Photovoltaic module cleaning method and device based on image processing and electronic equipment

Abstract

The invention discloses a photovoltaic module cleaning method and device based on image processing and electronic equipment, and belongs to the technical field of photovoltaic power generation. The method comprises the following steps: acquiring operation information, a visible light image, a near-infrared image and thermal imaging data of a photovoltaic module; the operation information, the visible light image, the thermal imaging data and the near-infrared image are input into a hot spot detection model, a hot spot detection result, output by the hot spot detection model, of the photovoltaic module is obtained, the hot spot detection result comprises a hot spot area and a hot spot level, and the hot spot detection model is obtained based on a plurality of first training samples; the first training sample comprises sample operation information, a sample visible light image, sample thermal imaging data, a sample near-infrared image and a hot spot detection label of the sample photovoltaic module; acquiring environment information and dust information of dust attached to the photovoltaic module, wherein the dust information comprises a dust type and an attachment form; the environment information, the dust information and the hot spot detection result are input into a dust cleaning prediction model, cleaning parameters of each hot spot output by the dust cleaning prediction model are obtained, and the dust cleaning prediction model is obtained based on a plurality of second training samples; the second training sample comprises sample environment information, sample dust information, a hot spot detection label and a cleaning parameter label of the sample photovoltaic module; and cleaning the photovoltaic module based on the cleaning parameters. According to the method, different types of dust can be effectively cleaned, and the dust cleaning effect of the photovoltaic module is improved.

Description

Technical Field

[0001] The present application belongs to the field of photovoltaic power generation technology, and in particular relates to a method, device and electronic equipment for cleaning photovoltaic components based on image processing. Background Art

[0002] Photovoltaic power generation is a technology that converts solar energy into electrical energy. Currently, due to the strong solar radiation and high annual power generation in desert areas, many countries around the world are vigorously developing desert photovoltaics. Combined with energy storage technology, it has become an important source of clean energy. However, in desert environments, there is a lot of sand and dust. In solar power generation, the dust (dust, sand, etc.) covering the surface of photovoltaic modules will hinder the incidence of light, resulting in the cells in the blocked areas being unable to generate electricity normally. These areas will instead become loads, consuming electricity from other areas, and thus causing local high temperatures. In addition, the dust layer will also hinder the heat dissipation of the module surface, causing temperature accumulation and the formation of hot spots. Hot spots will not only cause the power output of the cells to decrease, but may also accelerate the aging of the packaging materials of the photovoltaic modules, causing the solder joints to melt. In severe cases, it may even cause the glass to break or the module to burn, posing certain safety risks.

[0003] In order to avoid the safety risks brought by hot spots, existing technologies formulate cleaning strategies by comparing the predicted temperature of photovoltaic modules with the actual temperature, and spray cooling the photovoltaic modules according to the actual temperature.

[0004] However, the above method only relies on temperature data to clean and cool the hot spot area, and cannot effectively clean different types of dust, and the cleaning effect is average. Summary of the Invention

[0005] This application aims to solve at least one of the technical problems existing in the prior art. To this end, this application proposes a method, device, and electronic device for cleaning photovoltaic modules based on image processing, which can effectively clean different types of dust and improve the dust cleaning effect of photovoltaic modules.

[0006] In a first aspect, the present application provides a method for cleaning a photovoltaic module based on image processing, the method comprising:

[0007] Obtaining operational information, visible light images, near-infrared images, and thermal imaging data of photovoltaic modules;

[0008] Inputting the operating information, the visible light image, the thermal imaging data, and the near-infrared image into a hot spot detection model, obtaining a hot spot detection result of the photovoltaic module output by the hot spot detection model, the hot spot detection result including a hot spot area and a hot spot level, wherein the hot spot detection model is obtained based on a plurality of first training samples, wherein the first training samples include sample operating information, sample visible light images, sample thermal imaging data, sample near-infrared images, and hot spot detection labels of sample photovoltaic modules;

[0009] Acquiring environmental information and dust information of dust attached to the photovoltaic module, wherein the dust information includes dust type and attachment form;

[0010] Inputting the environmental information, the dust information, and the hot spot detection result into a dust cleaning prediction model to obtain a cleaning parameter for each hot spot output by the dust cleaning prediction model, wherein the dust cleaning prediction model is obtained based on a plurality of second training samples, the second training samples including sample environmental information, sample dust information, hot spot detection labels, and cleaning parameter labels of the sample photovoltaic modules;

[0011] The photovoltaic component is cleaned based on the cleaning parameters.

[0012] According to one embodiment of the present application, the hot spot detection model includes a feature extraction module, a feature fusion module, and an output module connected in sequence. The operation information, the visible light image, the thermal imaging data, and the near-infrared image are input into the hot spot detection model to obtain the hot spot detection result of the photovoltaic module output by the hot spot detection model, including:

[0013] By means of the feature extraction module, feature extraction is performed on the operation information, the visible light image, the thermal imaging data, and the near-infrared image to obtain corresponding multiple first features;

[0014] By means of the feature fusion module, the plurality of first features are subjected to spatiotemporal alignment and data fusion to obtain a fused feature;

[0015] The output module performs region segmentation and temperature regression on the fused features, and the obtained hot spot regions and hot spot levels are mapped to the visible light image and output as the hot spot detection results.

[0016] According to one embodiment of the present application, the feature extraction module includes a ResNet-50 network, a 3D convolutional network, a U-Net network, and an LSTM network. The multiple first features include texture features, temperature gradient features, hot spot area features, and abnormal fluctuation features of the dust. The feature extraction module performs feature extraction on the operation information, the visible light image, the thermal imaging data, and the near-infrared image, respectively, to obtain corresponding multiple first features, including:

[0017] Extracting the texture features of the dust from the visible light image using the ResNet-50 network;

[0018] The temperature gradient features extracted from the thermal imaging data by the 3D convolutional network;

[0019] Segmenting the hot spot area in the near-infrared image according to reflectivity using the U-Net network to obtain features of the hot spot area;

[0020] The abnormal fluctuation characteristics of the photovoltaic components are extracted from the operating information through the LSTM network.

