiment 3:

[0093] Based on Example 2, a photovoltaic panel cleanliness detection method, S102: based on the spectral curve of each remote sensing image pixel in the multispectral remote sensing image of the photovoltaic panel at each historical moment in the historical period, calculate the reflectivity balance coefficient in the near infrared band and the reflectivity balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image, including:

[0094] The spectral curves of the remote sensing image pixels at the same position in the multispectral remote sensing images of the photovoltaic panel at all historical moments in the historical period are time-series aligned to obtain the alignment curve combination of each remote sensing image pixel position;

[0095] Determine the average of all ordinate values ​​corresponding to each abscissa value in the alignment curve combination of each remote sensing image pixel position as the first average reflectivity of each abscissa value in the alignment curve combination of each remote sensing image pixel position;

[0096] The average value of all first average reflectances in the alignment curve combination of all horizontal coordinate values ​​belonging to the near-infrared band at each remote sensing image pixel position is regarded as the second average reflectance of the corresponding remote sensing image pixel position in the near-infrared band. At the same time, the average value of all first average reflectances in the alignment curve combination of all horizontal coordinate values ​​belonging to the red band at each remote sensing image pixel position is regarded as the second average reflectance of the corresponding remote sensing image pixel position in the red band.

[0097] Based on all the second average reflectances of all remote sensing image pixel positions in the near-infrared band and all the second average reflectances in the red band, the reflectance balance coefficient in the near-infrared band and the reflectance balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image are calculated.

[0098] The beneficial effects of the above technology are: by time-series alignment of the spectral curves of the remote sensing image pixels at the same position in the multispectral remote sensing images of the photovoltaic panels at all historical moments in the historical period, averaging the vertical coordinates, and further averaging the vertical coordinate values ​​within the band range, the reflectivity balance coefficient in the near-infrared band and the reflectivity balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image are calculated.

[0099] Embodiment 4:

[0100] On the basis of Example 3, the photovoltaic panel cleanliness detection method calculates the reflectivity balance coefficient in the near-infrared band and the reflectivity balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image based on all second average reflectivities in the near-infrared band and all second average reflectivities in the red band of all remote sensing image pixel positions, including:

[0101] The average of all the second average reflectances of all the remote sensing image pixel positions in the near-infrared band is used as the reference reflectance of the near-infrared band. At the same time, the average of all the second average reflectances of all the remote sensing image pixel positions in the red band is used as the reference reflectance of the red band.

[0102] The ratio of the difference between the second average reflectivity of each remote sensing image pixel in the near-infrared band and the reference reflectivity of the near-infrared band to the second average emissivity of the corresponding remote sensing image pixel in the near-infrared band is taken as the reflectivity balance coefficient of the corresponding remote sensing image pixel in the near-infrared band. At the same time, the ratio of the difference between the second average reflectivity of each remote sensing image pixel in the red band and the reference reflectivity of the red band to the second average emissivity of the corresponding remote sensing image pixel in the red band is taken as the reflectivity balance coefficient of the corresponding remote sensing image pixel in the red band.

[0103] The beneficial effect of the above technology is: by calculating the degree of deviation between the reference reflectivity of the near-infrared band (red band) and the second average emissivity of the remote sensing image pixel position in the near-infrared band (red band), the reflectivity balance coefficient of the remote sensing image pixel position in the near-infrared band (red band) is accurately calculated.

[0104] Embodiment 5:

[0105] On the basis of Example 2, a photovoltaic panel cleanliness detection method, S103: based on the brightness value of each remote sensing image pixel in the multispectral remote sensing image of the photovoltaic panel at each historical moment in the historical period in the red band and the near-infrared band and the reflectivity balance coefficient of the corresponding remote sensing image pixel position in the near-infrared band and the reflectivity balance coefficient in the red band, calculate the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel, including:

[0106]

[0107] Where Cleanliness is the local cleanliness of a single remote sensing image pixel in a multispectral remote sensing image at the corresponding position in the corresponding photovoltaic panel, NIR is the brightness value of a single remote sensing image pixel in a multispectral remote sensing image in the near infrared band, R is the brightness value of a single remote sensing image pixel in a multispectral remote sensing image in the red band, and Norm Cleanliness It is the ratio of the brightness value in the near-infrared band to the brightness value in the red band of the remote sensing image pixel at the same position as the currently calculated remote sensing image pixel in the standard dust-free photovoltaic panel under standard lighting conditions. α1 is the reflectivity balance coefficient of the remote sensing image pixel position corresponding to the currently calculated remote sensing image pixel in the near-infrared band. α2 is the reflectivity balance coefficient of the remote sensing image pixel position corresponding to the currently calculated remote sensing image pixel in the red band.

