A photovoltaic panel dust intelligent prevention and treatment method, device, equipment and storage medium

Abstract

The present application relates to the photovoltaic power station operation and maintenance technical field, especially point to a kind of photovoltaic panel dust intelligent prevention and control method, device, equipment and computer storage medium.The present application is based on computational fluid dynamics and discrete element method constructs dust deposition force life cycle model, quantifies the interaction of environment, dust and photovoltaic module;Through environmental sensor group and convolutional neural network identification dust image, output real-time dust data;Combined with multidimensional performance analysis model predicts power generation efficiency loss;Based on the prediction result through deep residual neural network dynamic optimization cleaning cycle and cleaning mode, generate cleaning decision and control intelligent cleaning equipment to execute cleaning operation.The present application realizes the accurate monitoring of dust and the high-precision prediction of efficiency loss, can dynamically optimize cleaning strategy, avoid cleaning insufficient or overcleaning, improve cleaning efficiency and reduce operation and maintenance cost.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power plant operation and maintenance technology, and in particular to a method, device, equipment and computer storage medium for intelligent prevention and control of dust accumulation on photovoltaic panels. Background Technology

[0002] Dust accumulation on photovoltaic (PV) panels is one of the core issues restricting the power generation efficiency and economic benefits of PV power plants. Current technologies primarily rely on traditional methods such as manual inspections, fixed-period cleaning, and natural rainfall to establish a basic dust control system. However, existing technologies lack quantitative analysis of dust deposition mechanisms and have failed to form a complete model coupling multiple factors between environmental parameters, dust characteristics, and PV modules. This results in the inability to accurately quantify the weight of dust adsorption forces, the influence of dust particle size, and the critical conditions for rainwater cleaning. Furthermore, existing solutions lack a real-time monitoring and power generation efficiency loss prediction system based on a dust accumulation mechanical lifecycle model. They typically rely directly on manual cleaning or simply on natural rainfall, leading to a lack of flexibility in selecting cleaning cycles and methods. Specific shortcomings include: insufficient rainfall not only fails to clean but can also create an adhesive layer, increasing dust density; manual cleaning is inefficient, costly, and cannot achieve precise operation and maintenance; fixed-period cleaning can easily lead to insufficient or excessive cleaning, severely impacting the power generation efficiency of PV power plants and increasing operation and maintenance costs.

[0003] Therefore, there is an urgent need for an intelligent prevention and control technology based on multi-factor coupled quantitative analysis. Summary of the Invention

[0004] Therefore, the technical problem to be solved by the present invention is to overcome the problems in the prior art caused by the lack of quantitative analysis and real-time monitoring and prediction system for dust deposition mechanism, which leads to blind selection of cleaning cycle and method, insufficient or excessive cleaning, inefficient use of natural precipitation, and thus affects power generation efficiency and operation and maintenance costs.

[0005] To address the aforementioned technical problems, this invention provides an intelligent method for preventing dust accumulation on photovoltaic panels, comprising:

[0006] Based on multi-factor coupling analysis combining computational fluid dynamics and discrete element method, a dust deposition mechanical life cycle model is constructed to quantify the interaction between environmental parameters, dust characteristics and photovoltaic modules, and to determine the weight of dust adsorption force, the influence law of dust particle size and the critical condition for precipitation cleaning. The system collects environmental parameters and photovoltaic panel performance parameters in real time using an environmental sensor array and a photovoltaic panel status acquisition device. It also uses a convolutional neural network to identify dust accumulation images on the photovoltaic panel surface and outputs real-time dust accumulation data. Based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanical life cycle model, combined with the multi-dimensional performance analysis model, the power generation efficiency loss of photovoltaic panels due to dust accumulation is predicted. The multi-dimensional performance analysis model includes at least the influence path of dust accumulation on the transmittance, reflectance and temperature of photovoltaic panels. Based on the predicted power generation efficiency loss, the cleaning cycle and cleaning method are dynamically optimized through a deep residual neural network to generate cleaning decisions and control the intelligent cleaning equipment to perform cleaning operations.

[0007] Preferably, the multi-factor coupled analysis based on computational fluid dynamics and the discrete element method, constructing a dust deposition mechanical life cycle model, quantifying the interaction between environmental parameters, dust characteristics, and photovoltaic modules, and determining the weight of dust adsorption force, the influence law of dust particle size, and the critical conditions for precipitation cleaning include: The van der Waals forces under dry conditions and the capillary forces under humid conditions are quantified as the forces for dust adsorption, and the weight of the capillary forces is determined according to the ambient humidity and incorporated into the prediction model. Clearly define the differences in migration and sedimentation between small-diameter and large-diameter particles, and set differentiated monitoring and cleaning thresholds for particles of different diameters; Precipitation intensity and dust density are determined as the joint criteria for determining the critical condition for precipitation cleaning. When the precipitation intensity is greater than the first threshold and the dust density is less than the second threshold, the precipitation is determined to have a cleaning effect. This critical condition is used as an auxiliary criterion for natural cleaning.

[0008] Preferably, the step of collecting environmental parameters and photovoltaic panel performance parameters in real time through an environmental sensor group and a photovoltaic panel status acquisition device, and using a convolutional neural network to identify dust accumulation images on the photovoltaic panel surface and output real-time dust accumulation data includes: The system uses an environmental sensor array to collect real-time data on wind speed, humidity, precipitation intensity, atmospheric particulate matter concentration, and industrial emission indicators around the photovoltaic power station. The high-definition imaging equipment of the photovoltaic panel status acquisition device acquires dust images on the surface of the photovoltaic panel, and simultaneously acquires the output power, component temperature, light transmittance and reflectance of the photovoltaic panel; The dust accumulation image is input into the convolutional neural network SolNet, and the dust density, dust particle size distribution and dust shape on the photovoltaic panel surface are identified by the image recognition algorithm, and the real-time dust accumulation density is output.

[0009] Preferably, the step of predicting the power generation efficiency loss of photovoltaic panels due to dust accumulation, based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanical life cycle model, combined with a multi-dimensional performance analysis model, includes: Substituting the real-time dust accumulation density into the semi-empirical efficiency prediction formula constructed based on the dust deposition mechanical life cycle model, the formula characterizes the exponential decay relationship between the actual power generation efficiency after dust accumulation and the rated power generation efficiency under clean conditions. The decay relationship is jointly determined by the correction coefficient, real-time dust accumulation density, dust particle size distribution, number of dust particles, dust accumulation duration, effective light-receiving area of ​​photovoltaic panel, and dust particle deposition rate. By considering the triple impact of dust accumulation on photovoltaic panel transmittance, reflectivity, and module temperature, the formula prediction results are corrected, and the predicted value of power generation efficiency loss is output.

[0010] Preferably, the step of dynamically optimizing the cleaning cycle and cleaning method based on the prediction results of the power generation efficiency loss, generating cleaning decisions, and controlling the intelligent cleaning equipment to perform cleaning operations includes: Based on the quantitative relationship between dust density and power generation efficiency, multiple efficiency loss thresholds are set. When the predicted efficiency loss reaches the corresponding threshold, a cleaning command is triggered. The optimal cleaning interval is dynamically calculated by combining the dust deposition rate and environmental changes to optimize the cleaning cycle. The cleaning method is selected based on dust density, dust particle size, photovoltaic panel installation tilt angle and power station site environment. When the dust density is below the first threshold, low-pressure spray cleaning by drone is used. When the dust density is above the second threshold or an adhesive layer is formed, physical wiping and spray cleaning by automatic cleaning robot is used. When rainfall meets the critical cleaning conditions, manual or intelligent cleaning is suspended and natural precipitation is used to assist in cleaning. Cleaning decisions are generated and intelligent cleaning equipment is controlled to perform cleaning operations.