[0021] According to one embodiment of the present application, a dust occlusion heat conduction model is embedded in the 3D convolutional network, and the dust occlusion heat conduction model includes:

[0022]

[0023] Where ΔT is the temperature rise of the PV module; is the photoelectric-thermal coupling term, J shunt is the bypass current density in the hot spot area, which is used to characterize the local current concentration effect caused by hidden cracks or contamination. cell (T) is the temperature-dependent resistance of the cell, which is used to reflect the negative temperature coefficient characteristics of the semiconductor material of the photovoltaic module, d dust is the effective thickness of the dust layer, The humidity-temperature dependent equivalent thermal conductivity is calculated. The heat conduction path is corrected by connecting the glass and dust heat conduction models in series. The humidity effect is dominated by the capillary condensation effect. is the dust porosity, which indicates the volume ratio of air or other pores in the dust layer, T is the local temperature, α h RH is the humidity correction factor, which is used to quantify the enhancement of heat conduction by the water film formed after the dust pores absorb moisture. h is the capillary condensation enhancement coefficient, which is used to ensure that the interface effect of nano-scale dust particles amplifies the humidity effect, and RH is the relative humidity; is the environmental dissipation term, P lossis the power loss per unit area, which is caused by the decrease in photoelectric conversion efficiency due to local shading. β(T) is the hot spot self-shielding factor, which is used to characterize the negative feedback suppression of the heat generation power caused by the intensified carrier recombination in the high temperature area. conv +h red is the comprehensive heat dissipation coefficient, h conv is the wind speed dependent convective heat transfer term, h red is the radiation heat transfer term; is the time dynamic term, τ(T) is the temperature-dependent relaxation time, which is used to characterize the nonlinear characteristics of heat capacity. is the temperature gradient dissipation correction term, which is used to suppress the violent oscillation of heat flow. γ is the gradient dissipation coefficient, which is used to suppress the violent oscillation of heat flow. is the temperature gradient.

[0024] According to one embodiment of the present application, a wind field-dust diffusion coupling model is embedded in the feature fusion module, and the wind field-dust diffusion coupling model includes:

[0025]

[0026] Boundary conditions on the surface of photovoltaic modules:

[0027]

[0028] in, is the convection term, C is the dust concentration field, which is used to characterize the dynamic spatial distribution of dust mass concentration per unit volume of air, t is time, u is the three-dimensional wind speed field, which is used to describe the air flow around the photovoltaic module, and D turb is the turbulent diffusion coefficient, which is related to the wind speed and the surface roughness of the photovoltaic module, k dep is the sedimentation rate constant, which is obtained based on the effects of gravity sedimentation, electrostatic adsorption and inertial collision. resusp is the dust resuspension source term, n is the boundary normal direction perpendicular to the photovoltaic panel surface, k adh is the surface adhesion coefficient of the photovoltaic module, k detach is the re-flying threshold, which is used to characterize the probability of dust detaching from the surface at the critical wind speed, d dust is the effective thickness of the dust layer.

[0029] According to one embodiment of the present application, the output module includes a hot spot detection head and a dust attribution head. The output module performs region segmentation and temperature regression on the fused features, maps the obtained hot spot regions and hot spot levels to the visible light image, and outputs them as the hot spot detection results, including:

[0030] The hot spot detection head performs pixel-level segmentation on the fusion features to generate a hot spot area marked by a binary mask, and generates a hot spot level of the hot spot area based on the coupling relationship between the local temperature rise and the current density in the hot spot area;

[0031] The dust attribution head quantifies the contribution of dust deposition to the temperature rise of hot spots based on the above-mentioned fusion features, generates an attribution heat map, corrects the hot spot area, maps the hot spot area and hot spot level from the infrared coordinate system to the visible light image, and generates hot spot detection results.

[0032] According to one embodiment of the present application, the dust cleaning prediction model includes a data input layer, a feature extraction layer, a prediction layer, and an output decision layer connected in sequence. The environmental information, the dust information, and the hot spot detection results are input into the dust cleaning prediction model to obtain the cleaning parameters of each hot spot output by the dust cleaning prediction model, including:

[0033] Through the data input layer, the environmental information, the dust information and the hot spot detection results are integrated to generate multi-source heterogeneous data;

[0034] By means of the feature extraction layer, the multi-source heterogeneous data is analyzed for dimensionality reduction and key influencing factors are screened to obtain a plurality of second features;

[0035] Processing the time series dependency of the second feature through the prediction layer to generate a prediction result;

[0036] The prediction results are mapped from the feature space to the parameter space through the output decision layer, and the cleaning parameters of each hot spot are generated and output.

[0037] According to one embodiment of the present application, after cleaning the photovoltaic assembly based on the cleaning parameters, the method further includes:

[0038] Calculating a first power generation efficiency of the photovoltaic assembly according to a first IV curve in the operation information;

[0039] Collect the second IV curve and feedback thermal imaging data of the cleaned photovoltaic modules;

[0040] calculating a second power generation efficiency of the photovoltaic assembly according to the second IV curve;

[0041] Cleaning of the photovoltaic assembly is optimized according to the thermal imaging data, the feedback thermal imaging data, and a deviation between the first power generation efficiency and the second power generation efficiency.

[0042] In a second aspect, the present application provides a photovoltaic module cleaning device based on image processing, the device comprising:

[0043] The first acquisition module is used to acquire operating information, visible light images, near-infrared images and thermal imaging data of photovoltaic modules;

[0044] a first processing module, configured to input the operating information, the visible light image, the thermal imaging data, and the near-infrared image into a hot spot detection model, and obtain a hot spot detection result of the photovoltaic module output by the hot spot detection model, wherein the hot spot detection result includes a hot spot area and a hot spot level, and the hot spot detection model is obtained based on a plurality of first training samples, wherein the first training samples include sample operating information, sample visible light images, sample thermal imaging data, sample near-infrared images, and hot spot detection labels of sample photovoltaic modules;

[0045] A second acquisition module is configured to acquire environmental information and dust information of dust attached to the photovoltaic module, wherein the dust information includes dust type and attachment form;

[0046] a second processing module, configured to input the environmental information, the dust information, and the hot spot detection result into a dust cleaning prediction model to obtain a cleaning parameter for each hot spot output by the dust cleaning prediction model, wherein the dust cleaning prediction model is obtained based on a plurality of second training samples, the second training samples including sample environmental information, sample dust information, hot spot detection labels, and cleaning parameter labels of the sample photovoltaic modules;

[0047] The third processing module is configured to clean the photovoltaic assembly based on the cleaning parameters.

[0048] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method for cleaning photovoltaic components based on image processing as described in the first aspect above is implemented.

[0049] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the photovoltaic component cleaning method based on image processing as described in the first aspect above.

[0050] In a fifth aspect, the present application provides a chip comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is configured to run programs or instructions to implement the method for cleaning a voltaic component as described in the first aspect.

[0051] In a sixth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the photovoltaic component cleaning method based on image processing as described in the first aspect above.

[0052] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become obvious from the description below, or will be learned through practice of the present application.

[0053] The present application provides a photovoltaic module cleaning method, device, and electronic device based on image processing, which have the following beneficial effects compared to the prior art:

[0054] (1) Through the comprehensive analysis of hot spot detection, environmental information, dust information and cleaning parameters, efficient, accurate and targeted cleaning of photovoltaic modules is achieved. With the assistance of hot spot detection and dust cleaning prediction model, the cleaning process can be intelligently guided, the cleaning efficiency of photovoltaic modules can be improved, and their service life can be extended. At the same time, unnecessary cleaning work can be reduced, thereby optimizing the maintenance process of photovoltaic modules, being able to effectively clean different types of dust, and improving the dust cleaning effect of photovoltaic modules.

