A method and algorithmic system for determining the cleaning opportunity of a photovoltaic module

Abstract

The present application relates to a method and algorithm system for determining the cleaning time of photovoltaic modules, which aims to solve the technical problems of insufficient accuracy and unclear physical mechanism of existing dust coverage evaluation methods. The method includes: first, obtaining four specific dust coverage physical parameters of the surface of the photovoltaic module: dust coverage average particle size (de), dust coverage thickness (d), module pollution rate (SR), and dust coverage visual color variation coefficient (k) obtained through a machine learning model; second, substituting the four parameters into a specific nonlinear logarithmic model to calculate the current power loss rate (PL); finally, converting the power loss rate (PL) into the economic cost of power generation loss, and comparing it with the preset cleaning cost to determine the optimal cleaning time. The present application can accurately quantify the impact of dust on the power generation of photovoltaic modules through a nonlinear logarithmic model based on specific multi-physical parameters, and scientifically determine the optimal cleaning time of photovoltaic power stations by combining economic analysis.

Description

Technical Field

[0001] This invention relates to the field of solar photovoltaic power generation technology, and in particular to a method and algorithm system for determining the timing of cleaning photovoltaic modules. Background Technology

[0002] As a crucial component of renewable energy, photovoltaic (PV) power generation's efficiency directly impacts the economic benefits and energy output of the power plant. During the long-term operation of a PV power plant, dust, sand, industrial emissions, and other pollutants inevitably accumulate on the surface of the PV modules. This dust significantly reduces the solar radiation intensity reaching the solar cell surface through optical shading, reflection, and scattering effects, leading to a decrease in the PV module's output power and resulting in power generation loss. Therefore, timely and economical cleaning of PV modules is a critical task in the operation and maintenance management of PV power plants.

[0003] In existing technologies, the main methods for determining the timing of photovoltaic module cleaning are as follows: The first approach is a fixed-cycle maintenance strategy, which involves cleaning all components at preset time intervals (e.g., monthly or quarterly). This method is simple to operate, but it completely ignores the dynamic changes in the rate of dust accumulation at different times and in different areas. This often leads to unnecessary cleaning when the pollution is light, resulting in wasted maintenance costs, or delayed cleaning when the pollution is severe, causing significant power generation losses.

[0004] The second type is an assessment method based on a single physical indicator. For example, some schemes determine the degree of pollution by measuring the dust deposition density (g / m²) per unit area of ​​photovoltaic modules. However, as a macroscopic indicator, deposition density cannot distinguish the differences in the impact of dust with different chemical compositions and particle size distributions on light attenuation. For instance, for the same mass, industrial dust with a darker color and smaller particle size causes a much greater power loss than desert sand dust with a lighter color and larger particle size. Therefore, relying solely on deposition density for decision-making results in a significant discrepancy between the assessment results and the actual power loss.

[0005] The third type is the machine learning-based evaluation method that has emerged in recent years with the development of artificial intelligence technology. This approach typically uses drones or fixed cameras to collect images of photovoltaic modules, and then utilizes deep learning models (such as convolutional neural networks, CNNs) to directly establish an end-to-end mapping between image features and power loss. While it can achieve good fitting results on specific datasets, the internal decision-making logic of these "black box" models is opaque and lacks clear physical mechanism support. Its generalization ability is limited by the diversity of training data; when encountering novel pollutants or environmental conditions not present in the training set, the model's predictive reliability drops significantly. Furthermore, the poor interpretability of the model also brings difficulties to system debugging and optimization.

[0006] In summary, existing technologies for determining the appropriate timing for cleaning photovoltaic modules generally suffer from shortcomings such as insufficient assessment accuracy, unclear physical mechanisms, and weak model generalization ability, making it difficult to meet the urgent needs of refined and economical operation and maintenance of photovoltaic power plants. Therefore, there is an urgent need for a new technological solution that can accurately and interpretably quantify the impact of dust with different physical characteristics on power generation and enable scientific and economical decision-making based on this information. Summary of the Invention

[0007] To help solve the technical problems existing in the prior art, the present invention provides a method and algorithm system for determining the timing of cleaning photovoltaic modules. It can accurately quantify the impact of dust on the power generation of photovoltaic modules through a nonlinear logarithmic model based on specific multiple physical parameters, and scientifically determine the optimal cleaning time for photovoltaic power plants by combining economic analysis.

[0008] This invention discloses a method for determining the timing of cleaning photovoltaic modules, comprising the following steps: S1. Parameter acquisition step: Obtain the physical parameters of the dust covering the surface of the photovoltaic module, wherein the physical parameters of the dust covering include at least: Average particle size (de) of dust cover obtained from sensor data. Dust thickness (d) obtained from sensor data; Component contamination rate (SR) obtained through a dedicated contamination rate measurement tool or image segmentation technique; The dust visual color variation coefficient (k) is obtained by analyzing the dust-covered images of the photovoltaic modules using a machine learning model. S2. Loss Rate Calculation Steps: Based on a preset nonlinear logarithmic model, the current power loss rate (PL) of the photovoltaic module is calculated using the dust-covered physical parameters. The mathematical expression of the nonlinear logarithmic model is:

[0009] Wherein, k is the base of the logarithmic model, and the value range of k is limited to (0, 1); S3. Decision-making steps: Convert the current power loss rate (PL) into the economic cost of power generation loss caused by dust accumulation, and determine the optimal time to clean the photovoltaic module based on the comparison between the preset cleaning cost and the economic cost of power generation loss.

[0010] It is understood that the core of the method for determining the optimal cleaning time for photovoltaic modules disclosed in this invention lies in accurately quantifying the impact of dust on power through an innovative multi-parameter physical model. This method first establishes a four-dimensional physical parameter system describing the dust-covered state through the S1 parameter acquisition step. This system not only includes the module contamination rate (SR), reflecting the macroscopic dust coverage, but also delves into the microscopic level, introducing the average particle size (de) and dust thickness (d), reflecting the three-dimensional physical structure of the dust. More importantly, this system creatively quantifies the complex chemical composition and optical properties of the dust into a unified, mathematically meaningful dust visual color variation coefficient (k) through a machine learning model. These four parameters together constitute a more comprehensive description of the physical characteristics of the dust layer than any single indicator in existing technologies.

