Photovoltaic panel dust deposition identification and cleaning auxiliary system and method based on deep learning

Abstract

The invention discloses a photovoltaic panel dust deposition identification and cleaning auxiliary system and method based on deep learning, and belongs to the technical field of photovoltaic panel dust deposition identification, and the method specifically comprises the steps: collecting a surface image of a photovoltaic panel, carrying out the preprocessing of the collected surface image of the photovoltaic panel, and carrying out the denoising, image enhancement and image segmentation operations, the method comprises the following steps: constructing a photovoltaic panel dust accumulation identification model, training the photovoltaic panel dust accumulation identification model, identifying dust accumulation of a photovoltaic panel, carrying out quantitative classification on dust accumulation grades, obtaining a dust accumulation degree of the photovoltaic panel, segmenting a dust accumulation region, extracting a dust accumulation region in the photovoltaic panel, and dynamically planning a cleaning path according to dust accumulation region distribution. The photovoltaic panel is cleaned; through identification processing and deep restoration of the shielding object, accumulated dust under the shielding object can be effectively identified, complex environmental factors can be better processed, and the accuracy and reliability of accumulated dust analysis are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic panel dust accumulation recognition, and specifically to a photovoltaic panel dust accumulation recognition and cleaning auxiliary system and method based on deep learning. Background Art

[0002] Photovoltaic panels (also known as solar panels) are an important part of the clean energy field and have been widely used in residential, industrial and commercial fields to convert solar energy into electrical energy. However, photovoltaic panels are easily affected by environmental factors during long-term use, especially under the influence of dust, rain, wind and other factors, dust, sand, leaves and other pollutants may accumulate on the surface of photovoltaic panels. These pollutants will reduce the amount of light received by the surface of photovoltaic panels, thereby reducing the energy efficiency of photovoltaic panels and may even cause permanent damage to the panels.

[0003] Traditional cleaning methods usually rely on manual regular inspection and cleaning, which is time-consuming and labor-intensive, may not be thorough, and cannot respond to changes in dirt on the surface of photovoltaic panels in a timely manner.

[0004] For example, the patent application with publication number CN116030409A discloses a method for identifying the dust accumulation state of photovoltaic panels based on adaptive image segmentation, including the following steps: S1, image preprocessing; S2, photovoltaic string segmentation; S3, dust segmentation; S4, image blocking; S5, feature extraction; S6, dust state identification; the invention takes into account that the collected image and the segmented image are both color images, so in order to improve the algorithm operation speed, the segmented image is processed in blocks, which not only improves the operation speed, but also increases the number of image texture features, and then uses the LBP algorithm to realize feature extraction of dust, and finally the state is identified by the softmax classifier; the identification of the dust accumulation state of photovoltaic panels can prevent hot spots and breakage of photovoltaic panels, reduce photovoltaic panel losses, and increase service life, thereby maximizing the utilization rate of photovoltaic panels, reducing operation and maintenance costs, and providing an evaluation basis for the subsequent health status of photovoltaic arrays.

[0005] For example, the patent application with publication number CN118628791A discloses a method for identifying the dust accumulation degree of photovoltaic panels, which specifically includes: step 1, image acquisition, step 2, image data preparation and preprocessing, step 3, constructing a dust accumulation degree recognition model of an improved DenseNet169 network, and step 4, training data set. After the training is completed, the dust accumulation degree recognition result of the photovoltaic panel is obtained; the invention adopts an improved DenseBlock module to replace the DenseBlock module in the original DenseNet169 model. In addition, during the training process, the Ranger optimization algorithm is used to update the model parameters, and the transfer learning method is introduced. The improved DenseNet169 network model improves the model accuracy while reducing the model parameter amount and floating-point calculation amount. Compared with the existing dust accumulation degree recognition model, it has a great improvement and is more practical.

[0006] The defects of the above patent are: the processing of complex environments is relatively rough, and the error in identifying dust accumulation under obstructions is large. Summary of the invention

[0007] In response to the shortcomings of the prior art, the present invention proposes a photovoltaic panel dust accumulation identification and cleaning assistance system and method based on deep learning, which can accurately identify the dust accumulation on the surface of the photovoltaic panel, and provide cleaning assistance suggestions based on the identification results, optimize the cleaning process of the photovoltaic panel, thereby improving the efficiency and economy of the photovoltaic power generation system.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A photovoltaic panel dust accumulation recognition and cleaning auxiliary system based on deep learning, comprising: an image acquisition module, an image preprocessing module, a dust accumulation degree analysis module, a dust accumulation area extraction module and an intelligent cleaning module;

[0010] The image acquisition module uses a drone or ground-based camera to acquire surface images of the photovoltaic panels;