[0055] (2) Through feature extraction and fusion of multi-source information, spatiotemporal alignment, regional segmentation and temperature regression processing, high-precision and high-reliability hot spot detection is achieved, which effectively improves the detection accuracy and visualization effect of hot spots, provides strong technical support for the monitoring and maintenance of photovoltaic modules, and improves detection efficiency and decision support capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0057] Figure 1 1 is a flow chart of a method for cleaning a photovoltaic module based on image processing provided in an embodiment of the present application;

[0058] Figure 2 Schematic diagram of the structure of the hot spot detection model provided in the embodiment of the present application;

[0059] Figure 3 It is a structural diagram of the dust cleaning prediction model provided in the embodiment of the present application;

[0060] Figure 4 1 is a schematic structural diagram of a photovoltaic module cleaning device based on image processing provided in an embodiment of the present application;

[0061] Figure 5 It is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0062] The following will be combined with the accompanying drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the embodiments described are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of this application.

[0063] The terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects, and are not used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of this application can be implemented in an order other than that illustrated or described herein, and that the objects distinguished by "first," "second," and the like are generally of the same type, and do not limit the number of objects; for example, the first object can be one or more. In addition, the term "and / or" in the specification and claims refers to at least one of the connected objects, and the character " / " generally indicates that the objects connected are in an "or" relationship.

[0064] The following, in combination with the accompanying drawings, describes in detail the photovoltaic component cleaning method based on image processing, the photovoltaic component cleaning device based on image processing, the electronic device and the readable storage medium provided in the embodiments of the present application through specific embodiments and their application scenarios.

[0065] The photovoltaic component cleaning method based on image processing can be applied to a terminal, and can be specifically executed by hardware or software in the terminal.

[0066] The terminal includes, but is not limited to, a portable communication device such as a mobile phone or tablet computer having a touch-sensitive surface (e.g., a touch screen display and / or a touch pad). It should also be understood that, in some embodiments, the terminal may not be a portable communication device, but a desktop computer having a touch-sensitive surface (e.g., a touch screen display and / or a touch pad).

[0067] In the following embodiments, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, a mouse, and a joystick.

[0068] The embodiment of the present application provides a method for cleaning photovoltaic components based on image processing. The execution subject of the method for cleaning photovoltaic components based on image processing can be an electronic device or a functional module or functional entity in the electronic device that can implement the method for cleaning the photovoltaic components. The electronic devices mentioned in the embodiment of the present application include but are not limited to mobile phones, tablet computers, computers, cameras and wearable devices, etc. The method for cleaning photovoltaic components based on image processing provided in the embodiment of the present application is explained below using an electronic device as an example of the execution subject.

[0069] like Figure 1 As shown, the photovoltaic module cleaning method based on image processing includes:

[0070] Step 110: Acquire operating information, visible light images, near infrared images, and thermal imaging data of the photovoltaic modules;

[0071] Step 120: Input the operating information, the visible light image, the thermal imaging data, and the near-infrared image into a hot spot detection model to obtain a hot spot detection result of the photovoltaic module output by the hot spot detection model, wherein the hot spot detection result includes a hot spot area and a hot spot level. The hot spot detection model is obtained based on a plurality of first training samples, wherein the first training samples include sample operating information, sample visible light images, sample thermal imaging data, sample near-infrared images, and hot spot detection labels of sample photovoltaic modules.

[0072] Step 130: Acquire environmental information and dust information of dust attached to the photovoltaic module, wherein the dust information includes dust type and attachment form;

[0073] Step 140: Input the environmental information, the dust information, and the hot spot detection result into a dust cleaning prediction model to obtain a cleaning parameter for each hot spot output by the dust cleaning prediction model, wherein the dust cleaning prediction model is obtained based on a plurality of second training samples, wherein the second training samples include sample environmental information, sample dust information, hot spot detection labels, and cleaning parameter labels of the sample photovoltaic modules;

[0074] Step 150: Clean the photovoltaic module based on the cleaning parameters.

[0075] Among them, the operating information includes the first IV curve that characterizes the volt-ampere characteristics of the photovoltaic module. The IV curve can reflect the current output of the photovoltaic module under different voltages, thereby judging whether its power generation performance is normal and providing basic data for subsequent hot spot detection.

[0076] Visible light images, near-infrared images, and thermal imaging data are collected by drones on photovoltaic panels during inspections.

[0077] It is understandable that visible light images intuitively present visible pollutants on the surface of photovoltaic modules; in near-infrared images, the reflectivity of materials such as photovoltaic modules' cells and glass cover plates is different from that of dust in the near-infrared band, which enables near-infrared images to more clearly distinguish between dust and photovoltaic modules themselves, helping to accurately identify the location and range of dust; thermal imaging data is used to monitor the temperature distribution on the surface of photovoltaic modules, providing a key basis for hot spot detection.

[0078] The hot spot detection result is a visible light image marked with the hot spot area and hot spot level, where the hot spot area is the coordinate position and range of the hot spot on the photovoltaic module. The hot spot level is determined based on the module temperature of the photovoltaic module and the hot spot temperature difference threshold, including mild, moderate and severe. The corresponding hot spot temperature difference threshold is dynamically set or adjusted according to the model of the photovoltaic module and the ambient temperature.

[0079] Environmental information includes light intensity, precipitation information and ambient temperature. This information can reflect the weather conditions during cleaning and provide a reference for selecting appropriate cleaning methods and time for photovoltaic modules.

[0080] Dust types can be identified through polarized reflectance spectral features. For example, a liquid crystal tunable filter (LCTF) can be added to a visible light camera to analyze the polarized reflectance spectrum of the dust to identify its chemical composition. This can be used as the dust type, for example, the encrusted type formed by cement solidification, salt and alkali, the loose type formed by sand, dust, pollen, etc., and the oily adherent type formed by industrial emissions.

[0081] The attachment form of dust on photovoltaic modules, such as uniform distribution or local agglomeration, is obtained by target recognition of dust in visible light images.

[0082] The cleaning parameters include the number of cleaning methods corresponding to different hot spot areas, such as dry cleaning, high-pressure water gun, robot scraping, chemical cleaning, and atomized spray cleaners; the number of cleaning tools, such as soft brushes, high-pressure water guns, and drone spray devices; cleaning frequency; the number of cleaning agent types; cleaning priority; priority treatment of areas with severe hot spot levels; priority cleaning of high-yield components during peak electricity price periods; and recommended cleaning time periods, for example, when the irradiance is <200W / m 2 period, or within 2 hours after sunset, to effectively avoid evaporation residue during the cleaning process, as well as mark special areas, etc.

[0083] Among them, the special area is the hot spot area that is not suitable for any cleaning method. The special area is marked as the key area for inspection and manual verification.

[0084] In actual implementation, the first IV curve and the ambient temperature are acquired in real time through a monitoring device connected to the photovoltaic module.