[0011] After obtaining the aforementioned four-dimensional parameters, the method proceeds to the S2 loss rate calculation step. This step is the core embodiment of the technical concept of this invention, which abandons traditional linear fitting or general black-box AI models and instead adopts a structure-specific nonlinear logarithmic model.

[0012] In this model, there are close, physically-based interactions among the parameters. The de / d ratio reflects the macroscopic impact of the microscopic stacking structure of the dust layer on light attenuation; that is, at the same thickness, a dust layer composed of smaller particles results in greater power loss due to its stronger scattering effect. The most original part of the model is... The physical significance of placing the color coefficient k at the base of the logarithmic model lies in the fact that the color / composition (k) of dust cover does not simply linearly scale the power loss value, but fundamentally alters the mathematical law governing the impact of pollution coverage (SR) on power loss. When the k value is small (e.g., dark carbonaceous dust), even small SR values ​​are amplified through the logarithmic relationship, leading to significant power loss; while when the k value is large (e.g., light-colored desert sand), the contribution of SR to power loss is relatively gradual. This design profoundly reflects the complex, nonlinear coupling effect between the "quality" (color / composition) and "quantity" (coverage) of dust cover.

[0013] Finally, in the S3 decision-making step, the method uses the highly accurate physical power loss rate (PL) calculated in S2 as input, combined with the actual economic parameters of the photovoltaic power plant (such as feed-in tariff and cleaning costs). By establishing a cost-benefit analysis model, the purely technical indicator (PL) is transformed into a direct economic indicator (the amount of power generation loss), and compared with the expenditure of cleaning operations, thereby deriving an optimal cleaning time with clear economic significance. This step completes a closed loop from accurate physical quantification to scientific economic decision-making, ensuring that the final cleaning timing recommendation has the highest economic rationality. The entire method, through the close coordination of these three steps, achieves in-depth quantification of the impact of dust accumulation and precise guidance for operation and maintenance decisions.

[0014] According to a method for determining the timing of photovoltaic module cleaning according to the present invention, in the parameter acquisition step (S1), a drone equipped with image acquisition equipment, a laser thickness sensor, and a laser scattering sensor is used to fly along a preset route generated based on the layout information of the photovoltaic power station to collect raw data for generating the dust-covered physical parameters. It can be understood that, in order to achieve efficient, comprehensive, and accurate data collection of dust-covered parameters for all modules in a large photovoltaic power station, the parameter acquisition step (S1) is performed using an automated inspection scheme employing a drone equipped with multiple sensors. Photovoltaic power stations cover a large area and have a large number of modules. Traditional methods, such as manual fixed-point sampling or deploying limited ground monitoring equipment, lack spatial representativeness and struggle to reflect the non-uniformity of dust distribution across the entire station. This scheme utilizes the high mobility of the drone, flying along a preset route generated based on the power station layout information, ensuring the repeatability of the detection path and comprehensive coverage of all modules. Simultaneously, integrating image acquisition equipment, laser thickness sensors, and laser scattering sensors onto the same UAV platform allows for the simultaneous acquisition of raw data necessary for calculating all four dust-covering physical parameters during a single flight mission. This significantly improves data acquisition efficiency and temporal synchronization between data points, providing a high-quality, high-spatiotemporal-resolution data foundation for subsequent accurate model calculations.

[0015] According to a method for determining the timing of cleaning photovoltaic modules according to the present invention, the method for determining the visual color variation coefficient (k) of dust cover includes: establishing a training dataset containing multiple sets of "dust cover image-k value label" sample pairs, wherein the k value label is obtained by manually calibrating the color type of the dust cover image, and the calibration rule is: the surface color of the photovoltaic module in a clean state is calibrated as 1, the exposed ground color of the photovoltaic module installation area is calibrated as 0, and other colors of dust cover are assigned a score between 0 and 1 according to their visual difference from the two reference colors; using the training dataset to train a neural network model, decision tree model or support vector machine model constrained by physical boundaries to obtain a trained model that can input dust cover images and output corresponding k values; in the parameter acquisition step (S1), the collected dust cover image of the photovoltaic module is input into the trained model to obtain the visual color variation coefficient (k) of dust cover.

[0016] To objectively and quantitatively represent the physical characteristic of dust color and thus make the core model engineering-feasible, this solution provides a specific method for determining the visual color variation coefficient k of dust. The method first establishes a manually calibrated training dataset containing multiple pairs of "dust image-k-value label" samples. The calibration rules have clear physical boundaries: the surface color of a clean photovoltaic module is labeled as 1, and the color of the exposed ground in the photovoltaic module installation area is labeled as 0. Other colors of dust are assigned a score between 0 and 1 as a k-value label by professionals based on their visual difference from the two benchmark colors. Subsequently, this training dataset is used to train a neural network model, decision tree model, or support vector machine model constrained by physical boundaries, ultimately resulting in a trained model that can automatically output the corresponding k-value when a dust image is input. This method effectively and indirectly quantifies the complex chemical composition and optical properties of dust through its external visual representation, solving the problem that traditional methods cannot uniformly represent the optical effects of different types of dust.

[0017] According to a method for determining the timing of cleaning photovoltaic modules according to the present invention, the input of the trained model is color features extracted based on the dust-covered image, wherein the color features are the average color value or color histogram of the dust-covered image in the RGB, HSV or Lab color space.

[0018] It is understandable that in this approach, by converting the dust-covered image into average color values ​​or color histograms in specific color spaces such as RGB, HSV, or Lab, the core color information of the image can be effectively represented with low-dimensional feature vectors. This aims to reduce the computational complexity of the model and the demand for computing resources. At the same time, this standardized feature extraction method also reduces the impact of non-critical factors such as image resolution and illumination changes, thereby enhancing the model's stability and generalization ability across different scenes.

[0019] According to a method for determining the timing of photovoltaic module cleaning according to the present invention, the nonlinear logarithmic model is expressed in the form of a changing-base formula during calculation:

[0020] Where ln is the natural logarithm. It can be understood that in the core loss rate calculation step (S2), to ensure the computational stability and universality of the nonlinear logarithmic model in software engineering practice, the model can be expressed using a changed-base formula: PL = (de / d) * [ln(SR) / ln(k)], where ln is the natural logarithm. This mathematically equivalent transformation aims to avoid the potential numerical instability or undefined problems that may arise from directly calculating the logarithm to base k when k is equal to or close to 1. The transformed expression has direct, efficient, and numerically stable implementations in all mainstream programming environments and computing libraries, ensuring the reliability of the core model in engineering applications.