[0011] The image preprocessing module is used to preprocess the collected surface image of the photovoltaic panel, including denoising, image enhancement and image segmentation operations;

[0012] The dust accumulation degree analysis module is used to construct a photovoltaic panel dust accumulation recognition model, train the photovoltaic panel dust accumulation recognition model, identify the dust accumulation of the photovoltaic panel, and quantify and classify the dust accumulation level;

[0013] The dust accumulation area extraction module is used to segment the dust accumulation area after obtaining the dust accumulation degree of the photovoltaic panel, and extract the dust accumulation area in the photovoltaic panel;

[0014] The intelligent cleaning module dynamically plans a cleaning path according to the distribution of dust accumulation areas to clean the photovoltaic panels.

[0015] Specifically, the dust accumulation degree analysis module includes: a model training unit and a dust accumulation identification unit;

[0016] The model training unit is used to construct a photovoltaic panel dust accumulation recognition model and perform training;

[0017] The dust accumulation recognition unit uses the trained photovoltaic panel dust accumulation recognition model to recognize dust accumulation on the surface of the photovoltaic panel and classifies the recognized dust accumulation on the surface of the photovoltaic panel.

[0018] Specifically, the photovoltaic panel dust accumulation recognition model is constructed and trained, including:

[0019] Obtain a public image dataset of photovoltaic panel dust accumulation, attach category labels to the photovoltaic panel dust accumulation images in the public image dataset of photovoltaic panel dust accumulation, and divide the public image dataset of photovoltaic panel dust accumulation into a training set and a test set in a ratio of 7:3;

[0020] Construct a photovoltaic panel dust accumulation recognition model, including: a convolution layer, used to extract local features in the photovoltaic panel dust accumulation image; a pooling layer, used for downsampling to reduce the data dimension and calculation amount; a fully connected layer, used to map the photovoltaic panel dust accumulation image features to the final category output; an activation function, used to output the final classification result;

[0021] Define the loss function of the photovoltaic panel dust accumulation recognition model. The specific formula is:

[0022]

[0023] Among them, Γ represents the loss function of the photovoltaic panel dust accumulation recognition model, C represents the number of photovoltaic panel dust accumulation categories, and yc i represents the true label of the photovoltaic panel dust accumulation image in the photovoltaic panel dust accumulation public image dataset, yc i ' represents the probability predicted by the photovoltaic panel dust accumulation recognition model; i represents the number index of photovoltaic panel dust accumulation categories;

[0024] The training set is input into the photovoltaic panel dust accumulation recognition model in batches. For each batch, forward propagation is performed to calculate the prediction value, and the loss is calculated through the loss function. Through iterative training, until the loss value converges or reaches the preset stopping condition, a trained photovoltaic panel dust accumulation recognition model is obtained.

[0025] Specifically, the category labels include: no dust accumulation, light dust accumulation, moderate dust accumulation, and heavy dust accumulation.

[0026] Specifically, the method of using the trained photovoltaic panel dust accumulation recognition model to identify dust accumulation on the surface of the photovoltaic panel and classifying the identified dust accumulation on the surface of the photovoltaic panel includes:

[0027] The pre-processed surface image of the photovoltaic panel is converted into the HSV color space and standardized;

[0028] The deep learning model is used to detect the occluders in the surface image of the photovoltaic panel, and the occluder set D in the surface image of the photovoltaic panel is obtained, where D = {(zb1,c1,p1),(zb2,c2,p2),...,(zb m ,c m ,p m )}, where zb m represents the coordinates of the mth occluder in the surface image of the photovoltaic panel, c m Represents the category of the mth occluder in the surface image of the photovoltaic panel, usually represented as a category label, p m Represents the confidence of the mth occluder in the surface image of the photovoltaic panel, that is, the probability of belonging to this category;

[0029] The regression model is used to predict the bounding box position of the object in the surface image of the photovoltaic panel. Combined with the surface image feature map of the photovoltaic panel extracted by the deep learning model during detection, the offset (Δx, Δy, Δw, Δh) of each occlusion candidate box is output, where Δx represents the offset of the horizontal coordinate of the occlusion candidate box, Δy represents the offset of the vertical coordinate of the occlusion candidate box, Δw represents the offset of the width of the occlusion candidate box, and Δh represents the offset of the height of the occlusion candidate box. The predicted bounding box is adjusted using the offset. The specific formula for adjusting the predicted bounding box is:

[0030]

[0031] in, represents the predicted bounding box position, b represents the true bounding box position, that is, the actual position of the occluder, and x center ,y center ,w,h represent the central horizontal coordinate, central vertical coordinate, width and height of the predicted occluder respectively. w * ,h * Indicates the actual center horizontal coordinate, center vertical coordinate, width and height of the occluder.

