A photovoltaic power station solar cell panel intelligent flushing system

Abstract

The present application relates to the technical field of solar panel maintenance, and particularly relates to a kind of photovoltaic power station solar cell panel intelligent flushing system, including image analysis module for monitoring the dirty data of solar cell panel surface, and dirty data is transmitted to intelligent control module in real time, the image analysis module includes multispectral camera, the multispectral camera gathers the surface image of current cell panel and next cell panel, surface image includes the cell panel surface image collected under different spectral bands or angle, obtains the dirty distribution and dirty degree of next cell panel;Utilize preset intelligent algorithm to generate cleaning strategy according to cleaning demand, output control instruction, control instruction is sent to intelligent control module by communication unit, and the working state of intelligent control module is adjusted flushing device.The present application improves the accurate identification ability of system to dirty area, ensures that water flow can effectively cover the part needing to be cleaned in cleaning process, and avoids waste to no dirty area.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of solar panel maintenance, and particularly relates to an intelligent flushing system for solar cell panels of a photovoltaic power station. BACKGROUND

[0002] In a photovoltaic power station, the efficiency of solar cell panels as key components is closely related to the cleaning degree. At present, most of the cleaning systems for the panels on the market adopt traditional methods such as manual cleaning, mechanical brushing and spraying cleaning. Although manual cleaning is effective in some specific scenarios, it cannot meet the needs of efficient operation and maintenance of modern photovoltaic power stations due to the large workload, low efficiency, difficulty in accurately controlling the cleaning strength and safety risks. Although mechanical brushing improves the automation level, it may cause mechanical damage to the surface of the panel and affect its long-term stability due to the fixed brush head and the difficulty in adapting to different cleaning intensities of different dirt conditions.

[0003] In recent years, spraying cleaning has become the mainstream way of cleaning photovoltaic power stations. The surface of the panel is washed by high-pressure water flow. However, most of the existing spraying systems are based on preset cleaning strategies and lack real-time monitoring and intelligent adjustment of the dirt degree of the panel surface. The traditional spraying system cannot adjust the nozzle position, pressure and flushing sequence according to the actual dirt condition of each panel, often leading to resource waste and uneven cleaning effect. In addition, with the continuous expansion of the scale of photovoltaic power stations, cleaning efficiency and resource utilization have become problems to be solved. Especially in the cleaning process of large-area panels, how to accurately clean each panel according to its specific dirt degree and avoid invalid cleaning and damage to the panel has become a difficulty in the current technology. SUMMARY

[0004] The present application provides an intelligent flushing system for solar cell panels of a photovoltaic power station.

[0005] An intelligent flushing system for solar cell panels of a photovoltaic power station, comprising an image analysis module, a central processing unit, a communication unit, an intelligent control module and a flushing device.

[0006] The image analysis module is used for monitoring the dirt data of the surface of the solar cell panel and transmitting the dirt data to the intelligent control module in real time. The image analysis module comprises a multispectral camera. The multispectral camera collects the surface images of the current panel and the next panel. The surface images include the panel surface images collected at different spectral bands or angles. The dirt distribution and degree of the next panel are obtained through reflection region recognition.

[0007] The central processing unit generates a cleaning strategy according to the cleaning requirement by using a preset intelligent algorithm based on the dirt distribution and the dirt degree of the next panel provided by the image analysis module, and outputs a control instruction, which is sent to the intelligent control module through the communication unit, and the intelligent control module adjusts the working state of the flushing device;

[0008] The flushing device includes a high-pressure water gun and an adjustable nozzle, and the intelligent control module adjusts the position and pressure of the adjustable nozzle of the flushing device in advance based on the control instruction, wherein:

[0009] Adjustable nozzle position: automatically adjust the jet position of the adjustable nozzle according to the dirt distribution of the next panel;

[0010] Water flow pressure adjustment: automatically adjust the water pressure of the high-pressure water gun according to the dirt degree, while avoiding damage to the panel caused by excessive water pressure.