[0085] Using drones equipped with visible light cameras, near-infrared cameras and thermal imagers, photovoltaic modules are photographed from multiple angles during the inspection process to obtain visible light images, near-infrared images and thermal imaging data of the photovoltaic modules.

[0086] The collected operational information, visible light images, thermal imaging data, and near-infrared images are formatted and aligned to ensure data accuracy and consistency.

[0087] The processed data is input into a pre-trained hot spot detection model. The model is trained based on a large number of first training samples and can automatically learn and identify the characteristic patterns of hot spots, thereby accurately determining whether there are hot spots in the photovoltaic module, and determining the specific area and level of the hot spots. The hot spot area and hot spot level are marked on the visible light image and output as the hot spot detection results of the photovoltaic module.

[0088] For example, if the temperature of a certain area of a photovoltaic module is significantly higher than that of the surrounding area and the temperature difference exceeds a set threshold, it is determined that a hot spot exists in the area, and the hot spot level is determined based on the size of the temperature difference.

[0089] Environmental information such as light intensity, precipitation information and ambient temperature can be obtained through weather stations or sensors installed in photovoltaic power plants.

[0090] The type of dust attached to photovoltaic modules is identified through polarized reflectance spectral characteristics; by analyzing the attachment form of dust, such as loose accumulation, tight adhesion, etc., a basis is provided for the generation of subsequent cleaning parameters.

[0091] After obtaining the hot spot detection results, the environmental information, dust information and hot spot detection results are integrated, and necessary data conversion and normalization are performed to make them meet the input requirements of the dust cleaning prediction model.

[0092] The processed data is fed into a dust cleaning prediction model. Based on a large number of secondary training samples, the model comprehensively considers the impact of various factors on cleaning effectiveness and predicts the optimal cleaning parameters for each hot spot. For example, for mild hot spots with loose dust, a dry cleaning solution is generated; for severe hot spots with oily, adherent dust, water washing and the addition of a cleaning agent may be required.

[0093] The model outputs the cleaning parameters for each hot spot. In addition, a preliminary assessment is conducted on the feasibility and effectiveness of the cleaning scheme corresponding to the cleaning parameters to ensure that it can effectively remove dust without causing damage to the PV panels.

[0094] When cleaning PV panels, choose cleaning equipment and tools that match the cleaning parameters. For example, for large PV arrays, use automated cleaning vehicles or cleaning robots for efficient cleaning. For hard-to-reach or heavily contaminated areas, use drones equipped with cleaning equipment for precise cleaning.

[0095] During the cleaning process, strictly follow the cleaning parameters. For example, when using water washing, set the water pressure, water temperature, water flow direction and other parameters according to the cleaning parameters to avoid impact damage to the photovoltaic modules.

[0096] In addition, when using cleaning agents, first test them on a small area to ensure that they are non-corrosive to the materials of the photovoltaic modules before applying them on a full scale.

[0097] After cleaning is complete, drones are used to capture images and data from the PV panels, comparing the changes before and after cleaning to evaluate the cleaning results. If the cleaning effect is not as expected, such as if there is still a lot of dust or hot spots in some areas, the cause is analyzed and cleaning parameters are adjusted, and a second cleaning is performed until the ideal cleaning effect is achieved, ensuring that the power generation efficiency of the PV panels is significantly improved and the risks caused by hot spots are avoided.

[0098] According to the image processing-based photovoltaic module cleaning method provided in the embodiment of the present application, efficient, accurate and targeted photovoltaic module cleaning is achieved through comprehensive analysis of hot spot detection, environmental information, dust information and cleaning parameters. With the assistance of hot spot detection and dust cleaning prediction models, the cleaning process can be intelligently guided, the cleaning efficiency of photovoltaic modules can be improved, and their service life can be extended. At the same time, unnecessary cleaning work can be reduced, thereby optimizing the maintenance process of photovoltaic modules, being able to effectively clean different types of dust, and improving the dust cleaning effect of photovoltaic modules.

[0099] In some embodiments, as Figure 2 As shown, the hot spot detection model includes a feature extraction module, a feature fusion module, and an output module connected in sequence. The operation information, the visible light image, the thermal imaging data, and the near-infrared image are input into the hot spot detection model to obtain the hot spot detection result of the photovoltaic module output by the hot spot detection model, including:

[0100] By means of the feature extraction module, feature extraction is performed on the operation information, the visible light image, the thermal imaging data, and the near-infrared image to obtain corresponding multiple first features;

[0101] By means of the feature fusion module, the plurality of first features are subjected to spatiotemporal alignment and data fusion to obtain a fused feature;

[0102] The output module performs region segmentation and temperature regression on the fused features, and the obtained hot spot regions and hot spot levels are mapped to the visible light image and output as the hot spot detection results.

[0103] In this embodiment, high-precision and high-reliability hot spot detection is achieved through feature extraction and fusion, spatiotemporal alignment, region segmentation, and temperature regression processing of multi-source information, which effectively improves the detection accuracy and visualization effect of hot spots, provides strong technical support for the monitoring and maintenance of photovoltaic modules, and improves detection efficiency and decision-making support capabilities.

[0104] In some embodiments, the feature extraction module includes a ResNet-50 network, a 3D convolutional network, a U-Net network, and an LSTM network. The multiple first features include texture features, temperature gradient features, hot spot area features, and abnormal fluctuation features of the dust. The feature extraction module performs feature extraction on the operation information, the visible light image, the thermal imaging data, and the near-infrared image, respectively, to obtain corresponding multiple first features, including:

[0105] Extracting the texture features of the dust from the visible light image using the ResNet-50 network;

[0106] The temperature gradient features extracted from the thermal imaging data by the 3D convolutional network;

[0107] Segmenting the hot spot area in the near-infrared image according to reflectivity using the U-Net network to obtain features of the hot spot area;

[0108] The abnormal fluctuation characteristics of the photovoltaic components are extracted from the operating information through the LSTM network.

[0109] Specifically, an FPN structure was added to the ResNet-50 network to capture hot spot characteristics at different scales. The heat conduction equation was added as a constraint term to the loss function to improve the physical rationality of the temperature regression.