[0032] Specifically, the method of using the trained photovoltaic panel dust accumulation recognition model to identify dust accumulation on the surface of the photovoltaic panel and classifying the identified dust accumulation on the surface of the photovoltaic panel also includes:

[0033] The obtained occluder prediction bounding box is removed from the surface image of the photovoltaic panel, and the blank area after removing the occluder is repaired using the mean filling method to obtain the initial repaired image I of the occluder prediction bounding box init (x, y), optimize the boundary of the occluder prediction bounding box. The specific formula is:

[0034] L boun =v (x',y')∈θ |▽I init (x',y')-▽I orig (x',y')|;

[0035] Among them, L boun represents the boundary difference between the initial repaired image of the occluder prediction bounding box and the surface image of the photovoltaic panel, that is, the difference in the boundary area of ​​the occluder prediction bounding box, (x', y') represents the pixel coordinates of the boundary area of ​​the occluder prediction bounding box, θ represents the boundary of the occluder prediction bounding box, that is, the boundary contour composed of all pixels in the repair area, |·| represents the absolute value function, ▽I init (x', y') represents the gradient of the initial repaired image of the predicted bounding box of the occluder at pixel (x', y'), ▽I orig (x', y') represents the gradient of the surface image of the photovoltaic panel at the pixel (x', y');

[0036] The surface image of the repaired photovoltaic panel is input into the trained photovoltaic panel dust accumulation recognition model to obtain the degree of dust accumulation of the photovoltaic panel.

[0037] Specifically, segmenting the dust accumulation area to extract the dust accumulation area in the photovoltaic panel includes:

[0038] The local adaptive threshold is set to perform region segmentation based on the brightness difference of the surface image of the repaired photovoltaic panel. The specific formula of the local adaptive threshold is:

[0039] T(x,y)=μ(x,y)-α×σ(x,y);

[0040] Wherein, T(x,y) represents the local adaptive threshold, i.e., the adaptive threshold of the pixel point (x,y), μ(x,y) represents the local average brightness of the neighborhood around the pixel point (x,y), σ(x,y) represents the standard deviation of the neighborhood around the pixel point (x,y), and α represents the adjustment factor; (x,y) represents the pixel point in the surface image of the repaired photovoltaic panel;

[0041] Based on the local adaptive threshold T(x,y), the surface image of the repaired photovoltaic panel is divided into dust accumulation area and non-dust accumulation area. The pixel value corresponding to the dust accumulation area is greater than T(x,y), while the pixel value corresponding to the non-dust accumulation area is less than T(x,y).

[0042] Specifically, the collected surface image of the photovoltaic panel is preprocessed, including denoising, image enhancement and image segmentation operations. Denoising uses algorithms such as Gaussian filtering to remove noise in the surface image of the photovoltaic panel; image enhancement is used to adjust the brightness or contrast of the surface image of the photovoltaic panel; and image segmentation operation is used to segment the photovoltaic panel area to distinguish the background from the photovoltaic panel area.

[0043] Specifically, the intelligent cleaning module includes: an intelligent planning unit and a cleaning unit;

[0044] The intelligent planning unit dynamically plans the cleaning path according to the distribution of dust accumulation areas;

[0045] The cleaning unit cleans the surface of the photovoltaic panel according to the planned path.

[0046] A photovoltaic panel dust accumulation identification and cleaning assistance method based on deep learning is implemented based on the photovoltaic panel dust accumulation identification and cleaning assistance system based on deep learning, and specifically includes the following steps:

[0047] Step 1: Collect the surface image of the photovoltaic panel;

[0048] Step 2: Preprocessing the collected surface images of the photovoltaic panels, including denoising, image enhancement and image segmentation operations;

[0049] Step 3: Build a photovoltaic panel dust accumulation recognition model, train the photovoltaic panel dust accumulation recognition model, identify the dust accumulation of the photovoltaic panel, and quantify and classify the dust accumulation level;

[0050] Step 4: After obtaining the dust accumulation degree of the photovoltaic panel, the dust accumulation area is segmented to extract the dust accumulation area in the photovoltaic panel;

[0051] Step 5: According to the distribution of dust accumulation areas, dynamically plan the cleaning path and clean the photovoltaic panels.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. The present invention proposes a photovoltaic panel dust accumulation identification and cleaning assistance system based on deep learning. Through the training of the deep learning model, it can realize the quantitative evaluation of the dust accumulation degree of the photovoltaic panel. This evaluation can not only identify the location of the dust accumulation area, but also determine the severity of the dust accumulation.