[0011] Preferably, after the intelligent cleaning device performs the cleaning operation, it further includes: After cleaning, the dust accumulation status and performance parameters of the photovoltaic panels are collected again by the environmental sensor group and the photovoltaic panel status acquisition device to verify the efficiency recovery after cleaning. If the expected effect is not achieved, the cleaning plan is optimized and a second cleaning is triggered.

[0012] Preferably, it also includes: incorporating cleaning effect data into a background database, continuously optimizing the parameters of the dust deposition mechanical life cycle model and the algorithm of the deep residual neural network, so as to realize the self-learning and self-optimization of the prevention and control system.

[0013] This invention provides an intelligent dust control device for photovoltaic panels, comprising: The dust accumulation mechanism quantification module is used to construct a dust deposition mechanical life cycle model based on multi-factor coupling analysis combining computational fluid dynamics and discrete element method, quantify the interaction between environmental parameters, dust characteristics and photovoltaic modules, and determine the weight of dust adsorption force, the influence law of dust particle size and the critical condition for precipitation cleaning. The real-time monitoring module is used to collect environmental parameters and photovoltaic panel performance parameters in real time through an environmental sensor group and a photovoltaic panel status acquisition device, and to use a convolutional neural network to identify dust accumulation images on the surface of the photovoltaic panel and output real-time dust accumulation data. The efficiency loss prediction module is used to predict the power generation efficiency loss of photovoltaic panels due to dust accumulation based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanical life cycle model, combined with a multi-dimensional performance analysis model. The multi-dimensional performance analysis model includes at least the influence path of dust accumulation on the transmittance, reflectance and temperature of the photovoltaic panel. A cleaning scheme optimization module is used to dynamically optimize the cleaning cycle and cleaning method based on the predicted power generation efficiency loss using a deep residual neural network, generating a cleaning decision. An intelligent cleaning execution module is used to control intelligent cleaning equipment to perform cleaning operations according to the cleaning decision. This invention also provides an intelligent dust control device for photovoltaic panels, comprising: Memory, used to store computer programs; A processor is used to implement the steps of the above-described intelligent dust prevention and control method for photovoltaic panels when executing the computer program.

[0014] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described intelligent dust prevention and control method for photovoltaic panels.

[0015] The technical solution of the present invention has the following advantages compared with the prior art: The intelligent dust control method for photovoltaic panels described in this invention quantifies the interaction between the environment, dust, and photovoltaic modules by constructing a dust deposition mechanical life cycle model. It combines convolutional neural networks for accurate identification of dust accumulation images and multi-dimensional performance analysis models for predicting power generation efficiency loss. Furthermore, it utilizes deep residual neural networks to dynamically optimize the cleaning cycle and cleaning method. This overcomes the shortcomings of blind cleaning decisions in existing technologies, achieving accurate real-time monitoring of dust accumulation status and high-precision prediction of power generation efficiency loss. It can adaptively generate the optimal cleaning strategy based on the actual dust accumulation situation and environmental changes, thereby ensuring the power generation efficiency of photovoltaic panels while effectively avoiding insufficient or excessive cleaning, and significantly reducing the operation and maintenance costs of photovoltaic power plants. Attached Figure Description

[0016] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart illustrating the implementation of an intelligent dust control method for photovoltaic panels provided by the present invention. Figure 2This is a structural block diagram of an intelligent dust control device for photovoltaic panels provided in an embodiment of the present invention. Detailed Implementation

[0017] The core of this invention is to provide a method, device, equipment, and computer storage medium for intelligent prevention and control of dust accumulation on photovoltaic panels, which can effectively quantify the dust accumulation mechanism and realize dynamic adaptive optimization of cleaning strategies, so as to overcome the problems of blind cleaning decisions and unpredictable efficiency losses in the prior art.

[0018] To enable those skilled in the art to better understand the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please refer to Figure 1. Figure 1 The flowchart illustrates the implementation of an intelligent dust control method for photovoltaic panels provided by this invention; the specific operation steps are as follows: S101: Based on multi-factor coupling analysis combining computational fluid dynamics and discrete element method, a dust deposition mechanical life cycle model is constructed to quantify the interaction between environmental parameters, dust characteristics and photovoltaic modules, and to determine the weight of dust adsorption force, the influence law of dust particle size and the critical condition for precipitation cleaning. This method first constructs a mechanophysiological model based on the coupled analysis of computational fluid dynamics and discrete element method to describe the entire life cycle of dust deposition on photovoltaic (PV) panel surfaces. This model aims to quantify the multi-factor interactions between environmental parameters, dust characteristics, and PV modules, thereby revealing the dynamic cycle of dust particle deposition, adhesion, rebound, removal, and resuspension on PV panel surfaces. Through this model, the key forces influencing the dust accumulation process and their weights can be identified. For example, the dominant position and contribution ratio of adsorption forces such as van der Waals forces and capillary forces under different humidity conditions are clearly defined. Simultaneously, the model is used to analyze the influence of dust particle size on deposition behavior, distinguishing the migration and settling characteristics of particles of different sizes, and setting differentiated monitoring and cleaning thresholds accordingly. Furthermore, the model further quantifies the cleaning effect of precipitation on dust accumulation. By analyzing the coupling relationship between precipitation intensity and dust density, the critical conditions for natural precipitation to effectively clean PV panels are determined; that is, precipitation has a cleaning effect only when precipitation intensity and dust density satisfy a specific relationship, otherwise it may aggravate dust accumulation. As one implementation method, the model can be specifically used to quantify the adsorption weights of van der Waals forces in dry environments and capillary forces in humid environments, clarify the difference in migration distance between particles with a diameter of less than 10 micrometers and particles with a diameter of more than 20 micrometers, and determine the critical thresholds for precipitation cleaning as a precipitation intensity of 38 mm / h and a dust accumulation density of 40 mg / m².

[0020] By constructing the aforementioned dust deposition mechanical lifecycle model, this step can provide precise physical mechanism support and key parameter basis for subsequent real-time monitoring, efficiency loss prediction, and cleaning scheme optimization, thereby significantly improving the scientificity and accuracy of the entire prevention and control method and avoiding blind cleaning decisions caused by a lack of quantitative understanding of dust accumulation mechanisms.

[0021] S102: Real-time acquisition of environmental parameters and photovoltaic panel performance parameters through an environmental sensor group and a photovoltaic panel status acquisition device, and recognition of dust accumulation images on the photovoltaic panel surface using a convolutional neural network to output real-time dust accumulation data; In the aforementioned intelligent dust control method for photovoltaic panels, real-time dust accumulation data characterizing the dust accumulation status is obtained by collecting environmental parameters from the photovoltaic power station site and the performance parameters of the photovoltaic panels themselves, combined with intelligent recognition of dust accumulation images on the photovoltaic panel surface. Specifically, this method utilizes an environmental sensor array to monitor and collect environmental parameters around the photovoltaic power station in real time. These parameters include, but are not limited to, key environmental factors affecting the dust accumulation process such as wind speed, humidity, precipitation intensity, and atmospheric particulate matter concentration. Simultaneously, a photovoltaic panel status acquisition device collects real-time performance parameters of the photovoltaic panels, such as output power, module temperature, light transmittance, and reflectivity, which reflect the working status of the photovoltaic panels and the degree of influence from dust accumulation. Based on this, the method further employs a convolutional neural network to process and recognize the collected dust accumulation images on the photovoltaic panel surface. This convolutional neural network, after training, can extract and identify key feature indicators such as dust density, dust particle size distribution, and dust shape from the images, thereby outputting accurate real-time dust accumulation data. In one implementation, the environmental sensor group may include various types of sensors to collect different environmental parameters, the photovoltaic panel status acquisition device may include a high-definition imaging device to acquire dust accumulation images, and the convolutional neural network may adopt the SolNet network structure, which can output data such as dust accumulation density with high recognition accuracy, providing an accurate and real-time data basis for subsequent efficiency loss prediction and cleaning decisions.