[0108] In this embodiment, the standard dust-free photovoltaic panel is a photovoltaic panel that meets the preset dust-free requirements.

[0109] In this embodiment, the standard lighting conditions are preset, for example, the light source should have a constant color temperature and spectral distribution to ensure that each part of all samples can receive the same light intensity, and it can be a continuously emitting light box, fluorescent lamp or artificial light source.

[0110] The beneficial effect of the above technology is: based on the above formula, the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel is calculated.

[0111] Embodiment 6:

[0112] Based on Example 1, a photovoltaic panel cleanliness detection method, S2: based on a quantification model of local cleanliness and overall cleanliness of all remote sensing image pixels in a multispectral remote sensing image at corresponding positions in a corresponding photovoltaic panel, determines the overall cleanliness of the photovoltaic panel, including:

[0113] The local cleanliness of all remote sensing image pixels in the multispectral remote sensing image at the corresponding position points in the corresponding photovoltaic panel is input into the overall cleanliness quantification model to obtain the overall cleanliness of the photovoltaic panel.

[0114] The beneficial effects of the above technology are: using the overall cleanliness quantification model to achieve preliminary acquisition of the overall cleanliness of the photovoltaic panel, providing data for the subsequent model verification process.

[0115] Embodiment 7:

[0116] Based on Example 1, the photovoltaic panel cleanliness detection method, S4: based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period, the overall cleanliness quantification model is obtained to obtain verification results, including:

[0117] Determine whether the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period meets the consistency judgment principle. If so, the overall cleanliness quantification model meets the consistency judgment principle as a verification result;

[0118] Otherwise, the overall cleanliness quantification model will be regarded as not meeting the consistency judgment principle as a verification result.

[0119] In this embodiment, judging whether the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period meets the consistency judgment principle includes:

[0120] In the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period, the overall cleanliness is represented by x, and the power generation efficiency is represented by a formula containing the x parameter;

[0121] The ratio of the difference between the power generation efficiencies expressed by the formula containing the x parameter in each two numerical relationships to the mean value is regarded as the deviation between the corresponding two numerical relationships;

[0122] Determine whether the maximum deviation in the numerical relationship of all sub-historical time periods does not exceed a preset value (the preset deviation threshold value, whose value must be less than 1). If so, it is determined that the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical time periods within the historical time period meets the consistency judgment principle. Otherwise, it is determined that the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical time periods within the historical time period does not meet the consistency judgment principle.

[0123] The beneficial effect of the above technology is: by judging whether the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period meets the consistency judgment principle, the verification result of the overall cleanliness quantification model is obtained.

[0124] Embodiment 8:

[0125] Based on Example 1, a photovoltaic panel cleanliness detection method, S5: based on the local cleanliness and verification results of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position point in the corresponding photovoltaic panel, obtain the overall cleanliness of the photovoltaic panel to be tested at the latest moment, including:

[0126] When the verification result shows that the overall cleanliness quantification model satisfies the consistency judgment principle, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is determined based on the local cleanliness of all remote sensing image pixels at the corresponding position points in the corresponding photovoltaic panel in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment and the overall cleanliness quantification model;

[0127] When the verification result shows that the overall cleanliness quantification model does not meet the consistency judgment principle, a new overall cleanliness quantification model is obtained based on the photovoltaic panel power generation efficiency of a large number of new photovoltaic panels at each moment in the new historical period and the local cleanliness of all remote sensing image pixels in the corresponding multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel, as well as the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period;

[0128] Based on the local cleanliness of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel and a new overall cleanliness quantification model, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is determined.