[0011] Optionally, the image analysis module specifically includes:

[0012] Multi-spectral image acquisition: the image analysis module uses a multi-spectral camera to acquire panel surface images under different spectral bands or angles, including at least visible light and infrared light images;

[0013] Constructing a reflection area prediction model: the image analysis module trains a neural network specially used for reflection area identification and prediction, which predicts and labels the reflection area according to image features under different light conditions, including reflection intensity, spectral differences between visible light and infrared light images, and morphological features of the reflection area, generates a reflection area layer, and obtains a reflection area map, which is removed from the original image to obtain a dirt area layer;

[0014] Differential processing: differential processing is performed on the dirt area layer to suppress the influence of reflection on the dirt area, and a dirt distribution map of the panel is obtained, and the dirt degree of each dirt area is calculated through image intensity information to ensure accurate assessment of dirt.

[0015] Optionally, the multi-spectral camera acquires solar panel surface images under multiple spectral bands, including visible light images and infrared light images, by adjusting the angle and spectral band of the camera, and the multi-spectral camera includes an adjustable band selector that switches within different spectral band ranges to capture panel surface reflection features under different spectra, providing multi-dimensional data support for dirt area identification and reflection area suppression.

[0016] Optionally, the reflection region prediction model adopts a convolutional neural network structure, and is trained by labeled panel surface image data. During the training process, the network learns image features under different lighting conditions, including reflection intensity, spectral differences between visible light and infrared light images, and morphological features of the reflection region.

[0017] Reflection intensity feature: the neural network extracts the reflection intensity feature in the visible light and infrared light images, and identifies the potential reflection region through the brightness and intensity difference of the reflected light.

[0018] Spectral difference feature: the neural network distinguishes the reflection region and the dirty region by using the spectral difference between the visible light and infrared light images. The reflection region appears as different brightness or color difference in the infrared light image.

[0019] Reflection region morphological feature: the neural network analyzes the morphological features of the reflection region to distinguish the reflection region from other interference regions.

[0020] Through the reflection intensity, the spectral difference between the visible light and infrared light images, and the morphological features of the reflection region, the neural network outputs the prediction result of the reflection region, and generates a reflection region layer R reflect (x,y), where R reflect (x,y) = 1 indicates that the position is a reflection region, and R reflect (x,y) = 0 indicates a non-reflection region. The generated reflection region layer is subjected to differential processing with the dirty region layer M(x,y). In the dirty region layer M(x,y), M(x,y) = 1 indicates that the position is a dirty region, and M(x,y) = 0 indicates a non-dirty region. Thus, the dirty distribution is obtained, and M(x,y) represents the dirty distribution. The differential processing includes suppressing the influence of reflection on the dirty region by superimposing the reflection region layer and the dirty region layer, shielding the region R reflect (x,y) marked by the reflection region, removing the interference of reflected light on the identification of the dirty region, and ensuring the identification accuracy of the dirty region M(x,y). The finally output dirty region layer M(x,y) presents the real dirty distribution of the panel after removing the reflection influence, and provides accurate data support for the subsequent cleaning strategy.

[0021] Optionally, the reflection region prediction model is trained using a convolutional neural network. The input of the neural network is multispectral data including visible light and infrared light images, and the output is a reflection region layer. The training process is performed by minimizing the following loss function:

[0022] L reflect =∑ (x,y) |R reflect (x,y)-R groundtruth (x,y)| 2 , where Rreflect (x,y) is the predicted reflection region layer of the neural network groundtruth (x,y) is the annotated real reflection region layer, obtained by manual annotation, using the reflection region prediction result output by the neural network to generate the reflection region layer R reflect (x,y) is the predicted reflection region layer of the neural network

[0023] Optionally, the differential processing is represented as: M final (x,y) = M(x,y)·1-R reflect (x,y) is the predicted reflection region layer of the neural network

[0024] The reflection region layer R reflect (x,y) is the predicted reflection region layer of the neural network, 1 indicates that the position is a reflection region, 0 indicates a non-reflection region, and the dirty region layer M(x,y) is 1 indicates a dirty region, 0 indicates a non-dirty region, and through differential processing, if R reflect (x,y) = 1, indicating that the position is a reflection region, at this time 1-R reflect (x,y) = 0, so the dirty region layer M(x,y) is suppressed at the reflection region position, ensuring that the interference of the reflection region on the dirty region is removed; if R reflect (x,y) = 0, indicating that the position is not a reflection region, at this time 1-R reflect (x,y) = 1, the dirty region layer M(x,y) remains unchanged.