[0054] 2. The present invention proposes a photovoltaic panel dust accumulation recognition and cleaning assistance system based on deep learning. Through the recognition and deep repair of obstructions, it can effectively identify the dust accumulation under the obstructions, and can better handle complex environmental factors (such as light, shadow, type of dirt, etc.), thereby improving the accuracy and reliability of dust accumulation analysis.

[0055] 3. The present invention proposes a photovoltaic panel dust accumulation identification and cleaning assistance system based on deep learning, which automatically completes dust accumulation detection and cleaning plan generation without human intervention, greatly reducing labor costs and operational risks. At the same time, due to the personalized design of the cleaning plan, it can also reduce unnecessary waste of resources and achieve more efficient and economical photovoltaic panel maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a deep learning-based photovoltaic panel dust accumulation identification and cleaning assistance system architecture diagram provided by the present invention;

[0057] Figure 2 A flow chart of a photovoltaic panel dust accumulation identification and cleaning assistance method based on deep learning provided by the present invention. DETAILED DESCRIPTION

[0058] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, in the description of the present invention, it should be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside" and the like indicate the orientation or position relationship based on the orientation or position relationship shown in the accompanying drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "No. 1", "No. 2" and "No. 3" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. The present invention is further explained below in conjunction with specific embodiments.

[0059] Example 1

[0060] See also Figure 1 , an embodiment provided by the present invention: a photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning, comprising: an image acquisition module, an image preprocessing module, a dust accumulation degree analysis module, a dust accumulation area extraction module and an intelligent cleaning module;

[0061] The image acquisition module uses a drone or ground camera to regularly capture and photograph the surface images of the photovoltaic panels;

[0062] The image acquisition frequency and angle should be set according to the layout of the photovoltaic panels and the actual application scenario. The images should include the surface status of the photovoltaic panels in different time periods and under different environmental conditions. The images should have sufficient resolution and clarity to ensure that the details of the dust can be captured;

[0063] The image preprocessing module is used to preprocess the collected surface images of the photovoltaic panels, including operations such as denoising, image enhancement and image segmentation, so as to improve the accuracy of model recognition;

[0064] The denoising operation uses algorithms such as Gaussian filtering to remove noise from the image and improve the image quality; the image enhancement operation adjusts the brightness or contrast of the image and enhances the model's adaptability to changes in illumination; the image segmentation operation segments the photovoltaic panel area and distinguishes the background from the photovoltaic panel area;

[0065] Image enhancement also includes: rotation, randomly rotating the image at a certain angle to simulate the situation under different shooting angles, flipping, randomly flipping the image horizontally or vertically to increase the diversity of the image, etc.

[0066] The dust accumulation degree analysis module is used to construct a photovoltaic panel dust accumulation recognition model, train the photovoltaic panel dust accumulation recognition model, identify the dust accumulation of the photovoltaic panel, and quantify and classify the dust accumulation level;

[0067] The dust accumulation degree analysis module includes: a model training unit and a dust accumulation identification unit;

[0068] The model training unit is used to construct a photovoltaic panel dust accumulation recognition model and perform training;

[0069] The dust accumulation recognition unit uses the trained photovoltaic panel dust accumulation recognition model to recognize dust accumulation on the surface of the photovoltaic panel and classifies the recognized dust accumulation on the surface of the photovoltaic panel.

[0070] The photovoltaic panel dust accumulation recognition model is constructed and trained, and the specific steps include:

[0071] Step S1: obtaining a public image dataset of photovoltaic panel dust accumulation, attaching category labels to the photovoltaic panel dust accumulation images in the public image dataset of photovoltaic panel dust accumulation, and dividing the public image dataset of photovoltaic panel dust accumulation into a training set and a test set in a ratio of 7:3;

[0072] Step S2: construct a photovoltaic panel dust accumulation recognition model, including: a convolution layer, used to extract local features in the photovoltaic panel dust accumulation image; a pooling layer, used to downsample and reduce the dimension and calculation amount of the data; a fully connected layer, used to map the photovoltaic panel dust accumulation image features to the final category output; an activation function, used to output the final classification result;

[0073] Step S3: Define the loss function of the photovoltaic panel dust accumulation recognition model. The specific formula is:

[0074]

[0075] Among them, Γ represents the loss function of the photovoltaic panel dust accumulation recognition model, C represents the number of photovoltaic panel dust accumulation categories, and yci represents the true label of the photovoltaic panel dust accumulation image in the photovoltaic panel dust accumulation public image dataset, yc i ' represents the probability predicted by the photovoltaic panel dust accumulation recognition model; i represents the number index of photovoltaic panel dust accumulation categories;

[0076] Step S4: Input the training set into the photovoltaic panel dust accumulation recognition model in batches, perform forward propagation to calculate the prediction value for each batch, and calculate the loss through the loss function. Through iterative training, until the loss value converges or reaches the preset stop condition, a trained photovoltaic panel dust accumulation recognition model is obtained.