[0022] This step achieves high-precision, real-time monitoring of dust accumulation on photovoltaic panels by collecting environmental parameters and photovoltaic panel performance parameters in real time, combined with intelligent image recognition of dust accumulation based on convolutional neural networks. This provides reliable data support for subsequent prediction of power generation efficiency loss and optimization of cleaning solutions, thereby significantly improving the accuracy and timeliness of dust accumulation prevention and control.

[0023] S103: Based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanical life cycle model, combined with the multi-dimensional performance analysis model, predict the power generation efficiency loss of the photovoltaic panel due to dust accumulation, wherein the multi-dimensional performance analysis model includes at least the influence path of dust accumulation on the photovoltaic panel transmittance, reflectance and component temperature. Based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanics lifecycle model, combined with a multi-dimensional performance analysis model, the power generation efficiency loss of photovoltaic panels due to dust accumulation is predicted. Specifically, this step aims to construct a comprehensive efficiency loss prediction mechanism based on the acquired real-time dust accumulation status data and referring to the key critical conditions related to dynamic processes such as dust deposition, adhesion, and removal, determined in advance through mechanistic analysis. The core of this mechanism is that it does not rely solely on a single dust density index, but rather couples real-time dust accumulation data with critical conditions reflecting the interaction between dust, environment, and photovoltaic modules (e.g., the environmental humidity threshold affecting dust adsorption and desorption, the critical point of the cleaning or aggravating effect of precipitation on dust accumulation, etc.) to determine the actual impact of the current dust accumulation state on the performance of the photovoltaic panels. On this basis, a multi-dimensional performance analysis model is further introduced. This model comprehensively evaluates the attenuation effect of dust accumulation on the power generation performance of photovoltaic panels from multiple physical dimensions, such as the shading effect of the dust layer on the light transmittance of the photovoltaic panel, the change in surface reflectivity, and the increase in module temperature due to local thermal effects. By combining these multi-dimensional performance degradation factors with a critical condition-based dust accumulation state assessment, a more accurate and comprehensive efficiency loss prediction model can be constructed, thereby outputting the predicted power generation efficiency loss of photovoltaic panels due to dust accumulation in the current and future periods. As one implementation method, real-time dust density and other data can be substituted into a semi-empirical efficiency prediction formula built based on a dust deposition mechanical lifecycle model, and corrected by combining real-time monitoring data of transmittance, reflectance, and module temperature changes to improve prediction accuracy.

[0024] This technical step combines real-time monitoring data with critical conditions based on mechanism analysis and multi-dimensional performance models to achieve accurate and dynamic prediction of power generation efficiency loss caused by dust accumulation on photovoltaic panels. It avoids the prediction bias caused by relying on a single parameter or empirical estimation in traditional methods, and provides a scientific and reliable data foundation for subsequent optimization decisions on cleaning solutions, thereby significantly improving the intelligence level and economic benefits of photovoltaic power plant operation and maintenance.

[0025] S104: Based on the predicted results of the power generation efficiency loss, the cleaning cycle and cleaning method are dynamically optimized through a deep residual neural network to generate cleaning decisions and control the intelligent cleaning equipment to perform cleaning operations.

[0026] Based on the predicted power generation efficiency loss, a cleaning decision is generated and the intelligent cleaning equipment is controlled to perform cleaning operations. Specifically, this is achieved through a decision optimization mechanism integrating a deep residual neural network. This mechanism uses the predicted power generation efficiency loss as the core input, combined with the current dust accumulation parameters of the photovoltaic panels and the dynamic trends of environmental parameters, to collaboratively optimize the cleaning cycle and cleaning method. Regarding the optimization of the cleaning cycle, the system triggers a corresponding cleaning command when the predicted efficiency loss reaches a certain threshold, based on preset multi-level efficiency loss thresholds. Simultaneously, the system dynamically calculates and adjusts the optimal time interval between two cleaning operations based on the dust deposition rate and changes in environmental factors, thereby achieving precise control over the timing of cleaning. Regarding the optimization of the cleaning method, the system automatically selects the optimal cleaning method from multiple preset cleaning modes based on dust density, dust particle size distribution, the installation tilt angle of the photovoltaic panels, and the environmental conditions at the power station site. For example, for different levels of dust accumulation and operating conditions, different methods such as low-pressure spraying, physical wiping, or a combination of cleaning methods can be selected. Finally, the system sends the optimized cleaning decision instructions to the intelligent cleaning equipment, controlling it to perform cleaning operations according to preset paths and parameters. As one implementation method, this deep residual neural network can be trained based on historical cleaning data and efficiency recovery data to continuously improve the accuracy of decisions; while the intelligent cleaning equipment can include automated cleaning robots and drone cleaning equipment, each suitable for cleaning tasks in different scenarios.

[0027] This technology introduces a deep residual neural network for dynamic decision-making, enabling intelligent and adaptive optimization of the cleaning cycle and cleaning method. This avoids the problems of insufficient or excessive cleaning caused by traditional fixed-cycle cleaning methods, thereby significantly reducing cleaning operation and maintenance costs and improving the overall economic benefits of photovoltaic power plants while ensuring the power generation efficiency of photovoltaic panels.

[0028] Based on the above embodiments, in some embodiments, the construction of a dust deposition mechanical lifecycle model based on multi-factor coupling analysis combining computational fluid dynamics and discrete element method, quantifying the interaction between environmental parameters, dust characteristics, and photovoltaic modules, and determining the weight of dust adsorption force, the influence law of dust particle size, and the critical conditions for precipitation cleaning includes: The van der Waals forces under dry conditions and the capillary forces under humid conditions are quantified as the forces for dust adsorption, and the weight of the capillary forces is determined according to the ambient humidity and incorporated into the prediction model. Clearly define the differences in migration and sedimentation between small-diameter and large-diameter particles, and set differentiated monitoring and cleaning thresholds for particles of different diameters; Precipitation intensity and dust density are determined as the joint criteria for determining the critical condition for precipitation cleaning. When the precipitation intensity is greater than the first threshold and the dust density is less than the second threshold, the precipitation is determined to have a cleaning effect. This critical condition is used as an auxiliary criterion for natural cleaning.