[0021] According to a method for determining the timing of photovoltaic module cleaning according to the present invention, the decision step (S3) further includes: obtaining the feed-in tariff or transaction tariff of the photovoltaic power plant, and calculating the economic cost of power generation loss per unit time in combination with the current power loss rate (PL); obtaining the estimated cost of a single cleaning operation as the cleaning cost; establishing a prediction function for the accumulation of the economic cost of power generation loss over time; calculating and determining the time point when the cumulative value of the economic cost of power generation loss first equals or exceeds the cleaning cost through the prediction function, and taking this time point as the optimal cleaning timing.

[0022] Understandably, in this scheme, to transform the physical power loss rate into an economically meaningful decision with direct commercial guidance, the decision-making step (S3) can execute a specific decision-making algorithm based on economic benefit optimization. This algorithm obtains the grid connection price or trading price of the photovoltaic power plant and combines it with the calculated current power loss rate (PL) to calculate the economic cost of power generation loss per unit time. Simultaneously, it obtains the estimated cost of a single cleaning operation as the cleaning cost. By establishing a prediction function for the cumulative economic cost of power generation loss over time, the algorithm can calculate and determine the point in time when the cumulative loss value first equals or exceeds the cleaning cost, and designates this point as the optimal cleaning time. This approach enables operation and maintenance decisions to be based on rigorous economic calculations, achieving scientific management of the input-output ratio of clean operation and maintenance.

[0023] According to a method for determining the timing of photovoltaic module cleaning according to the present invention, the decision step (S3) further includes an early warning mechanism: when the current power loss rate (PL) exceeds a fixed threshold preset based on historical data and operation and maintenance experience, the system triggers a cleaning early warning and marks the photovoltaic module with a bright color on the power plant heat map. It can be understood that, as a supplementary safety measure, the method of the present invention also includes a parallel real-time early warning mechanism. The technical purpose of this mechanism is to provide power plant operation and maintenance personnel with immediate and intuitive risk alerts. When the calculated current power loss rate (PL) exceeds a fixed threshold preset based on historical data and operation and maintenance experience, the system triggers a cleaning early warning. This early warning can be linked with the power plant monitoring system to mark heavily polluted photovoltaic modules or areas with a bright color on the power plant digital twin or heat map. This aims to respond quickly to sudden serious pollution events, guide operation and maintenance personnel to conduct emergency investigations, and prevent irreversible damage to the modules due to hot spot effects and other problems, thereby ensuring the safety of power plant assets.

[0024] According to a method for determining the timing of photovoltaic module cleaning according to the present invention, the method further includes the step of determining the optimal cleaning time interval: periodically performing step S1 to obtain time series data of the module contamination rate (SR) and the dust thickness (d); based on the time series data, establishing a fitting function to describe the changes of the SR value and the d value over time; using the fitting function, predicting and calculating the time length required from the start of the cleaning state until the cumulative value of the economic cost of power generation loss equals the cleaning cost, and determining this time length as the optimal cleaning time interval for the photovoltaic module.

[0025] It is understood that, in order to elevate the decision on the timing of a single cleaning operation to a periodic optimization of the cleaning strategy, the method of the present invention may further include a step of determining the optimal cleaning interval. This step aims to optimize the operation and maintenance strategy from a long-term planning perspective. By periodically performing the parameter acquisition step (S1), time-series data of component contamination rate (SR) and dust accumulation thickness (d) are acquired, and a fitting function describing the dust accumulation rate is established based on this series data. Using this function, the time length required from the start of the cleaning state until the cumulative value of the economic cost of power generation loss equals the cleaning cost can be predicted and calculated. This time length is determined as the optimal cleaning interval, which can serve as the core basis for formulating annual or quarterly operation and maintenance plans, helping power plants to achieve a shift from reactive operation and maintenance to proactive predictive operation and maintenance.

[0026] According to a method for determining the timing of photovoltaic module cleaning according to the present invention, the parameter acquisition step (S1) further includes a model correction step: a fixed ground dust monitoring system is deployed at a typical location of the photovoltaic power station to continuously collect high-precision dust physical parameters as reference data; the reference data is used to correct or train the analysis model used to process the raw data collected by the UAV to improve the calculation accuracy of the dust physical parameters.

[0027] It is understandable that, to ensure the accuracy of the UAV data analysis model during long-term operation, in one specific scheme, the parameter acquisition step (S1) also includes a model correction step. The technical purpose of this step is to establish a closed-loop accuracy assurance system. Specifically, a fixed ground dust monitoring system is deployed at a typical location in the photovoltaic power station. This system continuously collects high-precision dust physical parameters as reference data or true values. By periodically using these reference data to compare, correct, or retrain the analysis model that processes the UAV-collected data, it is possible to effectively compensate for and eliminate system errors introduced by factors such as flight attitude, changes in ambient light, and sensor drift, ensuring high reliability and high accuracy of the entire data acquisition and analysis system during long-term operation.