[0025] Optionally, the dirty degree of each dirty region is calculated based on image intensity information, which specifically includes:

[0026] The intensity value of each pixel of the multi-spectral image is I(x,y), where I(x,y) is the brightness value of the current pixel position, and the lower the intensity, the more dirt accumulated on the surface.

[0027] Dirty will cause the image intensity value to decrease (dirty regions are usually dark), and the dirty degree D(x,y) is calculated:

[0028] Where I(x,y) is the image intensity value of the current pixel position, I max is the maximum intensity value in the entire image (for example, 255 for an 8-bit image), and I min is the minimum intensity value in the entire image, and the dirty degree is quantified by the difference in image intensity, the lower the intensity, the higher the dirty degree, and D(x,y) takes a value in the range [0, 1].

[0029] Optionally, the cleaning strategy specifically includes:

[0030] The flushing sequence is adjusted according to the dirty distribution to adjust the jet position of the nozzle, and the flushing sequence is determined according to the dirty degree and the dirty distribution.

[0031] Water pressure adjustment: according to the degree of dirt, adjust the water pressure of the high-pressure water gun to ensure the cleaning effect of different areas.

[0032] Optionally, the flushing sequence includes priority calculation, represented as:

[0033] P 优 (x,y) = D(x,y) * M(x,y), wherein P 优 (x,y) is the cleaning priority of area (x,y), the higher the cleaning priority, the area should be cleaned first, and the cleaning sequence is sorted based on the cleaning priority.

[0034] Optionally, the water pressure P flow (x,y) adjustment is represented as: P flow (x,y) = P min +(P max -P min ) * D(x,y), wherein P min and P max are the minimum and maximum values of water pressure, respectively, and D(x,y) is the degree of dirt, which determines the range of water pressure change.

[0035] The beneficial effects of the present application are:

[0036] The present application effectively suppresses the interference of reflected light on the surface of the solar panel by combining multi-spectral image acquisition and reflection area prediction model, and accurately distinguishes the reflection area and the dirty area. The traditional flushing system is often affected by the reflected light source, resulting in misjudgment of the dirty area. However, the present application trains a special reflection area prediction network through a deep learning model, combines information of different spectral bands, and can efficiently identify reflection light intensity, spectral difference and morphological characteristics of reflection area, so as to accurately generate a reflection area layer and avoid the interference of reflection artifacts on the identification of dirty area, thereby improving the accurate identification ability of the system to the dirty area and ensuring that the water flow can effectively cover the part that needs to be cleaned during the cleaning process, avoiding waste to the non-dirty area.

[0037] Based on the real-time monitoring of the degree of dirt on the surface of the battery panel, the present application dynamically adjusts the water pressure and sequence of the flushing device through the intelligent control module, so as to accurately match the cleaning demand. Through multi-spectral image analysis and deep learning algorithm, the degree of dirt on the next battery panel can be predicted in advance during the cleaning process of the battery panel, and the working state of the flushing device is automatically adjusted. In particular, the dynamic adjustment of the water flow position is optimized according to the distribution of dirt, and the adjustment of the water flow pressure is refined according to the degree of dirt, which improves the cleaning efficiency, reduces the waste of water resources, and avoids excessive cleaning and damage to the battery panel by water flow. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description only illustrate the present application, and other drawings can also be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0039] Fig. 1 A schematic diagram of each module of the flushing system of the embodiment of the present application;

[0040] Fig. 2 A schematic diagram of the image processing module of the embodiment of the present application. DETAILED DESCRIPTION

[0041] The present application will be described in detail below with reference to the drawings and specific embodiments. It should be noted that in order to make the embodiments more detailed, the following embodiments are the best, preferred embodiments, and other alternative ways can also be used by those skilled in the art to implement them; and the drawings are only used to more specifically describe the embodiments, and are not intended to specifically limit the present application.

[0042] It should be noted that in the specification, "one embodiment", "embodiment", "exemplary embodiment", "some embodiments" and the like indicate that the described embodiments can include a specific feature, structure or property, but not necessarily every embodiment includes the specific feature, structure or property. In addition, when a specific feature, structure or property is described in combination with an embodiment, it should be within the knowledge of those skilled in the related art to realize this feature, structure or property in combination with other embodiments (whether or not explicitly described).