[0029] In this embodiment, step S101 constructs a dust deposition mechanical lifecycle model based on multi-factor coupling analysis combining computational fluid dynamics and discrete element method, and quantifies the interaction between environmental parameters, dust characteristics, and photovoltaic modules. This is achieved through the following sub-steps: First, the dust adsorption force is quantified. This process uses environmental humidity data as input. By analyzing the action mechanism of van der Waals forces in dry environments and capillary forces in humid environments, the dominant adsorption force type under different humidity conditions is determined. Specifically, when the environmental humidity is greater than 60%, capillary force becomes the main adsorption force, accounting for 98%. This force weight is output as a key parameter and incorporated into the subsequent prediction model to correct the adsorption effect in efficiency loss prediction. Second, the influence of dust particle size is clarified. This step uses dust particle size distribution data as input. By analyzing the migration and settling characteristics of particles of different sizes, it is determined that particles smaller than 10 μm can migrate thousands of kilometers, while particles larger than 20 μm mainly settle within a 40 km range. Based on this principle, differentiated monitoring and cleaning thresholds are set for particles of different sizes. For example, a more sensitive dust density monitoring frequency is set for fine particles, while a higher cleaning trigger threshold is set for coarse particles, thus outputting a particle size-based monitoring and cleaning strategy. Finally, the critical conditions for precipitation cleaning are determined. This step uses real-time precipitation intensity and dust density data as input. By comparing the precipitation intensity and dust density with preset critical thresholds, a precipitation intensity of 38 mm / h and a dust density of 40 mg / m² are determined as the critical conditions for precipitation cleaning. When the precipitation intensity is greater than 38 mm / h and the dust density is less than 40 mg / m², the precipitation is considered to have a cleaning effect, and a natural cleaning auxiliary judgment signal is output. Conversely, when the precipitation intensity is less than or equal to 38 mm / h, or the dust density is greater than or equal to 40 mg / m², the precipitation is considered to have no cleaning effect and may even aggravate dust accumulation, and an early warning signal is output to trigger manual or intelligent cleaning intervention. The outputs of the above three sub-steps—adsorption force weight, particle size classification threshold, and precipitation cleaning judgment signal—together constitute the core parameters of the dust deposition mechanical life cycle model, providing a quantitative basis for subsequent efficiency loss prediction and cleaning scheme optimization.

[0030] This specific implementation method achieves a precise analysis of the dust accumulation mechanism of photovoltaic panels by quantifying the dust adsorption force, particle size influence law, and precipitation cleaning critical conditions. It provides scientific and quantifiable basic parameters for subsequent monitoring, prediction, and cleaning decisions, significantly improving the pertinence and effectiveness of prevention and control strategies, and avoiding the blind cleaning caused by the ambiguity of the mechanism in traditional methods.

[0031] Based on the above embodiments, in some embodiments, the step of real-time acquisition of environmental parameters and photovoltaic panel performance parameters through an environmental sensor group and a photovoltaic panel status acquisition device, and the use of a convolutional neural network to identify dust accumulation images on the photovoltaic panel surface and output real-time dust accumulation data includes: The system uses an environmental sensor array to collect real-time data on wind speed, humidity, precipitation intensity, atmospheric particulate matter concentration, and industrial emission indicators around the photovoltaic power station. The high-definition imaging equipment of the photovoltaic panel status acquisition device acquires dust images on the surface of the photovoltaic panel, and simultaneously acquires the output power, component temperature, light transmittance and reflectance of the photovoltaic panel; The dust accumulation image is input into the convolutional neural network SolNet, and the dust density, dust particle size distribution and dust shape on the photovoltaic panel surface are identified by the image recognition algorithm, and the real-time dust accumulation density is output.

[0032] In this embodiment, step S102 specifically achieves real-time acquisition of environmental parameters and photovoltaic panel performance parameters through the coordinated operation of an environmental sensor group and a photovoltaic panel status acquisition device. A convolutional neural network is then used to perform high-precision recognition of the dust accumulation images, outputting real-time dust accumulation data. Specifically, the environmental sensor group is deployed within the photovoltaic power station to collect environmental parameters such as wind speed, humidity, precipitation intensity, atmospheric particulate matter concentration, and industrial emission indicators around the photovoltaic power station in real time. These sensors convert the collected raw analog signals into digital signals and transmit them to the back-end processing system via wired or wireless communication, serving as environmental input parameters for subsequent efficiency loss prediction and cleaning scheme optimization. Simultaneously, a high-definition imaging device in the photovoltaic panel status acquisition device, such as a high-resolution industrial camera or infrared thermal imager, images the photovoltaic panel surface at a preset acquisition frequency, acquiring high-resolution dust accumulation images. This high-definition imaging device also acquires performance parameters of the photovoltaic panel, such as output power, component temperature, transmittance, and reflectivity. These parameters reflect the actual working performance of the photovoltaic panel under the current dust accumulation state. The acquired dust accumulation images and performance parameter data are also transmitted to the back-end processing system. In the background processing system, the dust accumulation image is input into a pre-trained convolutional neural network, SolNet. This SolNet network extracts features from the image through its deep convolutional structure, using image recognition algorithms to accurately identify key indicators such as dust density, dust particle size distribution, and dust shape on the photovoltaic panel surface. As an example, the SolNet network outputs a multi-dimensional feature vector, which includes the dust density ρ (in g / m²) and the proportion of dust in each particle size range. In addition to dust shape classification labels. Finally, the back-end processing system integrates the real-time dust density ρ output by the SolNet network, along with the environmental parameters collected by the environmental sensor group and the performance parameters collected by the photovoltaic panel status acquisition device, into structured real-time dust data, which is then used by the efficiency loss prediction module in the subsequent step S3.

[0033] This specific implementation integrates an environmental sensor array, high-definition imaging equipment, and the SolNet convolutional neural network to achieve multi-dimensional, high-precision real-time monitoring of dust accumulation on photovoltaic panels. Its advantages lie in acquiring not only accurate dust density data but also simultaneously obtaining environmental parameters affecting dust accumulation behavior and photovoltaic panel parameters reflecting power generation performance. This provides a comprehensive and accurate data foundation for subsequent efficiency loss prediction based on semi-empirical formulas, thereby significantly improving the prediction accuracy and decision reliability of the entire prevention and control system.

[0034] Based on the above embodiments, in some embodiments, the step of predicting the power generation efficiency loss of photovoltaic panels due to dust accumulation, based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanical life cycle model, combined with a multi-dimensional performance analysis model, includes: Substituting the real-time dust accumulation density into the semi-empirical efficiency prediction formula constructed based on the dust deposition mechanical life cycle model, the formula characterizes the exponential decay relationship between the actual power generation efficiency after dust accumulation and the rated power generation efficiency under clean conditions. The decay relationship is jointly determined by the correction coefficient, real-time dust accumulation density, dust particle size distribution, number of dust particles, dust accumulation duration, effective light-receiving area of ​​photovoltaic panel, and dust particle deposition rate. By considering the triple impact of dust accumulation on photovoltaic panel transmittance, reflectivity, and module temperature, the formula prediction results are corrected, and the predicted value of power generation efficiency loss is output.