[0028] Based on the above, the present invention also discloses an algorithm system for determining the cleaning timing of photovoltaic modules to implement the above method, comprising: A parameter acquisition module is configured to receive or process raw data acquired through a sensor network to generate physical parameters of dust accumulation on the surface of the photovoltaic module. These physical parameters include at least: average dust particle size (de), dust thickness (d), module contamination rate (SR), and dust visual color variation coefficient (k). (This module is the core of the entire system's data input and preprocessing, and its technical purpose is to provide the necessary, high-quality four-dimensional dust accumulation physical parameters (de, d, SR, k) for subsequent calculations and decisions. This module is configured to receive raw data from the sensor network and process it into structured physical parameters.) The parameter acquisition module further includes: The drone control unit controls a drone equipped with image acquisition equipment and a laser sensor to collect raw data for generating the dust-covering physical parameters according to a preset flight path, and transmits the raw data through a 5G communication network or a local data interface. (The technical purpose of this module is to achieve automated and efficient inspection of the dust-covering status of large-area photovoltaic power plants. This unit controls a drone equipped with image acquisition equipment and a laser sensor to fly along a preset flight path generated based on the power plant layout information to collect the raw data required to generate the dust-covering physical parameters. After collection, the unit transmits the raw data to the system's data processing center in real time or in batches through a 5G communication network or a local data interface.) The color coefficient calculation submodule contains a machine learning model trained using a "dust-covered image - manually calibrated k value" sample pair. This model is used to calculate the visual color variation coefficient (k) of the dust cover based on the dust-covered image captured by the drone. (The technical purpose of this module is to achieve automated and objective quantification of the optical properties of dust color. This submodule contains a machine learning model trained using a "dust-covered image - manually calibrated k value" sample pair. When a dust-covered image is received from the drone, this submodule uses this model to calculate and directly outputs the visual color variation coefficient (k) of the dust cover corresponding to the image, providing a key input for the calculation of the core model.) The data calibration interface is used to receive reference data from the fixed ground dust monitoring system deployed at the power plant site. This data is used to perform online calibration or offline training on the analysis model that processes data collected by the UAV. (The technical purpose of this module is to establish a closed-loop accuracy assurance and improvement mechanism. This interface is used to receive high-precision reference data from the fixed ground dust monitoring system deployed at the power plant site. This reference data is used as the true value to perform online calibration or offline retraining on the analysis model that processes data collected by the UAV, thereby continuously compensating for and correcting system errors and ensuring that the dust physical parameters output by the system maintain high accuracy over a long period of time.) The loss rate calculation module is functionally connected to the parameter acquisition module and is configured to calculate the loss rate based on a preset mathematical expression.

[0029] The nonlinear logarithmic model is used to calculate the current power loss rate (PL) of the photovoltaic module using the dust-covered physical parameters, where k is the base of the logarithmic model. (The technical purpose of setting up this module is to execute the core physical model calculation of this invention. This module receives four dust-covered physical parameters generated by the parameter acquisition module and accurately calculates the current power loss rate (PL) of the photovoltaic module based on a preset nonlinear logarithmic model with the mathematical expression PL = (de / d) * log_k(SR). This module is a bridge connecting physical phenomena and quantitative evaluation.) The decision-making module, functionally connected to the loss rate calculation module, is configured to convert the current power loss rate (PL) into an economic cost of power generation loss, and based on a comparison of preset cleaning costs and the economic cost of power generation loss, generate decision recommendations for determining the optimal cleaning time for the photovoltaic modules. (The technical purpose of this module is to transform technical evaluation results into operation and maintenance decisions with direct economic guidance. This module receives power loss rate (PL) data and, in conjunction with preset economic parameters such as electricity prices and cleaning costs, converts power loss into an economic cost of power generation loss. By comparing the economic cost of power generation loss with the cleaning cost, this module generates decision recommendations for determining the optimal cleaning time for the photovoltaic modules, such as a specific cleaning date or triggering a cleaning work order.) The cleaning interval calculation module, functionally connected to the parameter acquisition module, is configured to receive and store time-series data of the module contamination rate (SR) and the dust thickness (d), and establish a fitting function based on the time-series data to calculate the optimal cleaning interval for the photovoltaic module. (The technical purpose of this module is to mine dust accumulation patterns from historical data to achieve predictive maintenance. This module is configured to receive and store time-series data of the module contamination rate (SR) and the dust thickness (d), and establish a fitting function describing the rate of pollutant accumulation based on this data. Using this function, the module can predict and calculate the time required to reach an economically viable cleaning threshold from a clean state, and output this time as the recommended optimal cleaning interval, providing a scientific basis for the power plant to formulate medium- and long-term operation and maintenance plans.)

[0030] The technical effects of the method and algorithm system for determining the timing of photovoltaic module cleaning according to the present invention include: First, the technical solution of this invention constructs a four-dimensional parameter system encompassing macroscopic and microscopic, physical and optical properties, and utilizes a specific nonlinear logarithmic model based on physical mechanisms for calculation. The accuracy and reliability of its evaluation results far surpass existing technologies relying on single indicators or general black-box models. Second, the "white-box" nature of this model provides excellent physical interpretability. When prediction results deviate, the error can be traced back to the specific physical parameter measurement stage, facilitating model maintenance and optimization and enhancing system robustness. Finally, this method directly links the precise physical loss rate with the economic cost model, ensuring that the final clean energy decision is no longer based on experience or rough estimation, but on scientific cost-benefit analysis. This effectively balances clean energy costs and power generation revenue, helping photovoltaic power plants minimize the cost per kilowatt-hour and maximize economic benefits throughout their entire lifecycle. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram illustrating the method for collecting quantitative parameters related to dust accumulation on photovoltaic power station modules in this embodiment of the invention. Figure 2 These are dust-covered images of the photovoltaic power plant surface obtained by using a drone to take high-resolution aerial photos in this embodiment of the invention. Figure 3 This is a schematic diagram of a dedicated measuring tool for measuring component contamination rate (SR) in an embodiment of the present invention; Figure 4 This is the calculation method for the optimal time for cleaning and dust covering in this invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0034] This embodiment provides a method and algorithm system for determining the timing of photovoltaic module cleaning. This embodiment uses a distributed photovoltaic power station located in Shenzhen (e.g., Figure 1 , Figure 2 As shown in the figure, the power station has deployed the algorithm system of the present invention in an application scenario to realize automated and precise decision-making on the timing of cleaning photovoltaic modules.

[0035] The method first performs the S1 parameter acquisition step to obtain the four-dimensional dust-covered physical parameters of the photovoltaic module surface. This step is performed by the parameter acquisition module in the algorithm system.

[0036] In a specific execution cycle, the parameter acquisition module, through its integrated drone control unit, dispatches a drone equipped with a high-resolution optical camera, laser thickness gauge, and laser scattering sensor to inspect the power plant. The drone flies along a preset route, scanning each photovoltaic module within the power plant and collecting the raw data needed to generate the physical parameters of the dust accumulation. This raw data includes high-resolution images of the module surfaces, as well as point cloud data returned by the laser sensor or directly measured thickness and particle size data. After collection, the raw data is transmitted in real-time via a 5G communication network to the algorithm system deployed on a local server.

[0037] After receiving the raw data, the parameter acquisition module performs parallel processing to generate four key parameters, specifically including: (1) Average particle size (de) and dust thickness (d) The processing unit within the module directly analyzes the data returned by the laser thickness gauge and the laser scattering sensor. By averaging the data from multiple measurement points on a single component, it obtains the average dust particle size de and dust thickness d of the component.