[0129] In this embodiment, based on the local cleanliness and overall cleanliness quantification model of all remote sensing image pixels at corresponding positions in the corresponding photovoltaic panel in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is determined, including:

[0130] The local cleanliness of all remote sensing image pixels at corresponding positions in the corresponding photovoltaic panel in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment is input into the overall cleanliness quantification model to obtain the overall cleanliness of the photovoltaic panel to be tested at the latest moment.

[0131] In this embodiment, the new photovoltaic panel is a photovoltaic panel different from the photovoltaic panel used in steps S1 to S4.

[0132] In this embodiment, the new historical period is a historical period different from the historical period used in steps S1 to S4.

[0133] In this embodiment, the method for determining the local cleanliness of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel at the corresponding position points in the corresponding photovoltaic panel at each moment in the new historical period is consistent with the method of obtaining "the local cleanliness of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel at the corresponding position points in the corresponding photovoltaic panel" in steps S1 and S2.

[0134] In this embodiment, based on the local cleanliness of all remote sensing image pixels at corresponding positions in the corresponding photovoltaic panel in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment and the new overall cleanliness quantification model, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is determined, including:

[0135] The local cleanliness of all remote sensing image pixels at corresponding positions in the corresponding photovoltaic panel in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment is input into the new overall cleanliness quantification model to obtain the overall cleanliness of the photovoltaic panel to be tested at the latest moment.

[0136] The beneficial effects of the above technology are: based on the verification results of the overall cleanliness quantification model, it is determined whether the current overall cleanliness quantification model needs to be re-optimized and trained, and when the current overall cleanliness quantification model is determined to need to be re-optimized and trained, a large number of new photovoltaic panels are used to obtain a new overall cleanliness quantification model by using the photovoltaic panel power generation efficiency at each moment in the new historical period and the local cleanliness of all remote sensing image pixels in the corresponding multispectral remote sensing image at the corresponding position points in the corresponding photovoltaic panel, as well as the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period, and by using the local cleanliness of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel and the new overall cleanliness quantification model, an accurate prediction of the overall cleanliness of the photovoltaic panel to be tested at the latest moment is achieved.

[0137] Embodiment 9:

[0138] On the basis of Example 8, the photovoltaic panel cleanliness detection method obtains a new overall cleanliness quantification model based on the photovoltaic panel power generation efficiency of a large number of new photovoltaic panels at each moment in a new historical period and the local cleanliness of all remote sensing image pixels in the corresponding multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel, as well as the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period, including:

[0139] Based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods, a standard numerical relationship between the overall cleanliness and power generation efficiency is obtained;

[0140] Substitute the photovoltaic panel power generation efficiency of all new photovoltaic panels at each moment in the new historical period into the standard numerical relationship between overall cleanliness and power generation efficiency to obtain the overall cleanliness of each new photovoltaic panel at each moment in the new historical period;

[0141] The local cleanliness and the corresponding overall cleanliness of all remote sensing image pixels in the multispectral remote sensing images of all new photovoltaic panels at each moment in the new historical period are used as model input and model output for neural network model training to obtain a new overall cleanliness quantification model.

[0142] In this embodiment, based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods, a standard numerical relationship between the overall cleanliness and power generation efficiency is obtained, including:

[0143] In the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period, the overall cleanliness is represented by x, and the power generation efficiency is represented by a formula containing the x parameter;

[0144] Among the two numerical relationships corresponding to the maximum deviation degree in the numerical relationships of all sub-historical periods, the numerical relationship with the largest deviation degree from the remaining numerical relationships is deleted;

[0145] The average value of the power generation efficiency expressed by the formula containing the x parameter in all the remaining numerical relationships is used as the standard numerical relationship between the overall cleanliness and the power generation efficiency;

[0146] In the standard numerical relationship, y can be used to represent the power generation efficiency, that is, y and the power generation efficiency represented by a formula containing the x parameter are connected with an equal sign to obtain the standard numerical relationship.