[0035] In this embodiment, the specific implementation method for predicting the power generation efficiency loss of photovoltaic panels due to dust accumulation in step S103, based on real-time dust accumulation data and the critical conditions in the dust deposition mechanics life cycle model, combined with a multi-dimensional performance analysis model, is as follows: First, the system obtains the real-time dust accumulation density ρ output by the SolNet identification unit from the real-time monitoring module. This data is in grams per square meter. Simultaneously, it obtains the dust particle size distribution data determined in the dust deposition mechanics life cycle model from the dust accumulation mechanism quantification module, including the particle diameter of the i-th type of dust particle size. and its proportion coefficient And the number of dust particles per unit area N, and the dust particle deposition rate The system also obtains parameters such as the effective light-receiving area A of the photovoltaic panel. In addition, it acquires parameters such as current ambient humidity and wind speed from an environmental sensor array to determine the correction coefficient 'a'. These data are then substituted into the semi-empirical efficiency prediction formula. Perform calculations, where The formula represents the rated power generation efficiency of the photovoltaic panel under clean conditions, where T is the duration of dust accumulation. This formula outputs the preliminary actual power generation efficiency. Based on this, the system further refines the preliminary prediction results using a multi-dimensional performance analysis model. Specifically, the photovoltaic panel status acquisition device in the real-time monitoring module continuously collects performance parameters such as the photovoltaic panel's transmittance, reflectance, and module temperature. The multi-dimensional performance analysis model corrects the formula prediction results based on the triple impact path of dust accumulation on the photovoltaic panel: decreased transmittance, decreased reflectance, and increased module temperature. For example, when a decrease in transmittance, a decrease in reflectance, or an increase in module temperature is detected, the system adjusts the prediction according to preset correction rules. Adjustments are made accordingly to output a more accurate predicted value for power generation efficiency loss. This predicted value is expressed as a percentage and is used as decision input for the subsequent cleaning scheme optimization module. Through the above steps, a complete processing flow from real-time dust accumulation data input to power generation efficiency loss prediction output is achieved, ensuring the accuracy and reliability of the prediction results. This specific implementation method, by combining a semi-empirical efficiency prediction formula with a multi-dimensional performance analysis model, significantly improves the accuracy of power generation efficiency loss prediction, providing a solid data foundation for the dynamic optimization of subsequent cleaning schemes, thereby effectively avoiding the problems of insufficient or excessive cleaning and reducing the operation and maintenance costs of photovoltaic power plants.

[0036] Based on the above embodiments, in some embodiments, the step of dynamically optimizing the cleaning cycle and cleaning method through a deep residual neural network based on the predicted results of the power generation efficiency loss, generating cleaning decisions, and controlling the intelligent cleaning equipment to perform cleaning operations includes: Based on the quantitative relationship between dust density and power generation efficiency, multiple efficiency loss thresholds are set. When the predicted efficiency loss reaches the corresponding threshold, a cleaning command is triggered. The optimal cleaning interval is dynamically calculated by combining the dust deposition rate and environmental changes to optimize the cleaning cycle. The cleaning method is selected based on dust density, dust particle size, photovoltaic panel installation tilt angle and power station site environment. When the dust density is below the first threshold, low-pressure spray cleaning by drone is used. When the dust density is above the second threshold or an adhesive layer is formed, physical wiping and spray cleaning by automatic cleaning robot is used. When rainfall meets the critical cleaning conditions, manual or intelligent cleaning is suspended and natural precipitation is used to assist in cleaning. Cleaning decisions are generated and intelligent cleaning equipment is controlled to perform cleaning operations.

[0037] Furthermore, after the intelligent cleaning device performs the cleaning operation, it also includes: After cleaning, the dust accumulation status and performance parameters of the photovoltaic panels are collected again by the environmental sensor group and the photovoltaic panel status acquisition device to verify the efficiency recovery after cleaning. If the expected effect is not achieved, the cleaning plan is optimized and a second cleaning is triggered.

[0038] In this embodiment, the specific implementation of step S104, which involves dynamically optimizing the cleaning cycle and cleaning method based on the predicted power generation efficiency loss, generating cleaning decisions, and controlling the intelligent cleaning equipment to perform cleaning operations, is as follows: First, the cleaning scheme optimization module receives the predicted power generation efficiency loss from the efficiency loss prediction module. This result includes the predicted percentage of efficiency degradation of the current photovoltaic panel. Based on the quantitative relationship between dust density and power generation efficiency, the system pre-sets multiple efficiency loss thresholds, such as key nodes like 10%, 20%, and 50% efficiency degradation. When the predicted efficiency loss reaches one of these thresholds, the system triggers the corresponding cleaning command. Simultaneously, combining the dust deposition rate with the changing trends of environmental parameters, the deep residual neural network dynamically calculates the optimal cleaning interval, thereby optimizing the cleaning cycle and avoiding insufficient or excessive cleaning. Second, regarding the selection of the cleaning method, the system makes a comprehensive decision based on the dust density, dust particle size distribution, photovoltaic panel installation tilt angle, and power plant site environmental parameters in the real-time dust data. When the dust density is low, the system selects a drone for low-pressure spray cleaning. When the dust density is high or the dust has formed an adhesive layer, the system selects an automatic cleaning robot for physical wiping and spray cleaning. When rainfall meets the critical cleaning conditions (rain intensity greater than 38 mm / h and dust density less than 40 mg / m²), the system suspends manual or intelligent cleaning commands, utilizes natural precipitation to assist cleaning, and monitors the dust status after the rainfall ends, only cleaning areas that were not properly cleaned. During the cleaning process, the intelligent cleaning equipment is controlled to perform cleaning operations in the following ways: the automatic cleaning robot is controlled to adapt to photovoltaic panels with different tilt angles, using path planning and autonomous obstacle avoidance functions, and employing a gentle wiping combined with high-pressure micro-mist cleaning method; or the drone cleaning equipment is controlled to perform light cleaning of large areas or high-altitude photovoltaic power stations using low-pressure spray. After cleaning, the environmental sensor group and photovoltaic panel status acquisition device collect the dust status and performance parameters of the photovoltaic panels again to verify the efficiency recovery after cleaning. If the expected effect is not achieved, the system re-optimizes the cleaning plan and triggers a second cleaning. Through the above steps, the dynamic optimization and precise execution of the cleaning cycle and cleaning method are achieved.

[0039] This specific implementation significantly improves the accuracy and timeliness of cleaning decisions by setting multi-level efficiency loss thresholds and combining them with deep residual neural networks to dynamically optimize the cleaning cycle. It avoids resource waste or insufficient cleaning caused by traditional fixed-cycle cleaning. At the same time, it flexibly selects cleaning methods according to dust accumulation and environmental conditions, and introduces natural precipitation to assist cleaning, effectively reducing cleaning costs and improving the overall operation and maintenance efficiency of photovoltaic power plants.

[0040] Based on the above embodiments, in some embodiments, cleaning effect data is incorporated into the background database to continuously optimize the parameters of the dust deposition mechanical life cycle model and the algorithm of the deep residual neural network, thereby realizing the self-learning and self-optimization of the prevention and control system.