[0110] In some embodiments, a dust occlusion heat conduction model is embedded in the 3D convolutional network, and the dust occlusion heat conduction model includes:

[0111]

[0112] Where ΔT is the temperature rise of the PV module; is the photoelectric-thermal coupling term, J shunt is the bypass current density in the hot spot area, which is used to characterize the local current concentration effect caused by hidden cracks or contamination. cell (T) is the temperature-dependent resistance of the cell, which is used to reflect the negative temperature coefficient characteristics of the semiconductor material of the photovoltaic module, d dust is the effective thickness of the dust layer, The humidity-temperature dependent equivalent thermal conductivity is calculated. The heat conduction path is corrected by connecting the glass and dust heat conduction models in series. The humidity effect is dominated by the capillary condensation effect. is the dust porosity, which indicates the volume ratio of air or other pores in the dust layer, T is the local temperature, α h RH is the humidity correction factor, which is used to quantify the enhancement of heat conduction by the water film formed after the dust pores absorb moisture. h is the capillary condensation enhancement coefficient, which is used to ensure that the interface effect of nano-scale dust particles amplifies the humidity effect, and RH is the relative humidity; is the environmental dissipation term, P loss is the power loss per unit area, which is caused by the decrease in photoelectric conversion efficiency due to local shading. β(T) is the hot spot self-shielding factor, which is used to characterize the negative feedback suppression of the heat generation power caused by the intensified carrier recombination in the high temperature area. conv +h red is the comprehensive heat dissipation coefficient, h conv is the wind speed dependent convective heat transfer term, h red is the radiation heat transfer term; is the time dynamic term, τ(T) is the temperature-dependent relaxation time, which is used to characterize the nonlinear characteristics of heat capacity. is the temperature gradient dissipation correction term, which is used to suppress the violent oscillation of heat flow. γ is the gradient dissipation coefficient, which is used to suppress the violent oscillation of heat flow. is the temperature gradient.

[0113] The U-Net network introduces an attention mechanism in the skip connection to dynamically enhance the weights of the hot spot-related feature channels. For example, the relevant feature channels correspond to areas with temperature gradients greater than 2°C / cm. In addition, by superimposing a temperature regression branch in the decoding stage, the segmentation accuracy and hot spot temperature prediction are jointly optimized.

[0114] When the thermal imaging data is obtained by performing multi-angle thermal imaging of photovoltaic modules by a thermal imager, the two-dimensional detection is expanded to three-dimensional volume analysis in the output layer, three-dimensional feature modeling is performed, and a Monte Carlo Dropout layer is added to output confidence assessment and perform uncertainty quantification.

[0115] It can be understood that the feature fusion module is used to stack texture features, temperature gradient features, hot spot area features and abnormal fluctuation features in the channel dimension, calculate the weights of each first feature through the Squeeze-and-Excitation (SE) module, realize attention weighting, and realize feature fusion through cross-modal feature interaction and information enhancement to obtain fused features.

[0116] In some embodiments, the feature fusion module is embedded with a wind field-dust diffusion coupling model, and the wind field-dust diffusion coupling model includes:

[0117]

[0118] Boundary conditions on the surface of photovoltaic modules:

[0119]

[0120] in, is the convection term, C is the dust concentration field, which is used to characterize the dynamic spatial distribution of dust mass concentration per unit volume of air, t is time, u is the three-dimensional wind speed field, which is used to describe the air flow around the photovoltaic module, and D turb is the turbulent diffusion coefficient, which is related to the wind speed and the surface roughness of the photovoltaic module, k dep is the sedimentation rate constant, which is obtained based on the effects of gravity sedimentation, electrostatic adsorption and inertial collision. resusp is the dust resuspension source term, n is the boundary normal direction perpendicular to the photovoltaic panel surface, k adh is the surface adhesion coefficient of the photovoltaic module, k detach is the re-flying threshold, which is used to characterize the probability of dust detaching from the surface at the critical wind speed, d dust is the effective thickness of the dust layer.

[0121] The wind field-dust diffusion coupling model is used to predict the dust distribution offset in the next 5 minutes, achieve refined mapping of the wind field dynamic characteristics and the dust diffusion process, and provide decision-making support for the operation and maintenance of clean energy facilities with both predictive accuracy and physical interpretability.

[0122] For example, the weight of the temperature gradient feature corresponding to the thermal imaging data is 62%.

[0123] The output module uses the DeepLabv3+ network to generate a binary mask for the fused features, locates the precise boundary of the hot spot, and superimposes a fully connected layer and ReLU as a temperature prediction head on the segmentation mask to predict the maximum temperature value in the hot spot area. The segmentation result is superimposed on the visible light image through bilinear interpolation, and the temperature level is annotated with pseudo color according to the hot spot temperature difference threshold. The abstract detection result is associated with the visible image, and the visible light image carrying the hot spot area and hot spot level is output as the hot spot detection result.

[0124] In some embodiments, the output module includes a hot spot detection head and a dust attribution head. The output module performs region segmentation and temperature regression on the fused features, maps the obtained hot spot regions and hot spot levels to the visible light image, and outputs them as the hot spot detection results, including:

[0125] The hot spot detection head performs pixel-level segmentation on the fusion features to generate a hot spot area marked by a binary mask, and generates a hot spot level of the hot spot area based on the coupling relationship between the local temperature rise and the current density in the hot spot area;

[0126] The dust attribution head quantifies the contribution of dust deposition to the temperature rise of hot spots based on the above-mentioned fusion features, generates an attribution heat map, corrects the hot spot area, maps the hot spot area and hot spot level from the infrared coordinate system to the visible light image, and generates hot spot detection results.

[0127] Among them, the hot spot detection head includes a region segmentation module and a temperature regression module. The region segmentation module is built based on the DeepLabv3+ network. The encoder uses ResNet-34 to extract multi-scale features. The decoder embeds the channel attention mechanism (CBAM) to enhance the hot spot edge segmentation capability, introduces the atrous spatial pyramid pooling (ASPP) to capture hot spot features of different scales, and can also use Dice Loss to optimize the small target segmentation accuracy.

[0128] The temperature regression module is constructed based on a fully convolutional network (FCN) combined with the Swish activation function. According to the coupling relationship between the local temperature rise and the current density in the hot spot area, the hot spot area is classified according to the classification threshold to generate the hot spot level.

[0129] Among them, when the ambient temperature fluctuation is greater than 5°C, the classification threshold is compensated.

[0130] The dust attribution head includes a dust-temperature rise coupling analysis module and a cross-modal feature mapping module. The dust-temperature rise coupling analysis module is a two-branch network. Branch one is used to extract dust deposition characteristics, and branch two is used to model the relationship between dust thickness and heat dissipation efficiency. The two-branch network can quantify the contribution of dust deposition to the temperature rise of hot spots, generate an attribution heat map, and correct the hot spot area based on the attribution heat map.

[0131] The cross-modal feature mapping module is constructed based on deformable convolutional layers (Deformable ConvNets v2) combined with affine transformation matrices. It realizes infrared and visible light image registration based on SIFT feature point matching, eliminates resolution differences through bilinear interpolation, maps hot spot areas and levels from the infrared coordinate system to the visible light image, achieves pixel-level alignment, generates hot spot detection results, and outputs them.

[0132] In some embodiments, as Figure 3 As shown, the dust cleaning prediction model includes a data input layer, a feature extraction layer, a prediction layer, and an output decision layer connected in sequence. The environmental information, the dust information, and the hot spot detection results are input into the dust cleaning prediction model to obtain the cleaning parameters of each hot spot output by the dust cleaning prediction model, including:

[0133] Through the data input layer, the environmental information, the dust information and the hot spot detection results are integrated to generate multi-source heterogeneous data;

[0134] By means of the feature extraction layer, the multi-source heterogeneous data is analyzed for dimensionality reduction and key influencing factors are screened to obtain a plurality of second features;

[0135] Processing the time series dependency of the second feature through the prediction layer to generate a prediction result;

[0136] The predicted values in the prediction results are projected to the feasible domain through the output decision layer to obtain the cleaning parameters of each hot spot and output them.