[0043] Generally, the terms can be understood at least in part from the use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular or can be used to describe combinations of features, structures, or characteristics that are combinable into one or more instances. In addition, the term "based on" can be understood as not necessarily intended to convey a set of exclusive factors, but can instead, depending at least in part on the context, allow for the existence of other factors that are not necessarily explicitly described.

[0044] As shown in Figs. 1-2 A photovoltaic power station solar cell panel intelligent flushing system, comprising an image analysis module, a central processing unit, a communication unit, an intelligent control module, a flushing device;

[0045] The image analysis module is used for monitoring the dirt data of the surface of the solar panel and transmitting the dirt data to the intelligent control module in real time. The image analysis module comprises a multispectral camera. The multispectral camera collects surface images of the current panel and the next panel. The surface images comprise panel surface images collected at different spectral bands or angles. The dirt distribution and the dirt degree of the next panel are obtained through reflection region identification.

[0046] The central processing unit generates a cleaning strategy according to the cleaning requirement by using a preset intelligent algorithm according to the dirt distribution and the dirt degree of the next panel provided by the image analysis module, and outputs a control instruction. The control instruction is sent to the intelligent control module through the communication unit. The intelligent control module adjusts the working state of the flushing device.

[0047] The flushing device comprises a high-pressure water gun and an adjustable nozzle. Based on the control instruction, the intelligent control module adjusts the position and pressure of the adjustable nozzle of the flushing device in advance.

[0048] Adjustable nozzle position: The spraying position of the adjustable nozzle is automatically adjusted according to the dirt distribution of the next panel.

[0049] Water pressure adjustment: The water pressure of the high-pressure water gun is automatically adjusted according to the dirt degree, while avoiding damage to the panel caused by excessive water pressure.

[0050] The image analysis module specifically comprises:

[0051] Multispectral image collection: The image analysis module uses a multispectral camera to collect panel surface images at different spectral bands or angles, including at least visible light and infrared light images.

[0052] Constructing a reflection region prediction model: The image analysis module trains a neural network specially used for reflection region identification and prediction. The neural network predicts and labels the reflection region according to image features under different light conditions, including reflection intensity, spectral differences between visible light and infrared light images, and morphological features of the reflection region. The generated reflection region layer is obtained by removing the reflection region from the original image to obtain a dirt region layer.

[0053] Differential processing: The reflection region layer is subjected to differential processing with the dirt region layer to suppress the influence of reflection on the dirt region, so as to obtain a dirt distribution map of the panel. The dirt degree of each dirt region is calculated through image intensity information to ensure accurate assessment of the dirt.

[0054] The image analysis module transmits the fused dirt distribution data and dirt degree data to the intelligent control module in real time. The dirt data includes the pollution distribution, the dirt degree and the preliminary judgment information of the cleaning requirement of the next panel, for further analysis and processing by the intelligent control module.

[0055] The multispectral camera adjusts the angle and spectral band of the camera to collect solar panel surface images at multiple spectral bands, including visible light images and infrared light images. The multispectral camera includes an adjustable band selector that switches within different spectral band ranges to capture the panel surface reflection characteristics under different spectra, providing multi-dimensional data support for dirty area identification and reflection area suppression.

[0056] The reflection area prediction model adopts a convolutional neural network structure and is trained by labeled panel surface image data. During the training process, the network learns image features under different lighting conditions, including reflection intensity, spectral differences between visible light and infrared light images, and morphological features of the reflection area.

[0057] Reflection intensity feature: the neural network extracts reflection intensity features from visible light and infrared light images and identifies potential reflection areas through the brightness and intensity differences of reflected light.

[0058] Spectral difference feature: the neural network distinguishes reflection areas and dirty areas using the spectral differences between visible light and infrared light images, with reflection areas appearing as different brightness or color differences in infrared light images.

[0059] Reflection area morphological feature: the neural network analyzes the morphological features of the reflection area to distinguish it from other interference areas.