[0077] After completing the photovoltaic panel dust recognition model training, it is necessary to use the test set to evaluate the trained photovoltaic panel dust recognition model and monitor the overfitting situation. The cross-validation method can be used to further improve the robustness of the model. In order to improve the recognition effect of the model, a transfer learning strategy can be used. Fine-tune based on pre-trained networks (such as ResNet, VGG, etc.), and use the feature representations obtained by these networks trained on large-scale data sets (such as ImageNet) to improve the performance of the model in the case of small sample learning.

[0078] The category labels in step S1 include: no dust accumulation, light dust accumulation, moderate dust accumulation, and heavy dust accumulation.

[0079] The method of using the trained photovoltaic panel dust accumulation recognition model to identify dust accumulation on the surface of the photovoltaic panel and classifying the dust accumulation on the identified photovoltaic panel surface specifically includes:

[0080] Step A1: converting the preprocessed surface image of the photovoltaic panel into the HSV color space and performing standardization processing;

[0081] Step A2: Use the deep learning model to detect the occluders in the surface image of the photovoltaic panel, and obtain the occluder set D in the surface image of the photovoltaic panel, D = {(zb1,c1,p1),(zb2,c2,p2),...,(zb m ,c m ,p m )}, where zb m represents the coordinates of the mth occluder in the surface image of the photovoltaic panel, c m Represents the category of the mth occluder in the surface image of the photovoltaic panel, usually represented as a category label, p m Represents the confidence of the mth occluder in the surface image of the photovoltaic panel, that is, the probability of belonging to this category;

[0082] In this embodiment, the deep learning model here is a target detection model, that is, a model established by a target detection algorithm. During the detection process, it is necessary to extract a feature map of the surface image of the photovoltaic panel for use in subsequent steps;

[0083] Step A3: Use the regression model to predict the bounding box position of the object in the surface image of the photovoltaic panel, and combine the surface image feature map of the photovoltaic panel extracted by the deep learning model during detection to output the offset (Δx, Δy, Δw, Δh) of each occlusion candidate box, where Δx represents the offset of the horizontal coordinate of the occlusion candidate box, Δy represents the offset of the vertical coordinate of the occlusion candidate box, Δw represents the offset of the width of the occlusion candidate box, and Δh represents the offset of the height of the occlusion candidate box. The predicted bounding box is adjusted using the offset. The specific formula for adjusting the predicted bounding box is:

[0084]

[0085] in, represents the predicted bounding box position, b represents the true bounding box position, that is, the actual position of the occluder, and x center ,y center ,w,h represent the central horizontal coordinate, central vertical coordinate, width and height of the predicted occluder respectively. w * ,h * Indicates the actual center horizontal coordinate, center vertical coordinate, width and height of the occluder;

[0086] Step A4: Remove the obtained occluder prediction bounding box from the surface image of the photovoltaic panel, and use the mean filling method to repair the blank area after removing the occluder to obtain the initial repaired image I of the occluder prediction bounding box init (x, y), optimize the boundary of the occluder prediction bounding box. The specific formula is:

[0087] L boun =∑ (x',y')∈θ |▽I init (x',y')-▽I orig (x',y')|;

[0088] Among them, L boun represents the boundary difference between the initial repaired image of the occluder prediction bounding box and the surface image of the photovoltaic panel, that is, the difference in the boundary area of ​​the occluder prediction bounding box, (x', y') represents the pixel coordinates of the boundary area of ​​the occluder prediction bounding box, θ represents the boundary of the occluder prediction bounding box, that is, the boundary contour composed of all pixels in the repaired area. Usually, the repaired area is marked by the mask image M, and the boundary is the boundary line from the repaired area to the non-repaired area in the mask image. |·| represents the absolute value function, ▽I init (x', y') represents the gradient of the initial repaired image of the predicted bounding box of the occluder at pixel (x', y'), ▽I orig(x', y') represents the gradient of the surface image of the photovoltaic panel at the pixel (x', y');

[0089] In this embodiment, ▽I init (x', y') represents the intensity change at that position, which usually reflects important features such as texture and edge in the image. orig (x', y') represents the changes in texture and edges in the surface image of the photovoltaic panel, which is usually used to maintain image details. boun The repair effect is measured by calculating the gradient difference between the repair area and the surface image of the photovoltaic panel at the boundary position, which is used to minimize the difference between the gradient of the repair area and the gradient of the surface image of the photovoltaic panel, making the transition of the repair area smoother and more natural;