[0137] Among them, the data input layer is constructed based on the Dynamic Time Twisting (DTT) algorithm and the affine transformation matrix.

[0138] In actual execution, in the data input layer, the environmental information, the dust information and the hot spot detection results are time-aligned through the DTT algorithm, and the coordinates of the hot spot area are mapped to the global coordinate system of the photovoltaic panel through the affine transformation matrix to achieve spatial alignment.

[0139] The data input layer also performs Z-Score standardization on environmental information and binning normalization on dust information. The processed standard data is used to construct a spatiotemporal tensor as multivariate heterogeneous data. The spatiotemporal tensor is defined by the time dimension, space dimension, and feature dimension.

[0140] In the feature extraction layer, the XG-Boost regression model is used to predict cleaning priorities based on hot spot levels. SHAP (Shapley Additive exPlanations) values are calculated to quantify the contribution of each feature. Features with the top 30% SHAP values, such as dust thickness, local wind speed, and hot spot ΔT gradient, are selected to assess their importance. For example, severe hot spot areas are treated with robot scraping and chemical cleaning, while mild hot spot areas are treated with drone atomization spraying.

[0141] Deep feature extraction is performed using a three-layer spatiotemporal convolutional network (ST-ConvNet). The first convolutional layer includes a 3D convolution kernel of 3×3×3 (time×height×width) and 64 output channels. The second convolutional layer introduces a channel attention mechanism (SENet Block) with a compression ratio of 16. The third convolutional layer is equipped with spatiotemporal pooling (MaxPool3d) to obtain a dimensionality-reduced feature vector.

[0142] By introducing a multi-head self-attention mechanism (Multi-Head Self-Attention) with 4 heads, the dependencies between features are calculated, an attention weight matrix is generated, and dynamic weighting of feature vectors is achieved to obtain multiple second features.

[0143] In the prediction layer, a timing modeling network constructed by a bidirectional LSTM (BiLSTM) stacking structure and a temporal convolutional network (TCN) is used to capture the dynamic time series dependency of the second feature. The recurrent structure of the bidirectional LSTM (BiLSTM) is used to capture the bidirectional temporal dependency. Taking into account the cumulative effect and lag effect of dust deposition, a dilated convolution layer with an exponential dilation rate (Dilation = 1, 2, 4) is used in the temporal convolutional network (TCN) to simultaneously capture the short-term fluctuations of hourly wind speed changes and the long-term trend of seasonal dust deposition. The underlying timing features are shared by a multi-task prediction head, and prediction results such as cleaning time, detergent usage, and priority classification are synchronously output.

[0144] In the output decision layer, based on the cleaning parameter generation rules, the predicted values in the prediction results are projected into the feasible domain through a constrained fully connected network to generate and output the cleaning parameters of each hot spot.

[0145] For example, in the cleaning parameter generation rules, if the dust thickness is greater than 0.5mm and the temperature rise ΔT of the hot spot is greater than 15°C, the emergency cleaning mode is triggered forcibly. According to environmental protection constraints, the upper limit of the cleaning agent dosage is set to 50mL / m 2 .

[0146] In addition, the output decision layer is also equipped with an A* algorithm to calculate the optimal moving path for cleaning photovoltaic modules.

[0147] In this embodiment, through multi-source data integration, feature extraction and dimensionality reduction, time series modeling, and feasible domain projection of the results, a more accurate, reliable, and efficient dust cleaning prediction can be achieved, providing a scientific basis for the optimization of cleaning work in practical applications and improving overall work efficiency and cleaning quality.

[0148] In some embodiments, after cleaning the photovoltaic assembly based on the cleaning parameters, the method further includes:

[0149] Calculating a first power generation efficiency of the photovoltaic assembly according to a first IV curve in the operation information;

[0150] Collect the second IV curve and feedback thermal imaging data of the cleaned photovoltaic modules;

[0151] calculating a second power generation efficiency of the photovoltaic assembly according to the second IV curve;

[0152] Cleaning of the photovoltaic assembly is optimized according to the thermal imaging data, the feedback thermal imaging data, and a deviation between the first power generation efficiency and the second power generation efficiency.

[0153] In actual implementation, the average power is calculated according to the first IV curve, and the first power generation efficiency of the component is evaluated by calculating the ratio between the average power and the solar irradiance.

[0154] The second IV curve of the cleaned photovoltaic module is collected, and the average power is calculated. The second power generation efficiency of the module is evaluated by calculating the ratio between the average power and the solar irradiance. The temperature data obtained by thermal imaging technology is used as feedback thermal imaging data to evaluate whether cleaning has an impact on hot spots and the overall temperature distribution.

[0155] Analyze the difference in power generation efficiency between the first power generation efficiency and the second power generation efficiency. If the power generation efficiency is significantly improved after cleaning and the thermal imaging data does not show overheating areas, it means that the cleaning is effective and the power generation efficiency of the photovoltaic module is improved.

[0156] If the second power generation efficiency after cleaning is improved compared to the first power generation efficiency, but the thermal imaging data shows that the component still has a hot spot problem, it means that the cleaning work is insufficient and there may be local dirt or shadow obstructions that have not been completely removed.

[0157] If the thermal imaging data is normal after cleaning, with no overheated areas, but the power generation efficiency has not improved significantly, it means that the hot spots may be caused by other factors such as component damage or electrical problems.

[0158] In areas with hot spots, clean more frequently or use a more delicate cleaning tool.

[0159] In areas where cleaning is ineffective, for example due to shadows or concentrated contaminants, the system automatically optimizes cleaning strategies, adjusts cleaning cycles and cleaning methods to ensure cleaning is performed at the most appropriate time, or schedules local cleaning to improve component life and power generation efficiency.

[0160] If cleaning fails to significantly improve power generation efficiency and hot spots are present, an inspection recommendation is generated, recommending further technical inspections, such as checking the PV module's cells and electrical connections, to determine if there are manufacturing defects or damage. All cleaning records are automatically recorded in a database for later analysis and evaluation of cleaning efficiency, further improving the performance of automated PV module cleaning.

[0161] In this embodiment, by evaluating, analyzing and optimizing the cleaning of photovoltaic modules, analyzing the power generation efficiency through IV curves, observing the temperature distribution through thermal imaging data, and optimizing the cleaning process based on the differences between the two, the power generation efficiency of the photovoltaic modules is improved, and problems such as hot spots are reduced, thereby ensuring the optimal operating state of the photovoltaic modules.