[0147] The beneficial effects of the above technology are: using the standard numerical relationship between the overall cleanliness and the power generation efficiency obtained based on the numerical relationship between the overall cleanliness and the power generation efficiency of the photovoltaic panels in all sub-historical time periods, and the photovoltaic panel power generation efficiency of all new photovoltaic panels at each moment in the new historical time period, the overall cleanliness of each new photovoltaic panel at each moment in the new historical time period used for model retraining is determined, and the local cleanliness and corresponding overall cleanliness of all remote sensing image pixels at corresponding positions in the corresponding photovoltaic panels in the multispectral remote sensing images of all new photovoltaic panels at each moment in the new historical time period are used to obtain a new overall cleanliness quantification model, so that the overall cleanliness of the photovoltaic panel predicted by the new overall cleanliness quantification model and the power generation efficiency of the photovoltaic panel at the corresponding moment satisfy a fixed numerical relationship.

[0148] Embodiment 10:

[0149] The present invention provides a photovoltaic panel cleanliness detection system for executing any one of the photovoltaic panel cleanliness detection methods described in embodiments 1 to 9, referring to Figure 3 ,include:

[0150] A local cleanliness calculation module is used to calculate the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at a corresponding position point in the corresponding photovoltaic panel based on the brightness value of each remote sensing image pixel in the red band and the near-infrared band and the corresponding spectral curve of each remote sensing image pixel in the multispectral remote sensing image at each historical moment in the historical period;

[0151] An overall cleanliness calculation module is used to determine the overall cleanliness of the photovoltaic panel based on the local cleanliness and overall cleanliness quantification model of all remote sensing image pixels in the multispectral remote sensing image at corresponding positions in the corresponding photovoltaic panel;

[0152] A numerical relationship extraction module is used to align the overall cleanliness of the photovoltaic panel and the photovoltaic panel power generation efficiency at all the same moments in each sub-historical period within the historical period and extract the numerical relationship, so as to obtain the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period;

[0153] A quantitative model verification module is used to obtain a verification result for the overall cleanliness quantification model based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period;

[0154] The overall cleanliness detection module is used to obtain the overall cleanliness of the photovoltaic panel to be tested at the latest moment based on the local cleanliness and verification results of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel.

[0155] The beneficial effects of the above technology are: combining the multispectral remote sensing images of photovoltaic panels with the neural network model to achieve accurate detection of the cleanliness of photovoltaic panels. There is no need to set up additional detection equipment to obtain detection data (such as weather change data and reflectivity data). Only the existing photovoltaic panel power generation efficiency data is required to be applied to the model verification process, that is, only the multispectral remote sensing images of the photovoltaic panels and the overall cleanliness quantification model are needed to obtain accurate photovoltaic panel cleanliness detection results, which simplifies the detection procedure, reduces the detection cost, reduces the data source items required for the detection process, and reduces the probability of detection result errors due to errors in the data source items.

[0156] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A photovoltaic panel cleanliness detection method, characterized in that: include: S1: Based on the brightness value of each remote sensing image pixel in the red band and near-infrared band of the photovoltaic panel at each historical moment in the historical period and the corresponding spectral curve, the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel is calculated; S2: Based on the local cleanliness and overall cleanliness quantification model of all remote sensing image pixels in the multispectral remote sensing image at the corresponding position points in the corresponding photovoltaic panel, the overall cleanliness of the photovoltaic panel is determined; S3: Align the overall cleanliness of the photovoltaic panel and the power generation efficiency of the photovoltaic panel at all the same moments in each sub-historical period within the historical period and extract the numerical relationship, so as to obtain the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period; S4: The overall cleanliness quantification model is verified based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panels in all sub-historical periods within the historical period, and the verification results are obtained, including: Determine whether the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period meets the consistency judgment principle. If so, the overall cleanliness quantification model meets the consistency judgment principle as a verification result; Otherwise, the overall cleanliness quantification model does not meet the consistency judgment principle as a verification result; S5: Based on the local cleanliness and verification results of all remote sensing image pixels at corresponding positions in the corresponding photovoltaic panel in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is obtained, including: When the verification result shows that the overall cleanliness quantification model satisfies the consistency judgment principle, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is determined based on the local cleanliness of all remote sensing image pixels at the corresponding position points in the corresponding photovoltaic panel in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment and the overall cleanliness quantification model; When the verification result shows that the overall cleanliness quantification model does not meet the consistency judgment principle, a new overall cleanliness quantification model is obtained based on the photovoltaic panel power generation efficiency of a large number of new photovoltaic panels at each moment in the new historical period and the local cleanliness of all remote sensing image pixels in the corresponding multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel, as well as the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period, including: Based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods, a standard numerical relationship between the overall cleanliness and power generation efficiency is obtained; Substitute the photovoltaic panel power generation efficiency of all new photovoltaic panels at each moment in the new historical period into the standard numerical relationship between overall cleanliness and power generation efficiency to obtain the overall cleanliness of each new photovoltaic panel at each moment in the new historical period; The local cleanliness and the corresponding overall cleanliness of all remote sensing image pixels at the corresponding positions in the corresponding photovoltaic panels in the multispectral remote sensing images of all new photovoltaic panels at each moment in the new historical period are used as model input and model output to train the neural network model, and a new overall cleanliness quantification model is obtained; Based on the local cleanliness of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel and a new overall cleanliness quantification model, the overall cleanliness of the photovoltaic panel to be tested at the latest moment is determined.