[0038] (2) Component contamination rate (SR) The module uses a dedicated contamination rate measurement tool (such as...) Figure 3 (As shown) to obtain the SR value. This tool utilizes optical closed-loop measurement of blue light pollutants (OMBP) technology to output an accurate component contamination rate (SR). In some embodiments, the SR value can also be obtained by performing AI image segmentation on high-resolution images captured by a drone to accurately identify and calculate the area ratio of dust-covered regions. In this embodiment, the tool measures the contamination rate (SR) of a typical component to be 68%.

[0039] (3) Visual color variation coefficient of dust cover (k) In selecting model parameters, this invention is based on in-depth research into the mechanism of dust accumulation. The average particle size (de) of the dust accumulation is chosen as the key parameter because particles of different sizes follow different scattering mechanisms. The dust accumulation visual color variation coefficient (k) is introduced because the chemical composition of dust varies greatly under different environments, and these components have distinct optical properties that cannot be characterized by simple physical parameters. Therefore, the introduction of the k value is precisely to quantify these complex optical properties determined by chemical composition. The dust accumulation visual color variation coefficient (k) is the key parameter of this invention and is calculated by the color coefficient calculation submodule in the parameter acquisition module. This submodule contains a pre-trained machine learning model. This model is pre-obtained as follows: First, hundreds of component images containing different types and colors of dust (such as desert sand, urban dust, industrial dust, and carbonaceous dust) are collected to build a training dataset. Subsequently, experienced technicians manually label each image in the dataset, assigning a k value between 0 and 1 based on the color and known components of the dust accumulation. The calibration rules are as follows: a completely clean component surface corresponds to k=1, representing minimal light attenuation; while dust covering the surface, matching the color of the exposed ground at the component installation location, corresponds to k=0, representing maximum light attenuation. Colors in between, such as yellow sand, are calibrated to 0.7; dark gray industrial dust to 0.3; and pure black carbon deposits to 0.1. These labeled sample pairs are used to train a neural network model constrained by physical boundaries, resulting in the trained model. In this execution, the submodule inputs dust-covered images captured by a drone into the model, calculating a k value of 0.9 corresponding to the current dust cover state.

[0040] In addition, to ensure long-term operational accuracy, the parameter acquisition module also periodically receives data from a high-precision ground dust monitoring system (such as...) deployed at fixed locations within the power plant via its data calibration interface. Figure 3 The system uses reference data (as shown) to calibrate the analysis model processing UAV data online or retrain it offline, thereby continuously correcting system errors. In actual photovoltaic power plants, dust deposition is affected by various factors such as wind, rain, and module tilt angle, resulting in a highly non-uniform distribution. Under natural conditions, dust accumulates heavily along the lower frame of the photovoltaic module, while the dust in the central area is relatively sparse. Traditional dust assessment methods, such as measuring the mass of dust per unit area (deposition density), often fail to distinguish this non-uniform distribution or include dust accumulation on the frame in the total mass, leading to significant deviations between the assessment results and the actual optical shading effect and power loss. To overcome the limitations of traditional single-index assessment methods, this invention proposes a novel multi-parameter, comprehensive assessment approach.

[0041] After acquiring the parameters, the method proceeds to the S2 loss rate calculation step. This step is executed by the loss rate calculation module in the algorithm system. This module acquires the following four key parameters: the component contamination rate SR measured using a specialized tool; the dust thickness d measured using a laser sensor; the average dust particle size de obtained through laser particle size analyzer analysis; and the dust visual color variation coefficient k calculated by analyzing the dust image on the component surface and inputting it into a pre-trained machine learning model. The loss rate calculation module receives these parameters: de = 12.6 μm, d = 410 μm, SR = 0.68, k = 0.9. Subsequently, the module calculates the loss rate based on a pre-set nonlinear logarithmic model.

[0042] The calculation is performed, where k is the base of the logarithmic model, and the value of k is limited to (0,1). To ensure the numerical stability of the calculation, the model adopts a change-of-base formula.

[0043] Finally, the method in this embodiment enters the S3 decision-making step, executed by the decision-making module in the algorithm system. This module takes the calculated power loss rate PL = 11.3% as input, and combines it with the current grid-connected electricity price of the power plant (e.g., 0.4 yuan / kWh) retrieved from the database and the estimated cost of a single cleaning cycle (e.g., 3000 yuan) to perform a cost-benefit analysis. The module first calculates the daily power generation loss caused by the 11.3% power loss and multiplies it by the electricity price to obtain the daily economic cost of power generation loss. Subsequently, the module predicts the cumulative cost of this loss over time and calculates the point in time when the cumulative loss first exceeds the cleaning cost (3000 yuan). This point in time is determined as the optimal cleaning time for this batch of components. The decision-making module presents this decision recommendation, along with the PL values ​​of each component, through a visualized heatmap interface (e.g., ...). Figure 2 The aerial photograph (which can be further processed into a heat map) is presented to the power plant operation and maintenance personnel. In the map, components with high PL values ​​are marked in red, indicating that they need to be cleaned first; those with medium PL values ​​are marked in yellow; and those with low PL values ​​are marked in green.

[0044] Meanwhile, the cleaning interval calculation module in the algorithm system also operates continuously in the background. This module periodically receives and stores the time-series data of SR and d generated by the parameter acquisition module, and establishes a fitting function describing the rate of pollutant accumulation based on this data. Using this function, the module can predict the length of time required from a clean state (SR=0, d=0) until the aforementioned economical cleaning threshold is reached. In the specific environment of this embodiment, the calculated optimal cleaning interval is 27 days. This calculation result provides a scientific basis for the power plant to formulate medium- and long-term, forward-looking operation and maintenance plans.

[0045] This invention decomposes the complex problem of dust accumulation into several precisely measurable physical parameters, quantifies them through a mathematical model with clear physical meaning, and ultimately outputs operation and maintenance decisions with direct economic guidance. This not only significantly improves the scientific rigor and accuracy of clean energy decisions, but also maximizes the economic benefits of photovoltaic power plants by balancing clean energy costs and power generation revenue.

[0046] Furthermore, in the above embodiments, the specific execution method of the S1 parameter acquisition step is described in further detail.