[0090] The basic principle of boundary optimization is to guide the filling of the repair area through gradient differences, so that the transition between the repair area and the surrounding area is smoother and more natural. In the image repair process, the boundary between the repair area and the unobstructed area in the original image is a key visual part. If the boundary between the repair area and the surrounding area is too abrupt, it may cause the repair result to be unnatural and produce obvious artifacts. The gradient reflects the speed of change in the image, that is, the degree of change in image brightness or color. In natural images, boundaries usually correspond to important structural information in the image, such as object edges, texture changes, etc. If the boundary of the repair area is very different from the boundary of the original image in gradient, the repaired part will look very abrupt. Therefore, by minimizing the gradient difference, the boundary of the repair area can be made smoother, making the repaired image more consistent with the original image;

[0091] For example, suppose there is an occluded leaf area in the surface image of a photovoltaic panel, and this area is filled by the repair algorithm. During the repair process, the transition between the repaired area and the surrounding photovoltaic panel background is the key. By minimizing the gradient difference, the repaired leaf area will not produce obvious jagged or unnatural texture transitions at the boundary with the background, but will blend into the background, giving a continuous visual effect;

[0092] Step A5: Input the surface image of the repaired photovoltaic panel into the trained photovoltaic panel dust accumulation recognition model to obtain the degree of dust accumulation of the photovoltaic panel.

[0093] The dust accumulation area extraction module is used to segment the dust accumulation area after obtaining the dust accumulation degree of the photovoltaic panel, and extract the dust accumulation area in the photovoltaic panel;

[0094] The dust accumulation area is segmented to extract the dust accumulation area in the photovoltaic panel, and the specific steps include:

[0095] Step B1: Set a local adaptive threshold and perform region segmentation based on the brightness difference of the surface image of the repaired photovoltaic panel. The specific formula of the local adaptive threshold is:

[0096] T(x,y)=μ(x,y)-α×σ(x,y);

[0097] Among them, T(x,y) represents the local adaptive threshold, that is, the adaptive threshold of the pixel point (x,y), μ(x,y) represents the local average brightness of the neighborhood around the pixel point (x,y), σ(x,y) represents the standard deviation of the neighborhood around the pixel point (x,y), α represents the adjustment factor, which is used to control the sensitivity of the threshold and determines the distance between the threshold and the local average brightness μ(x,y) under a given standard deviation σ(x,y). Adjusting α can balance the segmentation accuracy of the dust accumulation area and the non-dust accumulation area; (x,y) represents the pixel point in the surface image of the repaired photovoltaic panel;

[0098] For example, if α is large, the threshold T(x, y) will be farther from the average brightness μ(x, y), which may result in too many dusty areas being segmented; conversely, if α is small, some dusty areas may be missed;

[0099] The surrounding neighborhood is usually a window (such as a 3×3 or 5×5 pixel area), in which the average brightness is calculated. By calculating the local mean, it is possible to adapt to the brightness changes under different lighting conditions in the image and reduce the errors that may be caused by the global threshold method;

[0100] The standard deviation of the neighborhood around the pixel point (x, y) describes the discrete degree of brightness in the area. The larger the standard deviation, the greater the brightness change in the area, and vice versa. The standard deviation can capture the texture changes in the image and help distinguish between dusty areas (usually with small brightness changes) and non-dusty areas (with large brightness changes).

[0101] The brightness here is the grayscale value of the surface image of the repaired photovoltaic panel;

[0102] Step B2: Based on the local adaptive threshold T(x,y), the surface image of the repaired photovoltaic panel is divided into a dust accumulation area and a non-dust accumulation area. The pixel value corresponding to the dust accumulation area is greater than T(x,y), while the pixel value corresponding to the non-dust accumulation area is less than T(x,y).

[0103] The intelligent cleaning module dynamically plans a cleaning path according to the distribution of dust accumulation areas to clean the photovoltaic panels.

[0104] Intelligent cleaning module, including: intelligent planning unit and cleaning unit;

[0105] The intelligent planning unit dynamically plans the cleaning path according to the distribution of dust accumulation areas;

[0106] The cleaning unit cleans the surface of the photovoltaic panel according to the planned path.

[0107] Example 2

[0108] See also Figure 2 Another embodiment provided by the present invention is a photovoltaic panel dust accumulation identification and cleaning auxiliary method based on deep learning, and the specific steps include:

[0109] Step 1: Collect the surface image of the photovoltaic panel;

[0110] Step 2: Preprocessing the collected surface images of the photovoltaic panels, including denoising, image enhancement and image segmentation operations;

[0111] Step 3: Build a photovoltaic panel dust accumulation recognition model, train the photovoltaic panel dust accumulation recognition model, identify the dust accumulation of the photovoltaic panel, and quantify and classify the dust accumulation level;

[0112] Step 4: After obtaining the dust accumulation degree of the photovoltaic panel, the dust accumulation area is segmented to extract the dust accumulation area in the photovoltaic panel;

[0113] Step 5: According to the distribution of dust accumulation areas, dynamically plan the cleaning path and clean the photovoltaic panels.