[0162] The photovoltaic module cleaning method based on image processing provided in the embodiments of the present application can be executed by a photovoltaic module cleaning device based on image processing. In the embodiments of the present application, the photovoltaic module cleaning method based on image processing is executed by a photovoltaic module cleaning device based on image processing as an example to illustrate the photovoltaic module cleaning device based on image processing provided in the embodiments of the present application.

[0163] An embodiment of the present application also provides a photovoltaic component cleaning device based on image processing.

[0164] like Figure 4 As shown, the photovoltaic component cleaning device based on image processing includes:

[0165] A first acquisition module 410 is used to acquire operating information, visible light images, near infrared images, and thermal imaging data of photovoltaic modules;

[0166] A first processing module 420 is configured to input the operating information, the visible light image, the thermal imaging data, and the near-infrared image into a hot spot detection model to obtain a hot spot detection result of the photovoltaic module output by the hot spot detection model, wherein the hot spot detection result includes a hot spot area and a hot spot level. The hot spot detection model is obtained based on a plurality of first training samples, wherein the first training samples include sample operating information, sample visible light images, sample thermal imaging data, sample near-infrared images, and hot spot detection labels of sample photovoltaic modules;

[0167] A second acquisition module 430 is configured to acquire environmental information and dust information of dust attached to the photovoltaic module, wherein the dust information includes dust type and attachment form;

[0168] a second processing module 440 configured to input the environmental information, the dust information, and the hot spot detection result into a dust cleaning prediction model to obtain a cleaning parameter for each hot spot output by the dust cleaning prediction model, wherein the dust cleaning prediction model is obtained based on a plurality of second training samples, the second training samples including sample environmental information, sample dust information, hot spot detection labels, and cleaning parameter labels of the sample PV modules;

[0169] The third processing module 450 is configured to clean the photovoltaic assembly based on the cleaning parameters.

[0170] According to the image processing-based photovoltaic module cleaning device provided in the embodiment of the present application, hot spot detection, comprehensive analysis of environmental information, dust information and cleaning parameters realizes efficient, accurate and targeted photovoltaic module cleaning. With the assistance of hot spot detection and dust cleaning prediction models, it can intelligently guide the cleaning process, improve the cleaning efficiency of photovoltaic modules, extend their service life, and reduce unnecessary cleaning work, thereby optimizing the maintenance process of photovoltaic modules, being able to effectively clean different types of dust, and improving the dust cleaning effect of photovoltaic modules.

[0171] The cleaning device of the photovoltaic assembly based on image processing in the embodiment of the present application can be an electronic device, or a component in an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or other devices other than a terminal. Exemplary, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a handheld computer, a vehicle-mounted electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (Augmented Reality, AR) / virtual reality (Virtual Reality, VR) device, a robot, a wearable device, a super mobile personal computer (Ultra-Mobile Personal Computer, UMPC), a netbook or a personal digital assistant (Personal Digital Assistant, PDA), etc., and can also be a server, a network attached storage (Network Attached Storage, NAS) or a personal computer (Personal Computer, PC), etc., and the embodiment of the present application is not specifically limited.

[0172] The photovoltaic module cleaning device based on image processing in the embodiment of the present application can be a device having an operating system. The operating system can be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0173] The photovoltaic component cleaning device based on image processing provided in the embodiment of the present application can implement the various processes implemented in the photovoltaic component cleaning method based on image processing in the above embodiment. To avoid repetition, they are not described here.

[0174] In some embodiments, as Figure 5As shown, an embodiment of the present application also provides an electronic device 500, including a processor 501, a memory 502, and a computer program stored in the memory 502 and executable on the processor 501. When the program is executed by the processor 501, each process of the above-mentioned embodiment of the cleaning method for photovoltaic components based on image processing is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0175] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0176] An embodiment of the present application also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the above-mentioned embodiment of the photovoltaic component cleaning method based on image processing are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0177] The processor is the processor in the electronic device in the above embodiment. The readable storage medium includes a computer-readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0178] An embodiment of the present application further provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned photovoltaic component cleaning method based on image processing.

[0179] The processor is the processor in the electronic device in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0180] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, which are coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-mentioned embodiment of the photovoltaic component cleaning method based on image processing, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0181] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0182] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be noted that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0183] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the cleaning method of photovoltaic components based on image processing in each embodiment of the present application.

[0184] In the description of this application, "first feature" and "second feature" may include one or more such features.

[0185] In the description of this application, “plurality” means two or more.

[0186] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.

[0187] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0188] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and intent of the present application, and that the scope of the present application is defined by the claims and their equivalents.

Claims

1. A method for cleaning a photovoltaic module based on image processing, characterized in that: include: Obtaining operational information, visible light images, near-infrared images, and thermal imaging data of photovoltaic modules; Inputting the operating information, the visible light image, the thermal imaging data, and the near-infrared image into a hot spot detection model, obtaining a hot spot detection result of the photovoltaic module output by the hot spot detection model, the hot spot detection result including a hot spot area and a hot spot level, wherein the hot spot detection model is obtained based on a plurality of first training samples, wherein the first training samples include sample operating information, sample visible light images, sample thermal imaging data, sample near-infrared images, and hot spot detection labels of sample photovoltaic modules; Acquiring environmental information and dust information of dust attached to the photovoltaic module, wherein the dust information includes dust type and attachment form; Inputting the environmental information, the dust information, and the hot spot detection result into a dust cleaning prediction model to obtain a cleaning parameter for each hot spot output by the dust cleaning prediction model, wherein the dust cleaning prediction model is obtained based on a plurality of second training samples, the second training samples including sample environmental information, sample dust information, hot spot detection labels, and cleaning parameter labels of the sample photovoltaic modules; The photovoltaic component is cleaned based on the cleaning parameters.

2. The method for cleaning a photovoltaic module based on image processing according to claim 1, characterized in that: The hot spot detection model includes a feature extraction module, a feature fusion module, and an output module connected in sequence. The operation information, the visible light image, the thermal imaging data, and the near-infrared image are input into the hot spot detection model to obtain the hot spot detection result of the photovoltaic module output by the hot spot detection model, including: By means of the feature extraction module, feature extraction is performed on the operation information, the visible light image, the thermal imaging data, and the near-infrared image to obtain corresponding multiple first features; By means of the feature fusion module, the plurality of first features are subjected to spatiotemporal alignment and data fusion to obtain a fused feature; The output module performs region segmentation and temperature regression on the fused features, and the obtained hot spot regions and hot spot levels are mapped to the visible light image and output as the hot spot detection results.

3. The method for cleaning a photovoltaic module based on image processing according to claim 2, characterized in that: The feature extraction module includes a ResNet-50 network, a 3D convolutional network, a U-Net network, and an LSTM network. The multiple first features include texture features, temperature gradient features, hot spot area features, and abnormal fluctuation features of the dust. The feature extraction module performs feature extraction on the operation information, the visible light image, the thermal imaging data, and the near-infrared image, respectively, to obtain corresponding multiple first features, including: Extracting the texture features of the dust from the visible light image using the ResNet-50 network; The temperature gradient features extracted from the thermal imaging data by the 3D convolutional network; Segmenting the hot spot area in the near-infrared image according to reflectivity using the U-Net network to obtain features of the hot spot area; The abnormal fluctuation characteristics of the photovoltaic components are extracted from the operation information through the LSTM network.