[0047] Reference Figure 2 In this embodiment, the parameter acquisition step (S1) employs a drone equipped with image acquisition equipment, a laser thickness sensor, and a laser scattering sensor. The drone flies along a preset flight path generated based on the photovoltaic power plant layout information to collect raw data for generating the dust-covered physical parameters. The technical objective of this automated inspection scheme is to efficiently and comprehensively acquire the dust-covered status of all components in a large photovoltaic power plant. In specific implementation, the system first automatically plans an optimal flight path that covers all photovoltaic module arrays based on the power plant's CAD layout diagram or satellite map. The drone flies along this path, and its onboard high-resolution optical camera captures high-definition images of the module surface at a preset frame rate and angle, such as... Figure 1 As shown in the diagram. Simultaneously, laser thickness sensors and laser scattering sensors emit laser beams towards the component surface, and by analyzing the time difference and scattering characteristics of the reflected signals, real-time data on dust thickness and particle size distribution along the flight path are obtained. This integrated multi-sensor approach ensures that the raw data required to calculate all four dust clogging physical parameters can be collected synchronously in a single flight mission, improving data acquisition efficiency and guaranteeing high consistency across different parameters over time.

[0048] After acquiring the raw data, the system executes a method for determining the visual color variation coefficient (k) of the dust cover. This method is one of the core technologies of this invention, and its technical purpose is to transform the complex physical characteristic of dust cover color into an objective quantitative parameter that can be used in a mathematical model.

[0049] Specifically, before system deployment, a training dataset containing multiple sets of "dust-covered image-k-value label" sample pairs was first established. For this purpose, hundreds of representative dust-covered images were collected from photovoltaic power plants in different regions and seasons, covering various types from light-colored desert sand to dark industrial carbon black. Then, technicians with backgrounds in materials science and optics manually categorized the color types of the dust-covered images to obtain k-value labels. The labeling rule was: the surface color of clean photovoltaic modules was labeled as 1, and the color of the exposed ground in the photovoltaic module installation area was labeled as 0. Other colors of dust were numerically assigned based on their difference from these two baseline colors. The training dataset was used to train a neural network model constrained by physical boundaries, resulting in a trained model capable of taking a dust-covered image as input and outputting the corresponding k-value. This model was embedded in the color coefficient calculation submodule of the algorithm system.

[0050] Therefore, in the parameter acquisition step (S1), after the system receives the dust-covered image captured by the drone, it first preprocesses the image, and then inputs the acquired dust-covered image of the photovoltaic module into the trained model to obtain the visual color variation coefficient (k) of the dust cover. For example, for an image showing light yellow sand cover, the model outputs a k value of 0.85; while for another image showing dark gray oily soot, the model outputs a k value of 0.25. This process realizes the automated conversion from qualitative images to quantitative parameters.

[0051] To improve the computational efficiency and stability of the aforementioned machine learning model, in this embodiment, the input to the trained model is the color features extracted from the dust-covered image, rather than the complete original image. Specifically, before inputting the image into the model, the system first converts it to the HSV (Hue, Saturation, Lightness) color space. Then, it calculates the average color values ​​of the H, S, and V components of all pixels within the entire image region, and uses these three average values ​​to form a three-dimensional feature vector. Alternatively, it can calculate the color histogram of the image in the HSV space and use the histogram distribution data as a higher-dimensional feature vector. Using this low-dimensional color feature as the model input aims to significantly reduce the computational load and enhance the model's robustness to non-core factors such as uneven lighting and shadows in the image. The color feature is the average color value or color histogram of the dust-covered image in the RGB, HSV, or Lab color space. In this way, the system can process massive amounts of UAV inspection images faster and more stably, and output reliable k values.

[0052] Furthermore, in the above embodiments, the specific execution methods of the S2 loss rate calculation step and the S3 decision step are further described in detail.

[0053] In the S2 loss rate calculation step, the algorithm system's loss rate calculation module receives four parameters generated by the parameter acquisition module: average dust particle size de, average dust thickness d, contaminated coverage area percentage SR, and dust color coefficient k. To ensure the numerical stability and engineering feasibility of the calculation, the nonlinear logarithmic model is expressed using a changed-base formula during calculation:

[0054] Where ln is the natural logarithm. The technical purpose of using this form is to avoid the potential numerical instability or undefined problems that may arise from directly calculating the logarithm to the base k when k is equal to or close to 1. The loss rate calculation module substitutes the above parameter values ​​into this base-changing formula, and the calculation process is as follows:

[0055] The final calculated power loss rate (PL) of the photovoltaic module is 11.3%. This calculation result accurately quantifies the negative impact of the current dust cover on power generation performance.

[0056] After obtaining the accurate PL value, the system proceeds to the S3 decision-making step. In this embodiment, the decision-making step (S3) further includes a specific decision-making algorithm based on economic benefit optimization. This algorithm aims to transform the physical power loss rate into an economic decision with direct commercial guidance. The algorithm executes according to the following sub-steps: (1) Obtain the grid connection price or trading price of the photovoltaic power station, and calculate the economic cost of power generation loss per unit time based on the current power loss rate (PL). The system retrieves the grid connection price of the power station from the power trading center. At the same time, based on the rated power of the photovoltaic module and the local average effective sunshine hours, calculate the theoretical daily power generation under dust-free conditions. Then, multiply the theoretical daily power generation by the power loss rate PL caused by dust to obtain the daily power generation loss caused by dust. Finally, multiply the power generation loss by the daily electricity price to obtain the daily economic loss of power generation.

[0057] (2) Obtain the estimated cost of a single cleaning operation as the cleaning cost. The system queries the local labor market for unit man-hours and machine shifts, and estimates the total cost of a single full-station cleaning operation based on the power plant scale and the system's average labor and machine consumption quotas for the entire industry.

[0058] (3) Establish a prediction function for the cumulative economic cost of power generation loss over time. Based on historical data analysis, the system assumes that the dust accumulation rate is approximately linear in the short term, and thus establishes a prediction function describing the linear growth of the economic cost of power generation loss over time.

[0059] (4) Using the prediction function, calculate and determine the point in time when the cumulative value of the economic cost of power generation loss first equals or exceeds the cleaning cost, and take this point in time as the optimal cleaning time. The system uses the above function to extrapolate from the current time point to the future until the predicted cumulative power generation loss amount equals the cleaning cost. The specific date predicted is determined by the system as the optimal time to execute this cleaning operation, and a corresponding operation and maintenance work order is generated.