[0114] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0115] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning, characterized in that: include: Image acquisition module, image preprocessing module, dust accumulation degree analysis module, dust accumulation area extraction module and intelligent cleaning module; The image acquisition module uses a drone or ground-based camera to acquire surface images of the photovoltaic panels; The image preprocessing module is used to preprocess the collected surface image of the photovoltaic panel, including denoising, image enhancement and image segmentation operations; The dust accumulation degree analysis module is used to construct a photovoltaic panel dust accumulation recognition model, train the photovoltaic panel dust accumulation recognition model, identify the dust accumulation of the photovoltaic panel, and quantify and classify the dust accumulation level; The dust accumulation area extraction module is used to segment the dust accumulation area after obtaining the dust accumulation degree of the photovoltaic panel, and extract the dust accumulation area in the photovoltaic panel; The intelligent cleaning module dynamically plans a cleaning path according to the distribution of dust accumulation areas to clean the photovoltaic panels.

2. A photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning as claimed in claim 1, characterized in that: The dust accumulation degree analysis module includes: a model training unit and a dust accumulation identification unit; The model training unit is used to construct a photovoltaic panel dust accumulation recognition model and perform training; The dust accumulation recognition unit uses the trained photovoltaic panel dust accumulation recognition model to recognize dust accumulation on the surface of the photovoltaic panel and classifies the recognized dust accumulation on the surface of the photovoltaic panel.

3. A photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning as claimed in claim 2, characterized in that: The photovoltaic panel dust accumulation recognition model is constructed and trained, including: Obtain a public image dataset of photovoltaic panel dust accumulation, attach category labels to the photovoltaic panel dust accumulation images in the public image dataset of photovoltaic panel dust accumulation, and divide the public image dataset of photovoltaic panel dust accumulation into a training set and a test set in a ratio of 7:3; Construct a photovoltaic panel dust accumulation recognition model, including: a convolution layer, used to extract local features in the photovoltaic panel dust accumulation image; a pooling layer, used for downsampling to reduce the data dimension and calculation amount; a fully connected layer, used to map the photovoltaic panel dust accumulation image features to the final category output; an activation function, used to output the final classification result; Define the loss function of the photovoltaic panel dust accumulation recognition model. The specific formula is: Among them, Γ represents the loss function of the photovoltaic panel dust accumulation recognition model, C represents the number of photovoltaic panel dust accumulation categories, and yc i represents the true label of the photovoltaic panel dust accumulation image in the photovoltaic panel dust accumulation public image dataset, yc i ' represents the probability predicted by the photovoltaic panel dust accumulation recognition model; i represents the number index of photovoltaic panel dust accumulation categories; The training set is input into the photovoltaic panel dust accumulation recognition model in batches. For each batch, forward propagation is performed to calculate the prediction value, and the loss is calculated through the loss function. Through iterative training, until the loss value converges or reaches the preset stopping condition, a trained photovoltaic panel dust accumulation recognition model is obtained.

4. A photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning as claimed in claim 3, characterized in that: The category labels include: no dust accumulation, light dust accumulation, moderate dust accumulation, and heavy dust accumulation.

5. A photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning as claimed in claim 4, characterized in that: The method of using the trained photovoltaic panel dust accumulation recognition model to identify dust accumulation on the photovoltaic panel surface and classifying the identified dust accumulation on the photovoltaic panel surface includes: The pre-processed surface image of the photovoltaic panel is converted into the HSV color space and standardized; The deep learning model is used to detect the occluders in the surface image of the photovoltaic panel, and the occluder set D in the surface image of the photovoltaic panel is obtained, D = {(zb1,c1,p1),(zb2,c2,p2),...,(zb m ,c m ,p m )}, where zb m represents the coordinates of the mth occluder in the surface image of the photovoltaic panel, c m represents the category of the mth occluder in the surface image of the photovoltaic panel, p m Represents the confidence of the mth occluder in the surface image of the photovoltaic panel; The regression model is used to predict the bounding box position of the object in the surface image of the photovoltaic panel. Combined with the surface image feature map of the photovoltaic panel extracted by the deep learning model during detection, the offset (Δx, Δy, Δw, Δh) of each occlusion candidate box is output, where Δx represents the offset of the horizontal coordinate of the occlusion candidate box, Δy represents the offset of the vertical coordinate of the occlusion candidate box, Δw represents the offset of the width of the occlusion candidate box, and Δh represents the offset of the height of the occlusion candidate box. The predicted bounding box is adjusted using the offset. The specific formula for adjusting the predicted bounding box is: in, represents the predicted bounding box position, b represents the true bounding box position, that is, the actual position of the occluder, and x center ,y center ,w,h represent the central horizontal coordinate, central vertical coordinate, width and height of the predicted occluder respectively. w * ,h * Indicates the actual center horizontal coordinate, center vertical coordinate, width and height of the occluder.