[0041] In this embodiment, the method further includes incorporating cleaning effect data into a background database and continuously optimizing the parameters of the dust deposition mechanical lifecycle model and the algorithm of the deep residual neural network to achieve self-learning and self-optimization of the prevention and control system. Specifically, after the intelligent cleaning equipment completes the cleaning operation, the photovoltaic panel status acquisition device in the real-time monitoring module collects images of dust accumulation on the photovoltaic panel surface and performance parameters such as output power, component temperature, transmittance, and reflectance. This data is then input into the SolNet convolutional neural network recognition unit to output the real-time dust accumulation density ρ' after cleaning. Simultaneously, the environmental sensor group continues to collect environmental parameters such as wind speed, humidity, and precipitation intensity after cleaning. This cleaning effect data, including the dust accumulation density ρ' after cleaning, the performance parameter recovery values, the actual water consumption during cleaning, and the cleaning duration, is uniformly transmitted to the database of the background system. The background system associates and stores the cleaning effect data with the dust accumulation data before cleaning, the efficiency loss prediction results, and the cleaning decision. Subsequently, the system optimizes the key parameters in the dust deposition mechanical lifecycle model based on these newly entered cleaning effect data. For example, based on the correspondence between the actual change in dust density before and after cleaning and cleaning operation parameters (such as spray pressure and number of wipes), the coefficients in the model regarding dust removal efficiency can be corrected; or, based on the deviation between the actual value and the predicted value of efficiency recovery after cleaning, the correction coefficient 'a' in the semi-empirical efficiency prediction formula can be adjusted. Simultaneously, the Deep Residual Neural Network (DRNN) algorithm also utilizes this new data as training samples for incremental learning, optimizing its internal weight parameters to improve the accuracy of future decisions regarding cleaning cycles and methods. For instance, if multiple cleaning data indicate that, under specific environmental conditions, after using low-pressure spray cleaning with a drone, the dust density recovers to the trigger cleaning threshold within a short period (e.g., 3 days), the DRNN algorithm will learn that the effective cycle of this cleaning method is short under those conditions. Consequently, in subsequent decisions, when similar environmental conditions are predicted, it will tend to choose an automated cleaning robot solution with a longer-lasting cleaning effect, or shorten the cleaning cycle. By continuously feeding cleaning performance data back to the model and algorithm, the entire dust control system can constantly adjust and optimize itself based on the actual operating data of the photovoltaic power plant and environmental changes. This allows the accuracy of dust monitoring, efficiency loss prediction, and cleaning decisions to gradually improve over time, achieving the system's self-learning and self-optimization capabilities. The beneficial effects of this specific implementation method are that by incorporating cleaning performance data into the database and continuously optimizing model parameters and algorithms, the system can adapt to the long-term operating environment and dust characteristics of different photovoltaic power plants. This avoids prediction and decision-making biases caused by model rigidity, significantly improving the system's monitoring accuracy, prediction accuracy, and the rationality of cleaning decisions during long-term operation, thereby further reducing operation and maintenance costs and improving power generation efficiency.

[0042] Please refer to Figure 2 , Figure 2 This invention provides a structural block diagram of an intelligent dust control device for photovoltaic panels; the specific device may include: The Dust Accumulation Mechanism Quantification Module 100 is used to construct a dust deposition mechanical life cycle model based on multi-factor coupling analysis combining computational fluid dynamics and discrete element method, quantify the interaction between environmental parameters, dust characteristics and photovoltaic modules, and determine the weight of dust adsorption force, the influence law of dust particle size and the critical conditions for precipitation cleaning. The real-time monitoring module 200 is used to collect environmental parameters and photovoltaic panel performance parameters in real time through an environmental sensor group and a photovoltaic panel status acquisition device, and to use a convolutional neural network to identify dust accumulation images on the surface of the photovoltaic panel and output real-time dust accumulation data. The efficiency loss prediction module 300 is used to predict the power generation efficiency loss of photovoltaic panels due to dust accumulation based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanical life cycle model, combined with a multi-dimensional performance analysis model. The multi-dimensional performance analysis model includes at least the influence path of dust accumulation on the transmittance, reflectance and component temperature of the photovoltaic panel. The cleaning scheme optimization module 400 is used to dynamically optimize the cleaning cycle and cleaning method based on the predicted results of the power generation efficiency loss through a deep residual neural network, and generate a cleaning decision; the intelligent cleaning execution module is used to control the intelligent cleaning equipment to perform cleaning operations according to the cleaning decision.

[0043] The intelligent dust control device for photovoltaic panels in this embodiment is used to implement the aforementioned intelligent dust control method for photovoltaic panels. Therefore, the specific implementation of the intelligent dust control device for photovoltaic panels can be found in the previous embodiment section of the intelligent dust control method for photovoltaic panels. For example, the dust accumulation mechanism quantification module 100, the real-time monitoring module 200, the efficiency loss prediction module 300, and the cleaning scheme optimization module 400 are respectively used to implement steps S101, S102, S103, and S104 in the aforementioned intelligent dust control method for photovoltaic panels. Therefore, the specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0044] Based on the above embodiments, in some embodiments, the following further includes: The intelligent cleaning execution module is used to control intelligent cleaning equipment to perform cleaning operations based on the cleaning decision, wherein the intelligent cleaning equipment includes automated cleaning robots and / or drone cleaning equipment.

[0045] In this embodiment, the complete implementation of the intelligent dust prevention and control method for photovoltaic panels will be described in detail, including the system architecture, the specific implementation method of each module, the data flow path, parameter configuration, algorithm logic, and boundary condition handling, etc., to ensure that those skilled in the art can implement the technical solution based on the description of this embodiment.

[0046] The method described in this embodiment relies on a system composed of multiple collaborative modules, including a dust accumulation mechanism quantification module, a real-time monitoring module, an efficiency loss prediction module, a cleaning scheme optimization module, and an intelligent cleaning execution module. The dust accumulation mechanism quantification module serves as the theoretical foundation of the entire system. It constructs a dust deposition mechanical lifecycle model using a multi-factor coupled analysis method combining computational fluid dynamics (CFD) and discrete element method (DEM). This model quantifies the interaction between environmental parameters, dust characteristics, and photovoltaic modules. Specifically, in a dry environment, van der Waals forces are the primary adsorption force; in a humid environment, capillary forces are the primary adsorption force. When the ambient humidity is greater than 60%, capillary forces account for 98%, and this force weight is incorporated into the subsequent prediction model. Simultaneously, the influence of dust particle size is clarified: particles smaller than 10 μm can migrate thousands of kilometers, while particles larger than 20 μm mainly settle within 40 km. Differentiated monitoring and cleaning thresholds are set for particles of different sizes. In addition, the critical conditions for precipitation cleaning were determined, namely, a rainfall intensity of 38 mm / h and a dust accumulation density of 40 mg / m² as the critical thresholds. When the rainfall intensity is greater than 38 mm / h and the dust accumulation density is less than 40 mg / m², the precipitation is judged to have a cleaning effect; otherwise, the precipitation is judged to have no cleaning effect or even aggravate dust accumulation. This critical condition serves as an auxiliary basis for judging natural cleaning.

[0047] The real-time monitoring module consists of an environmental sensor array, a photovoltaic panel status acquisition device, and a SolNet convolutional neural network recognition unit. The environmental sensor array collects environmental parameters such as wind speed, humidity, precipitation intensity, atmospheric particulate matter concentration, and industrial emission indicators around the photovoltaic power station in real time, and transmits them to the backend processing system via wired or wireless communication. The photovoltaic panel status acquisition device acquires dust images of the photovoltaic panel surface using high-definition imaging equipment, and simultaneously collects performance parameters such as the photovoltaic panel's output power, module temperature, light transmittance, and reflectivity. The acquired dust images are input into the SolNet convolutional neural network, which achieves a recognition accuracy of 98.2%. Through image recognition algorithms, it accurately identifies key indicators such as dust density, dust particle size distribution, and dust shape on the photovoltaic panel surface, ultimately outputting the real-time dust density ρ in g / m². In one possible implementation, the high-definition imaging equipment can be deployed at a density of one unit per 100 photovoltaic panels to ensure comprehensive monitoring coverage.