[0060] In a more preferred embodiment, the above-mentioned calculation method for economic decisions can achieve a higher degree of dynamism and accuracy, and its detailed logical flow is as follows: Figure 4 As shown. In this preferred embodiment: (a) The calculation of the "economic cost of power generation loss" is not based on a fixed feed-in tariff, but rather on real-time acquisition of the "clearing price of photovoltaic power plants in the electricity spot market" and dynamic calculation combined with power generation data obtained from the "inverter side" or "grid connection point," thus obtaining a more accurate "loss of grid-connected power generation revenue due to dust accumulation." (b) The "cleaning cost" is also not a fixed estimated value, but is dynamically calculated by multiplying the "dynamic unit price of labor and machinery shifts" by a preset "consumption quota" to reflect the fluctuations in "maintenance costs for cleaning component surfaces" in real time. By dynamically calculating both revenue loss and maintenance costs, the final determined "actual optimal cleaning time" can more closely align with market changes and actual costs, thereby achieving a higher level of economic efficiency optimization.

[0061] Through the precise calculations based on the economic model described above, this embodiment can determine a cleaning time that maximizes the overall economic benefits of the power plant, avoiding economic losses caused by cleaning too early or too late.

[0062] It should be noted that the calculation of the power loss rate (PL) and its economic evaluation described above are mainly applicable to photovoltaic power plants operating at full load without power curtailment. In actual operation, photovoltaic power plants may face curtailment scenarios due to unsuccessful bids in power trading or power optimization adjustments by dispatching agencies. In such cases, the actual power generation of the modules is not entirely determined by the photovoltaic efficiency caused by dust accumulation. To address these special scenarios, the method of this invention provides two correction strategies based on this core formula: one is to use calculus principles to split and then merge the power generation data to eliminate the power generation loss caused by curtailment; the other is to add a correction coefficient with certain rules based on historical curtailment data and power plant operating experience. These strategies ensure that this method can still perform accurate power loss assessment and economic analysis under a wider range of power plant operating conditions.

[0063] Furthermore, as a supplement and safety guarantee to the aforementioned economic decision-making model, the decision-making step (S3) also includes an early warning mechanism. The technical purpose of this mechanism is to provide an immediate response to severe contamination situations that could lead to component safety issues. The system internally presets a fixed threshold for the power loss rate, for example, 15%. In step S2, when the current power loss rate (PL) exceeds this preset threshold, the system immediately triggers a cleaning warning. This warning information is pushed to the operation and maintenance management personnel via SMS, email, or system message. Simultaneously, the system also marks the severely contaminated photovoltaic modules with a bright color (e.g., red) on the heat map of the power plant monitoring interface. This guides operation and maintenance personnel to conduct emergency inspections and treatments in specific areas to prevent "hot spot effects" caused by severe localized dust accumulation (such as bird droppings or thick sludge), thereby avoiding irreversible physical damage to the photovoltaic modules and ensuring the long-term safe and stable operation of the power plant assets.

[0064] Furthermore, in a preferred embodiment, the method of this embodiment further includes a step of determining the optimal cleaning time interval. This step aims to expand from a single optimal timing decision to the periodic optimization of the cleaning strategy, thereby providing data support for the development of medium- and long-term operation and maintenance plans for the power plant. The specific implementation process of this step is as follows: (1) Step S1 is executed periodically to obtain time series data of the component contamination rate (SR) and the dust cover thickness (d). The algorithm system executes the parameter acquisition step at a fixed time frequency (such as daily or weekly) and records the SR value and d value obtained each time, along with the timestamp, in the database to form a historical data sequence describing the evolution of these two key parameters over time.

[0065] (2) Based on the time series data, establish a fitting function to describe the changes of the SR value and d value over time. The algorithm system calls the built-in mathematical fitting tool to analyze the above time series data in order to find the Sigmoid function model that best describes the dust accumulation law, i.e., the S-curve.

[0066] (3) Using the fitting function, predict and calculate the time required from the start of the clean state until the cumulative economic cost of the power generation loss equals the cleaning cost, and determine this time length as the optimal cleaning time interval for the photovoltaic module. The system substitutes SR(t) and d(t) into the core power loss model.

[0067] (Here, it is assumed that de and k are constants in the short term), thus obtaining a function describing the change of PL over time. Subsequently, the system combines the electricity price and the rated power of the components to convert PL(t) into a function Cost_loss(t) describing the change of the economic cost of daily power generation loss over time. By integrating Cost_loss(t) from 0 to T, the cumulative loss cost function Cost_total(T) is obtained. Finally, by solving the equation Cost_total(T) = Cleaning_Cost (where Cleaning_Cost is the cost of a single cleaning), the time length T can be obtained. This value of T is the optimal cleaning interval calculated from an economic perspective under this environment. To ensure the accuracy of all the above calculations and model analyses, in a preferred embodiment, the parameter acquisition step (S1) also includes a model correction step. The technical purpose of this step is to establish a closed-loop accuracy assurance system to ensure the continuous improvement and maintenance of the accuracy of the UAV data analysis model, because the measurement accuracy of the UAV's onboard sensors may be affected by various factors such as flight attitude, changes in ambient light, and sensor drift.

[0068] The specific implementation steps for model correction are as follows: A fixed ground dust monitoring system is deployed at a typical location in the photovoltaic power station, such as... Figure 3 As shown, the ground system employs high-precision sensors to continuously and stably collect dust physical parameters at its location, serving as high-confidence reference data or "true values." This reference data is used to correct or train the analytical model processing the raw data collected by the UAV, thereby improving the accuracy of the dust physical parameters calculation. Specifically, when the UAV flies over the area where the ground monitoring system is located, the system simultaneously records the parameter values ​​analyzed by the UAV and the parameter values ​​measured by the ground system. By comparing these two sets of data, the system error can be calculated, and a correction coefficient can be generated for real-time calibration of subsequent UAV-collected data. Furthermore, the accumulated reference data can be periodically used for incremental training or retraining of the machine learning model for the color coefficient k value, thereby continuously optimizing the model's performance and ensuring high reliability and accuracy of the entire data acquisition and analysis system during long-term operation.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining the timing of cleaning photovoltaic modules, characterized in that, Includes the following steps: S1. Parameter acquisition step: Obtain the physical parameters of the dust covering the surface of the photovoltaic module, wherein the physical parameters of the dust covering include at least: Average particle size (de) of dust cover obtained from sensor data. Dust thickness (d) obtained from sensor data; Component contamination rate (SR) obtained through a dedicated contamination rate measurement tool or image segmentation technique; The dust visual color variation coefficient (k) is obtained by analyzing the dust-covered images of the photovoltaic modules using a machine learning model. S2. Loss Rate Calculation Steps: Based on a preset nonlinear logarithmic model, the current power loss rate (PL) of the photovoltaic module is calculated using the dust-covered physical parameters. The mathematical expression of the nonlinear logarithmic model is: Wherein, k is the base of the logarithmic model, and the value range of k is limited to (0, 1); S3. Decision-making steps: Convert the current power loss rate (PL) into the economic cost of power generation loss caused by dust accumulation, and determine the optimal time to clean the photovoltaic module based on the comparison between the preset cleaning cost and the economic cost of power generation loss.