4. The method for cleaning a photovoltaic module based on image processing according to claim 3, characterized in that: The 3D convolutional network is embedded with a dust occlusion heat conduction model, which includes: Where ΔT is the temperature rise of the PV module; is the photoelectric-thermal coupling term, J shunt is the bypass current density in the hot spot area, which is used to characterize the local current concentration effect caused by hidden cracks or contamination. cell (T) is the temperature-dependent resistance of the cell, which is used to reflect the negative temperature coefficient characteristics of the semiconductor material of the photovoltaic module, d dust is the effective thickness of the dust layer, The humidity-temperature dependent equivalent thermal conductivity is calculated. The heat conduction path is corrected by connecting the glass and dust heat conduction models in series. The humidity effect is dominated by the capillary condensation effect. is the dust porosity, which indicates the volume ratio of air or other pores in the dust layer, T is the local temperature, α h RH is the humidity correction factor, which is used to quantify the enhancement of heat conduction by the water film formed after the dust pores absorb moisture. h is the capillary condensation enhancement coefficient, which is used to ensure that the interface effect of nano-scale dust particles amplifies the humidity effect, and RH is the relative humidity; is the environmental dissipation term, P loss is the power loss per unit area, which is caused by the decrease in photoelectric conversion efficiency due to local shading. β(T) is the hot spot self-shielding factor, which is used to characterize the negative feedback suppression of the heat generation power caused by the intensified carrier recombination in the high temperature area. conv +h red is the comprehensive heat dissipation coefficient, h conv is the wind speed dependent convective heat transfer term, h red is the radiation heat transfer term; is the time dynamic term, τ(T) is the temperature-dependent relaxation time, which is used to characterize the nonlinear characteristics of heat capacity. is the temperature gradient dissipation correction term, which is used to suppress the violent oscillation of heat flow. γ is the gradient dissipation coefficient, which is used to suppress the violent oscillation of heat flow. is the temperature gradient.

5. The method for cleaning a photovoltaic module based on image processing according to claim 2, characterized in that: The feature fusion module is embedded with a wind field-dust diffusion coupling model, which includes: Boundary conditions on the surface of photovoltaic modules: in, is the convection term, C is the dust concentration field, which is used to characterize the dynamic spatial distribution of dust mass concentration per unit volume of air, t is time, u is the three-dimensional wind speed field, which is used to describe the air flow around the photovoltaic module, and D turb is the turbulent diffusion coefficient, which is related to the wind speed and the surface roughness of the photovoltaic module, k dep is the sedimentation rate constant, which is obtained based on the effects of gravity sedimentation, electrostatic adsorption and inertial collision. resusp is the dust resuspension source term, n is the boundary normal direction perpendicular to the photovoltaic panel surface, k adh is the surface adhesion coefficient of the photovoltaic module, k detach is the re-flying threshold, which is used to characterize the probability of dust detaching from the surface at the critical wind speed, d dust is the effective thickness of the dust layer.

6. The method for cleaning a photovoltaic module based on image processing according to claim 2, characterized in that: The output module includes a hot spot detection head and a dust attribution head. The output module performs region segmentation and temperature regression on the fused features, maps the obtained hot spot regions and hot spot levels to the visible light image, and outputs them as the hot spot detection results, including: The hot spot detection head performs pixel-level segmentation on the fusion features to generate a hot spot area marked by a binary mask, and generates a hot spot level of the hot spot area based on the coupling relationship between the local temperature rise and the current density in the hot spot area; The dust attribution head quantifies the contribution of dust deposition to the temperature rise of hot spots based on the above-mentioned fusion features, generates an attribution heat map, corrects the hot spot area, maps the hot spot area and hot spot level from the infrared coordinate system to the visible light image, and generates hot spot detection results.

7. The method for cleaning a photovoltaic module based on image processing according to claim 1, characterized in that: The dust cleaning prediction model includes a data input layer, a feature extraction layer, a prediction layer, and an output decision layer connected in sequence. The environmental information, the dust information, and the hot spot detection results are input into the dust cleaning prediction model to obtain the cleaning parameters of each hot spot output by the dust cleaning prediction model, including: Through the data input layer, the environmental information, the dust information and the hot spot detection results are integrated to generate multi-source heterogeneous data; By means of the feature extraction layer, the multi-source heterogeneous data is analyzed for dimensionality reduction and key influencing factors are screened to obtain a plurality of second features; Processing the time series dependency of the second feature through the prediction layer to generate a prediction result; The prediction results are mapped from the feature space to the parameter space through the output decision layer, and the cleaning parameters of each hot spot are generated and output.

8. The method for cleaning a photovoltaic module based on image processing according to claim 1, characterized in that: After cleaning the photovoltaic assembly based on the cleaning parameters, the method further includes: Calculating a first power generation efficiency of the photovoltaic assembly according to a first IV curve in the operation information; Collect the second IV curve and feedback thermal imaging data of the cleaned photovoltaic modules; calculating a second power generation efficiency of the photovoltaic assembly according to the second IV curve; Cleaning of the photovoltaic assembly is optimized according to the thermal imaging data, the feedback thermal imaging data, and a deviation between the first power generation efficiency and the second power generation efficiency.

9. A photovoltaic module cleaning device based on image processing, characterized in that: include: The first acquisition module is used to acquire operating information, visible light images, near-infrared images and thermal imaging data of photovoltaic modules; a first processing module, configured to input the operating information, the visible light image, the thermal imaging data, and the near-infrared image into a hot spot detection model, and obtain a hot spot detection result of the photovoltaic module output by the hot spot detection model, wherein the hot spot detection result includes a hot spot area and a hot spot level, and the hot spot detection model is obtained based on a plurality of first training samples, wherein the first training samples include sample operating information, sample visible light images, sample thermal imaging data, sample near-infrared images, and hot spot detection labels of sample photovoltaic modules; A second acquisition module is configured to acquire environmental information and dust information of dust attached to the photovoltaic module, wherein the dust information includes dust type and attachment form; a second processing module, configured to input the environmental information, the dust information, and the hot spot detection result into a dust cleaning prediction model to obtain a cleaning parameter for each hot spot output by the dust cleaning prediction model, wherein the dust cleaning prediction model is obtained based on a plurality of second training samples, the second training samples including sample environmental information, sample dust information, hot spot detection labels, and cleaning parameter labels of the sample photovoltaic modules; The third processing module is configured to clean the photovoltaic assembly based on the cleaning parameters.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for cleaning a photovoltaic component based on image processing according to any one of claims 1 to 8 is implemented.

Images