[0060] Through reflection intensity, spectral differences between visible light and infrared light images, and morphological features of the reflection area, the neural network outputs the prediction results of the reflection area, generating a reflection area layer R reflect (x,y), where R reflect (x,y) = 1 indicates that the position is a reflection area, and R reflet (x,y) = 0 indicates a non-reflection area. The generated reflection area layer is differentiated with the dirty area layer M(x,y), where M(x,y) = 1 indicates that the position is a dirty area, and M(x,y) = 0 indicates a non-dirty area. Thus, the dirty distribution is obtained, represented by M(x,y). The differentiation process includes suppressing the influence of reflection on the dirty area by overlaying the reflection area layer and the dirty area layer, masking the area R reflect (x,y) marked by the reflection area, removing the interference of reflected light on dirty area identification, and ensuring the accuracy of dirty area M(x,y) identification. The final output of the dirty area layer M(x,y) presents the true dirty distribution of the panel after removing the reflection effect, providing accurate data support for subsequent cleaning strategies.

[0061] Reflection area intensity R intensity(x, y) represents the reflection intensity at a certain position (x, y), which is calculated by comparing the brightness difference between the visible light and infrared light images, and is expressed using the following formula:

[0062] R intensity (x, y) = I visible (x, y) - I infrared (x, y) |, where I visible (x, y) is the pixel value of the visible light image at position (x, y), I infrared (x, y) is the pixel value of the infrared light image at position (x, y), R intensity (x, y) reflects the intensity difference of reflected light;

[0063] Spectral difference R spectral (x, y) measures the spectral difference by calculating the pixel value difference between the visible light and infrared light bands, and is calculated as follows: where I visible (x, y) and I infrared (x, y) are the pixel values of the visible light image and the infrared light image at position (x, y), respectively, R spectral (x, y) is the spectral difference value at this position, reflecting the relative difference between the two spectra;

[0064] Reflection region morphology feature extraction: morphological feature R shape (x, y) extracts the morphological features of the reflection region through morphological operations, and detects the reflection region by calculating the morphological features of the region. Morphological operations include erosion, dilation, opening operation, closing operation, and highlighting the boundary of the reflection region.

[0065] The reflection region prediction model is trained using a convolutional neural network, and the input of the neural network is multispectral data including visible light and infrared light images, and the output is a reflection region layer. The training process is performed by minimizing the following loss function:

[0066] L reflect = Σ (x,y) | R reflect (x, y) - R groundtruth (x, y) | 2 , where R reflect (x, y) is the reflection region layer predicted by the neural network, R groundtruth (x, y) is the labeled true reflection region layer, which is obtained by manual labeling. The reflection region prediction result output by the neural network is used to generate the reflection region layer R reflect (x, y).

[0067] The differential processing is represented as: M final (x, y) = M(x, y) · 1 - Rreflect (x,y)) ;

[0068] Reflection area layer R reflect (x,y), 1 means the position is a reflection area, 0 means a non-reflection area, in the dirt area layer M(x,y), 1 means a dirt area, 0 means a non-dirt area, through differential processing, if R reflect (x,y) = 1, it means that the position is a reflection area, at this time 1-R reflect (x,y) = 0, so the dirt area layer M(x,y) at the reflection area position will be suppressed, ensuring that the interference of the reflection area on the dirt area is removed; if R reflect (x,y) = 0, it means that the position is not a reflection area, at this time 1-R reflect (x,y) = 1, the dirt area layer M(x,y) remains unchanged. M final (x,y) is the dirt area layer after the suppression of the reflection area, wherein the influence of the reflection area is suppressed, ensuring the accurate identification of the dirt area.

[0069] The dirt degree of each dirt area is calculated through image intensity information, which specifically includes:

[0070] The intensity value of each pixel of the multi-spectrum image is I(x,y), wherein I(x,y) is the brightness value of the current pixel position, and the lower the intensity, the more dirt accumulated on the surface;

[0071] Dirt will cause the image intensity value to decrease (dirt area is usually dark), and the dirt degree D(x,y) is calculated:

[0072] Wherein, I(x,y) is the image intensity value of the current pixel position, I max is the maximum intensity value in the entire image (for example, 255 for an 8-bit image), and I min is the minimum intensity value in the entire image, and the dirt degree is quantified by the difference in image intensity, and the lower the intensity, the higher the dirt degree;

[0073] In order to ensure that the dirt degree D(x,y) is within the range of 0 to 1, normalization is performed according to the minimum and maximum values of the image, and I max and I min in the formula are dynamically determined according to the brightness distribution of the image, ensuring that the calculated D(x,y) has a value range of [0, 1].