[0048] The efficiency loss prediction module, based on the quantification results of dust accumulation mechanisms and real-time dust accumulation data output by the real-time monitoring module, accurately predicts the power generation efficiency loss of photovoltaic panels caused by dust accumulation through a semi-empirical efficiency prediction formula and a multi-dimensional performance analysis model. The core semi-empirical efficiency prediction formula is:

[0049] in, This represents the actual power generation efficiency of the photovoltaic panels after dust accumulation. ρ represents the rated power generation efficiency of the photovoltaic panel under clean conditions, α is a correction factor (dynamically determined by parameters such as ambient humidity and wind speed), and ρ is the real-time dust density. Let be the particle diameter (in μm) of the i-th type of dust. Let be the proportion coefficient of dust with the i-th particle size, N be the number of dust particles per unit area, T be the dust accumulation duration (in hours), and A be the effective light-receiving area of ​​the photovoltaic panel (in m²). This represents the dust particle deposition rate (in m / h). The multi-dimensional performance analysis model incorporates the three-way impact of dust accumulation on photovoltaic panels: decreased transmittance (maximum decrease of 35%), decreased reflectance (maximum decrease of 9.4%), and increased module temperature (maximum increase of 10°C). This data is used to correct the formula's prediction results, thereby improving the accuracy of efficiency loss prediction. For example, when a 17.5% decrease in transmittance and a 5°C increase in module temperature are detected, the efficiency degradation value calculated by the formula can be corrected, resulting in a more accurate predicted value for power generation efficiency loss.

[0050] The cleaning scheme optimization module, as the core decision-making module, dynamically optimizes the cleaning cycle and cleaning method based on real-time monitoring data and efficiency loss prediction results, generating the optimal cleaning decision through a deep residual neural network (DRNN) algorithm. Regarding cleaning cycle optimization, based on the quantitative relationship between dust density and power generation efficiency (e.g., when the dust density reaches 14.45 g / m², the output power decreases by nearly 50%), multiple efficiency loss thresholds are set, such as 10%, 20%, and 50% decrease. When the predicted efficiency loss reaches the corresponding threshold, a cleaning command is triggered. Simultaneously, combining the dust deposition rate and environmental change trends, the optimal cleaning interval is dynamically calculated to achieve precise control of the cleaning cycle. Regarding cleaning method optimization, the optimal cleaning method is selected based on dust density, dust particle size, photovoltaic panel installation tilt angle, and the power plant site environment. Specifically, when the dust density is low, low-pressure spray cleaning is carried out using drones; when the dust density is high or an adhesive layer forms, an automatic cleaning robot is used for physical wiping and spray cleaning, with a cleaning efficiency of up to 95%; when rainfall meets the critical cleaning conditions (rainfall intensity greater than 38 mm / h and dust density less than 40 mg / m²), manual or intelligent cleaning is suspended, and natural precipitation is used to assist in cleaning to reduce cleaning costs.

[0051] The intelligent cleaning execution module serves as the implementation unit for the solution. It incorporates cleaning decisions output by the cleaning solution optimization module and combines automated cleaning robots and drone cleaning equipment to achieve intelligent and automated cleaning of photovoltaic panels. The automated cleaning robot is adaptable to photovoltaic panels with different tilt angles, such as a standard 25° tilt angle. It features path planning and autonomous obstacle avoidance, employing a combination of flexible wiping and high-pressure micro-mist cleaning, achieving a cleaning efficiency of 95%, a 32% improvement over traditional cleaning methods. The drone cleaning equipment is suitable for large-area, high-altitude photovoltaic power plants. It achieves light cleaning of accumulated dust through low-pressure spraying, featuring a wide operating range and high cleaning efficiency. In one possible implementation, the power plant can be configured with one automated cleaning robot for every 500 photovoltaic panels and one drone cleaning equipment for every 1000 photovoltaic panels. After cleaning, the real-time monitoring module collects the dust accumulation status and performance parameters of the photovoltaic panels again, transmitting this data to the efficiency loss prediction module to verify the efficiency recovery after cleaning. If the expected results are not achieved, the cleaning solution is re-optimized and a second cleaning is triggered. In addition, the back-end system incorporates cleaning performance data into the database and continuously optimizes the parameters of the dust deposition mechanical life cycle model and the algorithm of the deep residual neural network to achieve self-learning and self-optimization of the prevention and control system.

[0052] Taking the photovoltaic power station in the Shagohuang desert as a specific implementation scenario, the photovoltaic panels are installed at a tilt angle of 25°. This area has an arid climate, scarce rainfall, and high concentrations of atmospheric particulate matter, resulting in severe dust accumulation; natural precipitation is almost ineffective for cleaning. During system deployment, environmental sensor groups are deployed regionally within the photovoltaic power station. One high-definition imaging device is provided for every 100 photovoltaic panels. The backend system incorporates a dust accumulation mechanism quantification model, a SolNet recognition unit, and a DRNN decision algorithm. The power station is equipped with an automated cleaning robot (one robot for every 500 photovoltaic panels) and a drone cleaning device (one drone for every 1000 photovoltaic panels). During real-time monitoring, the environmental sensor groups collect data on wind speed, humidity (typically less than 60% year-round, with van der Waals forces as the primary adsorption force), and atmospheric particulate matter concentration (mainly fine particles smaller than 10μm) in the desert area. The high-definition imaging device captures images of dust accumulation on the photovoltaic panel surface. After SolNet recognition, the real-time dust density ρ is output. Initially, ρ = 0 g / m², indicating the photovoltaic panels are clean. As dust accumulation continued, after 15 days of operation, a density of ρ = 7.225 g / m² was monitored. Substituting this data into the semi-empirical efficiency prediction formula, combined with the correction coefficient 'a' for this area, it was predicted that the photovoltaic panel's output power would decrease by approximately 25%. Simultaneously, a 17.5% decrease in photovoltaic panel transmittance and a 5°C increase in module temperature were observed. After correcting the prediction, a 24% efficiency decrease was confirmed. Based on the predicted 24% efficiency decrease and the dust deposition rate in the area, the cleaning solution optimization module used a DRNN algorithm to determine that cleaning was required within 3 days. Since the dust density did not reach 14.45 g / m², a low-pressure spray cleaning method using a drone was selected. After receiving the cleaning command, the drone performed low-pressure spray cleaning on the photovoltaic panel according to the preset path. After the operation was completed, the real-time monitoring module collected data again, showing ρ = 0.8 g / m², and the photovoltaic panel's output power recovered to 98% of the clean state, indicating that the cleaning effect met the standards. The backend system recorded the duration of dust accumulation from 0 g / m² to 7.225 g / m² as 15 days. Considering subsequent changes in environmental parameters (such as increased particulate matter concentration due to sandstorms), the system predicted the next time dust accumulation would reach the 20% attenuation threshold in 10 days. The cleaning cycle was optimized to 10 days, and the cleaning data was incorporated into the database, with the formula correction coefficient 'a' optimized. If the power station experiences light rainfall (with a rainfall intensity of 25 mm / h to 38 mm / h), the system determines that the precipitation will form a sticky layer, immediately triggering an early warning and initiating light cleaning by the robot to prevent a reversal in dust density. If a rare heavy rain occurs (with a rainfall intensity of 45 mm / h and a dust density of 35 mg / m² to 40 mg / m²), the system determines that natural precipitation has a cleaning effect, suspends the preset cleaning commands, and monitors the dust accumulation after the rainfall ends, only cleaning areas that were not properly cleaned.

[0053] This embodiment achieves closed-loop management of the entire process, from quantifying dust accumulation mechanisms, real-time monitoring, efficiency loss prediction, cleaning scheme optimization to intelligent cleaning execution, through the collaborative work of the above modules. It also has self-learning and self-optimization capabilities, which can significantly improve the intelligence level and cleaning efficiency of photovoltaic panel dust prevention and control, and reduce operation and maintenance costs.