2. The method for determining the timing of photovoltaic module cleaning according to claim 1, characterized in that, In the parameter acquisition step (S1), an unmanned aerial vehicle (UAV) equipped with an image acquisition device, a laser thickness sensor, and a laser scattering sensor is used to fly along a preset route generated based on the layout information of the photovoltaic power station in order to collect and generate the raw data of the dust physical parameters.

3. The method for determining the timing of photovoltaic module cleaning according to any one of claims 1 or 2, characterized in that, The method for determining the visual color variation coefficient (k) of the dust cover includes: A training dataset containing multiple sets of "dust-covered image-k-value label" sample pairs is established. The k-value label is obtained by manually calibrating the color type of the dust-covered images. The calibration rule is as follows: the surface color of the clean photovoltaic module is calibrated as 1, the exposed ground color of the photovoltaic module installation area is calibrated as 0, and other colors of dust are assigned a score between 0 and 1 based on their visual difference from the two reference colors. The training dataset is used to train a neural network model, decision tree model, or support vector machine model constrained by physical boundaries to obtain a trained model that can take a dust-covered image as input and output the corresponding k value. In the parameter acquisition step (S1), the collected dust-covered image of the photovoltaic module is input into the trained model to obtain the dust-covered visual color variation coefficient (k).

4. The method for determining the timing of photovoltaic module cleaning according to claim 3, characterized in that, The input to the trained model is the color features extracted from the dusty image, where the color features are the average color value or color histogram of the dusty image in the RGB, HSV, or Lab color space.

5. The method for determining the timing of photovoltaic module cleaning according to claim 1, characterized in that, The nonlinear logarithmic model is expressed using a change-base formula during computation: Where ln is the natural logarithm.

6. The method for determining the timing of photovoltaic module cleaning according to claim 1, characterized in that, The decision-making step (S3) further includes: Obtain the grid connection price or trading price of the photovoltaic power plant, and calculate the economic cost of power generation loss per unit time in combination with the current power loss rate (PL); The estimated cost per cleaning operation is used as the cleaning cost. Establish a prediction function for the cumulative economic cost of the power generation loss over time; The prediction function is used to calculate and determine the point in time when the cumulative value of the economic cost of power generation loss first equals or exceeds the cleaning cost, and this point in time is taken as the optimal cleaning time.

7. The method for determining the timing of photovoltaic module cleaning according to claim 1, characterized in that, The decision-making step (S3) also includes an early warning mechanism: when the current power loss rate (PL) exceeds a fixed threshold preset based on historical data and operation and maintenance experience, the system triggers a cleaning early warning and marks the photovoltaic module with a bright color on the power plant heat map.

8. The method for determining the timing of photovoltaic module cleaning according to claim 1, characterized in that, The method also includes the step of determining the optimal cleaning time interval: Step S1 is performed periodically to obtain time-series data of the component contamination rate (SR) and the dust cover thickness (d); Based on the time series data, a fitting function is established to describe the changes of the SR value and d value over time; Using the fitting function, the time required from the start of the clean state until the cumulative value of the economic cost of power generation loss equals the cleaning cost is predicted and calculated, and this time length is determined as the optimal cleaning time interval for the photovoltaic module.

9. The method for determining the timing of photovoltaic module cleaning according to claim 2, characterized in that, The parameter acquisition step (S1) also includes a model correction step: In typical locations of photovoltaic power plants, a fixed ground dust monitoring system is deployed to continuously collect high-precision dust physical parameters as reference data. The reference data is used to correct or train the analytical model used to process the raw data collected by the UAV, so as to improve the accuracy of the calculation of the dust physical parameters.

10. An algorithm system for determining the timing of cleaning photovoltaic modules, characterized in that, include: A parameter acquisition module is configured to receive or process raw data acquired through a sensor network to generate physical parameters of dust covering the surface of a photovoltaic module. These physical parameters include at least: average dust particle size (de), dust thickness (d), module contamination rate (SR), and dust visual color variation coefficient (k). The parameter acquisition module further includes: The drone control unit is used to control a drone equipped with an image acquisition device and a laser sensor to collect raw data of the dust physical parameters according to a preset route, and to transmit the raw data through a 5G communication network or a local data interface. The color coefficient calculation submodule contains a machine learning model trained using a "dust-covered image - manually calibrated k value" sample pair, which is used to calculate the dust visual color variation coefficient (k) based on the dust-covered image collected by the UAV. The data calibration interface is used to receive reference data from a fixed ground dust monitoring system deployed at the power plant site, so as to perform online calibration or offline training on the analysis model that processes data collected by UAVs. The loss rate calculation module is functionally connected to the parameter acquisition module and is configured to calculate the loss rate based on a preset mathematical expression. A nonlinear logarithmic model is used to calculate the current power loss rate (PL) of the photovoltaic module using the dust-covered physical parameters, where k is the base of the logarithmic model; The decision module is functionally connected to the loss rate calculation module and is configured to convert the current power loss rate (PL) into the economic cost of power generation loss, and generate decision suggestions for determining the optimal time to clean the photovoltaic module based on the comparison result between the preset cleaning cost and the economic cost of power generation loss. The cleaning interval calculation module is functionally connected to the parameter acquisition module and is configured to receive and store time series data of the component contamination rate (SR) and the dust thickness (d), and establish a fitting function based on the time series data to calculate the optimal cleaning interval of the photovoltaic module.

Images