[0074] The cleaning strategy specifically includes:

[0075] The flushing sequence is adjusted according to the dirt distribution, and the flushing sequence is determined according to the dirt degree and the dirt distribution;

[0076] Water pressure adjustment: adjust the water pressure of the high-pressure water gun according to the degree of dirtiness to ensure the cleaning effect of different areas.

[0077] The flushing sequence includes priority calculation, which is represented as:

[0078] P 优 (x,y) = D(x,y)·M(x,y), where P 优 (x,y) is the cleaning priority of area (x,y), the higher the cleaning priority, the area should be cleaned first, and the cleaning sequence is sorted based on the cleaning priority.

[0079] Water pressure P flow (x,y) adjustment is represented as: P flow (x,y) = P min +(P max -P min )·D(x,y), where P min and P max are the minimum and maximum values of water pressure, respectively, and D(x,y) is the degree of dirtiness, which determines the range of water pressure change.

[0080] The present application covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present application. In order to make the public have a thorough understanding of the present application, specific details are described in the following preferred embodiments of the present application, and the present application can also be fully understood without the description of these details to those skilled in the art. In addition, in order to avoid unnecessary confusion to the essence of the present application, well-known methods, processes, procedures, elements and circuits, etc. are not described in detail.

[0081] The above is only the preferred embodiment of the present application, it should be pointed out that for ordinary skilled in the art, without departing from the principles of the present application, can also make a number of improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A photovoltaic power plant solar panel intelligent washing system, characterized in that, The solar panel cleaning device comprises an image analysis module, a central processing unit, a communication unit, an intelligent control module and a flushing device. The image analysis module is used for monitoring the dirt data of the surface of the solar panel and transmitting the dirt data to the intelligent control module in real time. The image analysis module comprises a multispectral camera. The multispectral camera collects the surface images of the current solar panel and the next solar panel. The surface images comprise the images of the solar panel surface collected under different spectral bands or angles. The image analysis module specifically comprises: Multispectral image collection: the image analysis module uses a multispectral camera to collect the surface images of the solar panel under different spectral bands or angles, including at least visible light and infrared light images. Construction of a reflection region prediction model: the image analysis module trains a neural network specially used for reflection region identification and prediction. The neural network predicts and labels the reflection region according to the image features under different light conditions, including reflection intensity, spectral difference between visible light and infrared light images and morphological features of the reflection region. The generated reflection region layer is obtained from the original image by excluding the reflection region, and the dirt region layer is obtained by differentiating the reflection region layer. The neural network outputs a prediction result of the reflection region through spectral differences of reflection intensity, visible light and infrared light images and morphological characteristics of the reflection region, to generate a reflection region layer wherein, represents that the position is a reflection region, represents a non-reflection region, and the generated reflection region layer and the dirt region layer are subjected to differential processing, and the dirt region layer wherein, represents that the position is a dirt region, represents a non-dirt region, thereby obtaining a dirt distribution, is a representation of the dirt distribution, and the differential processing includes suppressing the influence of reflection on the dirt region by superimposing the reflection region layer and the dirt region layer, and marking out the region marked by the reflection region is shielded, the interference of reflected light on the identification of the dirt region is removed, and the identification accuracy of the dirt region is ensured; Differential processing: the reflection region layer is differentiated with the dirt region layer to suppress the influence of reflection on the dirt region, and the dirt distribution map of the solar panel is obtained. The dirt degree of each dirt region is calculated through the image intensity information. The reflection region prediction model adopts a convolutional neural network structure and is trained through labeled solar panel surface image data. During the training process, the network learns the image features under different light conditions, including reflection intensity, spectral difference between visible light and infrared light images and morphological features of the reflection region. Reflection intensity feature: the neural network extracts the reflection intensity feature in the visible light and infrared light images and identifies the potential reflection region through the brightness and intensity difference of the reflected light. Spectral difference feature: the neural network distinguishes the reflection region and the dirt region by using the spectral difference between the visible light and infrared light images. Reflection region morphological feature: the neural network analyzes the morphological features of the reflection region to distinguish the reflection region from other interference regions. The central processing unit generates a cleaning strategy according to the cleaning requirements by using a preset intelligent algorithm according to the dirt distribution and degree of the next solar panel provided by the image analysis module. The control instruction is output and sent to the intelligent control module through the communication unit. The intelligent control module adjusts the working state of the flushing device. The flushing device comprises a high-pressure water gun and an adjustable nozzle. Based on the control instruction, the intelligent control module adjusts the position and pressure of the adjustable nozzle of the flushing device in advance. Adjustable nozzle position: the adjustable nozzle position is automatically adjusted according to the dirt distribution of the next solar panel. Water pressure adjustment: the water pressure of the high-pressure water gun is automatically adjusted according to the dirt degree, and damage to the solar panel caused by excessive water pressure is avoided.