2. The photovoltaic panel cleanliness detection method according to claim 1, characterized in that: S1: Based on the brightness value of each remote sensing image pixel in the red band and near-infrared band of the photovoltaic panel at each historical moment in the historical period and the corresponding spectral curve, the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel is calculated, including: S101: Obtain the brightness value of each remote sensing image pixel in the red band and the near infrared band of the photovoltaic panel in the multispectral remote sensing image at each historical moment in the historical period and the corresponding spectral curve; S102: based on the spectral curve of each remote sensing image pixel in the multispectral remote sensing image of the photovoltaic panel at each historical moment in the historical period, calculating the reflectivity balance coefficient in the near infrared band and the reflectivity balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image; S103: Based on the brightness values ​​of each remote sensing image pixel in the red band and the near-infrared band of the photovoltaic panel in the multispectral remote sensing image at each historical moment in the historical period and the reflectivity balance coefficient of the corresponding remote sensing image pixel position in the near-infrared band and the reflectivity balance coefficient in the red band, calculate the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel.

3. The photovoltaic panel cleanliness detection method according to claim 2, characterized in that: S102: Based on the spectral curve of each remote sensing image pixel in the multispectral remote sensing image of the photovoltaic panel at each historical moment in the historical period, the reflectivity balance coefficient in the near infrared band and the reflectivity balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image are calculated, including: The spectral curves of the remote sensing image pixels at the same position in the multispectral remote sensing images of the photovoltaic panel at all historical moments in the historical period are time-series aligned to obtain the alignment curve combination of each remote sensing image pixel position; Determine the average of all ordinate values ​​corresponding to each abscissa value in the alignment curve combination of each remote sensing image pixel position as the first average reflectivity of each abscissa value in the alignment curve combination of each remote sensing image pixel position; The average value of all first average reflectances in the alignment curve combination of all horizontal coordinate values ​​belonging to the near-infrared band at each remote sensing image pixel position is regarded as the second average reflectance of the corresponding remote sensing image pixel position in the near-infrared band. At the same time, the average value of all first average reflectances in the alignment curve combination of all horizontal coordinate values ​​belonging to the red band at each remote sensing image pixel position is regarded as the second average reflectance of the corresponding remote sensing image pixel position in the red band. Based on all the second average reflectances of all remote sensing image pixel positions in the near-infrared band and all the second average reflectances in the red band, the reflectance balance coefficient in the near-infrared band and the reflectance balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image are calculated.