6. A photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning as claimed in claim 5, characterized in that: The method of using the trained photovoltaic panel dust accumulation recognition model to identify dust accumulation on the photovoltaic panel surface and classifying the identified dust accumulation on the photovoltaic panel surface also includes: The obtained occluder prediction bounding box is removed from the surface image of the photovoltaic panel, and the blank area after removing the occluder is repaired using the mean filling method to obtain the initial repaired image I of the occluder prediction bounding box init (x, y), optimize the boundary of the occluder prediction bounding box. The specific formula is: Among them, L boun represents the boundary difference between the initial repaired image of the occluder prediction bounding box and the surface image of the photovoltaic panel, that is, the difference in the boundary area of ​​the occluder prediction bounding box, (x', y') represents the pixel coordinates of the boundary area of ​​the occluder prediction bounding box, θ represents the boundary of the occluder prediction bounding box, that is, the boundary contour composed of all pixels in the repair area, |·| represents the absolute value function, ▽I init (x', y') represents the gradient of the initial repaired image of the predicted bounding box of the occluder at pixel (x', y'), ▽I orig (x', y') represents the gradient of the surface image of the photovoltaic panel at the pixel (x', y'); The surface image of the repaired photovoltaic panel is input into the trained photovoltaic panel dust accumulation recognition model to obtain the degree of dust accumulation of the photovoltaic panel.

7. A photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning as claimed in claim 6, characterized in that: The step of segmenting the dust accumulation area to extract the dust accumulation area in the photovoltaic panel includes: The local adaptive threshold is set to perform region segmentation based on the brightness difference of the surface image of the repaired photovoltaic panel. The specific formula of the local adaptive threshold is: T(x,y)=μ(x,y)-α×σ(x,y), Wherein, T(x,y) represents the local adaptive threshold, i.e., the adaptive threshold of the pixel point (x,y), μ(x,y) represents the local average brightness of the neighborhood around the pixel point (x,y), σ(x,y) represents the standard deviation of the neighborhood around the pixel point (x,y), and α represents the adjustment factor; (x,y) represents the pixel point in the surface image of the repaired photovoltaic panel; Based on the local adaptive threshold T(x,y), the surface image of the repaired photovoltaic panel is divided into dust accumulation area and non-dust accumulation area. The pixel value corresponding to the dust accumulation area is greater than T(x,y), while the pixel value corresponding to the non-dust accumulation area is less than T(x,y).

8. A photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning as claimed in claim 7, characterized in that: The collected surface image of the photovoltaic panel is preprocessed, including denoising, image enhancement and image segmentation operations, denoising, using a Gaussian filter algorithm to remove noise in the surface image of the photovoltaic panel; Image enhancement is used to adjust the brightness or contrast of the surface image of the photovoltaic panel; image segmentation operation is used to segment the photovoltaic panel area to distinguish the background from the photovoltaic panel area.

9. A photovoltaic panel dust accumulation identification and cleaning auxiliary system based on deep learning as claimed in claim 8, characterized in that: The intelligent cleaning module comprises: an intelligent planning unit and a cleaning unit; The intelligent planning unit dynamically plans the cleaning path according to the distribution of dust accumulation areas; The cleaning unit cleans the surface of the photovoltaic panel according to the planned path.

10. A photovoltaic panel dust accumulation identification and cleaning assistance method based on deep learning, implemented based on a photovoltaic panel dust accumulation identification and cleaning assistance system based on deep learning as described in any one of claims 1-9, characterized in that: The specific steps include: Step 1: Collect the surface image of the photovoltaic panel; Step 2: Preprocessing the collected surface images of the photovoltaic panels, including denoising, image enhancement and image segmentation operations; Step 3: Build a photovoltaic panel dust accumulation recognition model, train the photovoltaic panel dust accumulation recognition model, identify the dust accumulation of the photovoltaic panel, and quantify and classify the dust accumulation level; Step 4: After obtaining the dust accumulation degree of the photovoltaic panel, the dust accumulation area is segmented to extract the dust accumulation area in the photovoltaic panel; Step 5: According to the distribution of dust accumulation areas, dynamically plan the cleaning path and clean the photovoltaic panels.

Images