2. The intelligent washing system for solar panels of a photovoltaic power station according to claim 1, characterized in that, The multispectral camera adjusts the angle and spectral band of the camera to collect solar panel surface images in multiple spectral bands, including visible light images and infrared light images. The multispectral camera includes an adjustable band selector that switches between different spectral band ranges to capture the reflection characteristics of the panel surface under different spectra.

3. The intelligent washing system for solar panels of a photovoltaic power station according to claim 1, characterized in that, The reflection area prediction model is trained using a convolutional neural network, with the input being multispectral data including visible light and infrared light images, and the output being a reflection area layer. The training process is performed by minimizing the following loss function: wherein, is a neural network predicted reflection region map layer, is a labeled real reflection region map layer, obtained by manual labeling, using the reflection region prediction result output by the neural network, to generate a reflection region map layer .

4. The intelligent washing system for solar panels of a photovoltaic power station according to claim 2, characterized in that, The differentiated processing is represented as: ; reflective region layer In the reflective region layer, 1 indicates that the position is a reflective region, and 0 indicates a non-reflective region. In the dirty region layer, 1 indicates a dirty region, and 0 indicates a non-dirty region. If the value of the reflective region layer is 1, it indicates that the position is a reflective region, and at this time The value of the dirty region layer is 0, so the dirty region layer The position of the reflective region is suppressed, ensuring that the interference of the reflective region on the dirty region is removed; if If the value of the reflective region layer is 0, it indicates that the position is not a reflective region, and at this time The value of the dirty region layer is 1, so the dirty region layer Remains unchanged.

5. The intelligent washing system for solar panels of a photovoltaic power station according to claim 1, characterized in that, The calculation of the dirt degree of each dirty area based on image intensity information specifically includes: The intensity value of each pixel of the multispectral image is wherein is the brightness value of the current pixel position, the lower the intensity, the more dirt the surface area has accumulated; Dirt can cause a decrease in image intensity values, the degree of dirt is calculated : wherein, is the image intensity value of the current pixel position, is the maximum intensity value in the entire image, is the minimum intensity value in the entire image, the dirtiness degree is quantified by the difference of image intensity, the lower the intensity, the higher the dirtiness degree, with a value range of [0, 1].

6. The intelligent washing system for solar panels of a photovoltaic power station according to claim 1, characterized in that, The cleaning strategy specifically includes: Rinse sequence: adjust the spray position of the nozzle according to the dirt distribution, and determine the rinse sequence according to the dirt degree and dirt distribution; Water pressure adjustment: adjust the water pressure of the high-pressure water gun according to the dirt degree to ensure the cleaning effect of different areas.

7. A photovoltaic power plant solar panel intelligent washing system according to claim 6, characterized in that, The rinse sequence includes priority calculation, which is represented as: wherein, is a region a cleaning priority, the higher the cleaning priority, the region should be cleaned first, based on the cleaning priority, the cleaning sequence is sorted.

8. A photovoltaic power plant solar panel intelligent washing system according to claim 7, characterized in that, The water pressure The adjustment is expressed as: wherein, and are the minimum and maximum values of the water pressure, respectively, is the degree of dirtiness, which determines the range of variation of the water pressure.

Images