4. The photovoltaic panel cleanliness detection method according to claim 3, characterized in that: Based on all the second average reflectances of all remote sensing image pixel positions in the near infrared band and all the second average reflectances in the red band, the reflectance balance coefficient in the near infrared band and the reflectance balance coefficient in the red band of each remote sensing image pixel position in the multispectral remote sensing image are calculated, including: The average of all the second average reflectances of all the remote sensing image pixel positions in the near-infrared band is used as the reference reflectance of the near-infrared band. At the same time, the average of all the second average reflectances of all the remote sensing image pixel positions in the red band is used as the reference reflectance of the red band. The ratio of the difference between the second average reflectivity of each remote sensing image pixel in the near-infrared band and the reference reflectivity of the near-infrared band to the second average emissivity of the corresponding remote sensing image pixel in the near-infrared band is taken as the reflectivity balance coefficient of the corresponding remote sensing image pixel in the near-infrared band. At the same time, the ratio of the difference between the second average reflectivity of each remote sensing image pixel in the red band and the reference reflectivity of the red band to the second average emissivity of the corresponding remote sensing image pixel in the red band is taken as the reflectivity balance coefficient of the corresponding remote sensing image pixel in the red band.

5. The photovoltaic panel cleanliness detection method according to claim 2, characterized in that: S103: Based on the brightness value of each remote sensing image pixel in the red band and the near infrared band of the photovoltaic panel at each historical moment in the historical period and the reflectivity balance coefficient of the corresponding remote sensing image pixel position in the near infrared band and the reflectivity balance coefficient in the red band, the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at the corresponding position point in the corresponding photovoltaic panel is calculated, including: In the formula, is the local cleanliness of a single remote sensing image pixel in the multispectral remote sensing image at the corresponding position in the corresponding photovoltaic panel, is the brightness value of a single remote sensing image pixel in the near infrared band in the multispectral remote sensing image, is the brightness value of a single remote sensing image pixel in the red band in the multispectral remote sensing image, It is the ratio of the brightness value in the near-infrared band to the brightness value in the red band of the remote sensing image pixel in the same position as the currently calculated remote sensing image pixel in the standard dust-free photovoltaic panel under standard lighting conditions. is the reflectivity balance coefficient of the remote sensing image pixel position corresponding to the currently calculated remote sensing image pixel in the near-infrared band, It is the reflectance balance coefficient of the remote sensing image pixel position in the red band corresponding to the currently calculated remote sensing image pixel.

6. The photovoltaic panel cleanliness detection method according to claim 1, characterized in that: S2: Based on the local cleanliness and overall cleanliness quantification model of all remote sensing image pixels in the multispectral remote sensing image at the corresponding position points in the corresponding photovoltaic panel, the overall cleanliness of the photovoltaic panel is determined, including: The local cleanliness of all remote sensing image pixels in the multispectral remote sensing image at the corresponding position points in the corresponding photovoltaic panel is input into the overall cleanliness quantification model to obtain the overall cleanliness of the photovoltaic panel.

7. A photovoltaic panel cleanliness detection system, characterized in that: A photovoltaic panel cleanliness detection method for performing any one of claims 1 to 6, comprising: A local cleanliness calculation module is used to calculate the local cleanliness of each remote sensing image pixel in the multispectral remote sensing image at a corresponding position point in the corresponding photovoltaic panel based on the brightness value of each remote sensing image pixel in the red band and the near-infrared band and the corresponding spectral curve of each remote sensing image pixel in the multispectral remote sensing image at each historical moment in the historical period; An overall cleanliness calculation module is used to determine the overall cleanliness of the photovoltaic panel based on the local cleanliness and overall cleanliness quantification model of all remote sensing image pixels in the multispectral remote sensing image at corresponding positions in the corresponding photovoltaic panel; A numerical relationship extraction module is used to align the overall cleanliness of the photovoltaic panel and the photovoltaic panel power generation efficiency at all the same moments in each sub-historical period within the historical period and extract the numerical relationship, so as to obtain the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in each sub-historical period; A quantitative model verification module is used to verify the overall cleanliness quantification model based on the numerical relationship between the overall cleanliness and power generation efficiency of the photovoltaic panel in all sub-historical periods within the historical period to obtain a verification result; The overall cleanliness detection module is used to obtain the overall cleanliness of the photovoltaic panel to be tested at the latest moment based on the local cleanliness and verification results of all remote sensing image pixels in the multispectral remote sensing image of the photovoltaic panel to be tested at the latest moment at the corresponding position points in the corresponding photovoltaic panel.

Images