[0054] A specific embodiment of the present invention also provides an intelligent dust control device for photovoltaic panels, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the aforementioned intelligent dust control method for photovoltaic panels.

[0055] A specific embodiment of the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described intelligent dust prevention and control method for photovoltaic panels.

[0056] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0057] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0058] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0059] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0060] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for intelligent dust control on photovoltaic panels, characterized in that, Includes the following steps: Based on multi-factor coupling analysis combining computational fluid dynamics and discrete element method, a dust deposition mechanical life cycle model is constructed to quantify the interaction between environmental parameters, dust characteristics and photovoltaic modules, and to determine the weight of dust adsorption force, the influence law of dust particle size and the critical condition for precipitation cleaning. The system collects environmental parameters and photovoltaic panel performance parameters in real time using an environmental sensor array and a photovoltaic panel status acquisition device. It also uses a convolutional neural network to identify dust accumulation images on the photovoltaic panel surface and outputs real-time dust accumulation data. Based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanical life cycle model, combined with the multi-dimensional performance analysis model, the power generation efficiency loss of photovoltaic panels due to dust accumulation is predicted. The multi-dimensional performance analysis model includes at least the influence path of dust accumulation on the transmittance, reflectance and temperature of photovoltaic panels. Based on the predicted power generation efficiency loss, the cleaning cycle and cleaning method are dynamically optimized through a deep residual neural network to generate cleaning decisions and control the intelligent cleaning equipment to perform cleaning operations.

2. The method according to claim 1, characterized in that, The multi-factor coupled analysis based on computational fluid dynamics and discrete element method constructs a dust deposition mechanical life cycle model, quantifies the interaction between environmental parameters, dust characteristics, and photovoltaic modules, and determines the weight of dust adsorption force, the influence law of dust particle size, and the critical conditions for precipitation cleaning, including: The van der Waals forces under dry conditions and the capillary forces under humid conditions are quantified as the forces for dust adsorption, and the weight of the capillary forces is determined according to the ambient humidity and incorporated into the prediction model. Clearly define the differences in migration and sedimentation between small-diameter and large-diameter particles, and set differentiated monitoring and cleaning thresholds for particles of different diameters; Precipitation intensity and dust density are determined as the joint criteria for determining the critical condition for precipitation cleaning. When the precipitation intensity is greater than the first threshold and the dust density is less than the second threshold, the precipitation is determined to have a cleaning effect. This critical condition is used as an auxiliary criterion for natural cleaning.

3. The method according to claim 1, characterized in that, The process involves real-time acquisition of environmental parameters and photovoltaic panel performance parameters via an environmental sensor array and a photovoltaic panel status acquisition device, and the use of a convolutional neural network to identify dust accumulation images on the photovoltaic panel surface, outputting real-time dust accumulation data, including: The system uses an environmental sensor array to collect real-time data on wind speed, humidity, precipitation intensity, atmospheric particulate matter concentration, and industrial emission indicators around the photovoltaic power station. The high-definition imaging equipment of the photovoltaic panel status acquisition device acquires dust images on the surface of the photovoltaic panel, and simultaneously acquires the output power, component temperature, light transmittance and reflectance of the photovoltaic panel; The dust accumulation image is input into the convolutional neural network SolNet, and the dust density, dust particle size distribution and dust shape on the photovoltaic panel surface are identified by the image recognition algorithm, and the real-time dust accumulation density is output.

4. The method according to claim 1, characterized in that, The prediction of power generation efficiency loss of photovoltaic panels due to dust accumulation, based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanical life cycle model, combined with a multi-dimensional performance analysis model, includes: Substituting the real-time dust accumulation density into the semi-empirical efficiency prediction formula constructed based on the dust deposition mechanical life cycle model, the formula characterizes the exponential decay relationship between the actual power generation efficiency after dust accumulation and the rated power generation efficiency under clean conditions. The decay relationship is jointly determined by the correction coefficient, real-time dust accumulation density, dust particle size distribution, number of dust particles, dust accumulation duration, effective light-receiving area of ​​photovoltaic panel, and dust particle deposition rate. By considering the triple impact of dust accumulation on photovoltaic panel transmittance, reflectivity, and module temperature, the formula prediction results are corrected, and the predicted value of power generation efficiency loss is output.

5. The method according to claim 1, characterized in that, The process of dynamically optimizing the cleaning cycle and cleaning method based on the predicted power generation efficiency loss using a deep residual neural network, generating cleaning decisions, and controlling intelligent cleaning equipment to perform cleaning operations includes: Based on the quantitative relationship between dust density and power generation efficiency, multiple efficiency loss thresholds are set. When the predicted efficiency loss reaches the corresponding threshold, a cleaning command is triggered. The optimal cleaning interval is dynamically calculated by combining the dust deposition rate and environmental changes to optimize the cleaning cycle. The cleaning method is selected based on dust density, dust particle size, photovoltaic panel installation tilt angle and power station site environment. When the dust density is below the first threshold, low-pressure spray cleaning by drone is used. When the dust density is above the second threshold or an adhesive layer is formed, physical wiping and spray cleaning by automatic cleaning robot is used. When rainfall meets the critical cleaning conditions, manual or intelligent cleaning is suspended and natural precipitation is used to assist in cleaning. Cleaning decisions are generated and intelligent cleaning equipment is controlled to perform cleaning operations.

6. The method according to claim 5, characterized in that, After the intelligent cleaning device is controlled to perform the cleaning operation, the following is also included: After cleaning, the dust accumulation status and performance parameters of the photovoltaic panels are collected again by the environmental sensor group and the photovoltaic panel status acquisition device to verify the efficiency recovery after cleaning. If the expected effect is not achieved, the cleaning plan is optimized and a second cleaning is triggered.

7. The method according to claim 1, characterized in that, Also includes: By incorporating cleaning performance data into a backend database, the parameters of the dust deposition mechanical life cycle model and the algorithm of the deep residual neural network are continuously optimized to achieve self-learning and self-optimization of the prevention and control system.

8. A smart dust control device for photovoltaic panels, characterized in that, include: The dust accumulation mechanism quantification module is used to construct a dust deposition mechanical life cycle model based on multi-factor coupling analysis combining computational fluid dynamics and discrete element method, quantify the interaction between environmental parameters, dust characteristics and photovoltaic modules, and determine the weight of dust adsorption force, the influence law of dust particle size and the critical condition for precipitation cleaning. The real-time monitoring module is used to collect environmental parameters and photovoltaic panel performance parameters in real time through an environmental sensor group and a photovoltaic panel status acquisition device, and to use a convolutional neural network to identify dust accumulation images on the surface of the photovoltaic panel and output real-time dust accumulation data. The efficiency loss prediction module is used to predict the power generation efficiency loss of photovoltaic panels due to dust accumulation based on the real-time dust accumulation data and the critical conditions in the dust deposition mechanical life cycle model, combined with a multi-dimensional performance analysis model. The multi-dimensional performance analysis model includes at least the influence path of dust accumulation on the transmittance, reflectance and temperature of the photovoltaic panel. The cleaning scheme optimization module is used to dynamically optimize the cleaning cycle and cleaning method based on the predicted results of the power generation efficiency loss, and generate cleaning decisions through a deep residual neural network. The intelligent cleaning execution module is used to control the intelligent cleaning equipment to perform cleaning operations based on the cleaning decision.

9. A smart dust control device for photovoltaic panels, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the intelligent dust prevention method for photovoltaic panels as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the intelligent dust prevention and control method for photovoltaic panels as described in any one of claims 1 to 7.

Images