Method for predicting meteorological soiling and cleaning needs of a photovoltaic module based on a numerical model

Abstract

The application discloses a method for predicting meteorological dust accumulation cleaning requirements of a photovoltaic module based on a numerical mode, selects a SSP-RCP scenario and future simulation data of multiple global climate modes and historical precipitation data from a CMIP6 data set; for a target region, the future and historical precipitation of each grid point is calculated; then, the annual effective deposition of each grid point is calculated, the core of which is to correct the conversion of vertical deposition to the inclined surface of the photovoltaic module through the effective deposition coefficient, and to quantify the wind loss through the wind erosion correction; finally, based on the annual effective deposition, the theoretical annual cleaning frequency of the photovoltaic module of each grid point is calculated, thereby providing a quantitative basis for differentiated operation and maintenance.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance technology for photovoltaic power generation systems, and specifically to a method for predicting the demand for meteorological dust cleaning of photovoltaic modules based on numerical models. Background Technology

[0002] As a clean energy source, photovoltaic power generation's efficiency directly impacts the economic returns of power plants. However, photovoltaic modules, exposed to the natural environment for extended periods, inevitably accumulate dust, sand, pollen, and industrial pollutants on their surfaces. This "dust accumulation" phenomenon leads to the following problems:

[0003] Significantly reduced power generation efficiency: Dust blocks light, leading to a decrease in module transmittance. Studies have shown that in heavily polluted or arid regions, uncleaned modules can result in an annual power generation loss of 6%-25%, and a single instance of severe dust accumulation can even lead to an instantaneous power loss of over 30%.

[0004] It can cause hot spot effect: uneven dust accumulation can cause excessively high local temperature of the module, accelerate the aging of the cells and encapsulation materials, and even cause fire hazards and shorten the service life of the module.

[0005] Given the above issues, the decision regarding the cleaning requirements for photovoltaic modules becomes particularly important.

[0006] Currently, the industry's commonly used methods for making decisions about cleaning needs can be mainly categorized as follows:

[0007] 1) Manual or mechanical cleaning at fixed time intervals is not adaptable to changing climate and pollution conditions, and is prone to over-cleaning or under-cleaning, resulting in low overall cost-effectiveness.

[0008] 2) Decisions are made based on limited meteorological experience or rough observations of power generation; experience is difficult to quantify and is highly regional, making it impossible to form universal and accurate guidance, and there is a lack of accurate assessment of the cleaning effect of rainfall and the accumulation rate of dust carried by light winds.

[0009] 3) By deploying dust sensors or analyzing the output power of components in real time, cleaning can be triggered when the efficiency drops to a threshold. This requires additional hardware investment and involves installation, calibration, and maintenance issues. Furthermore, single-point measurements are difficult to represent the overall situation and cleaning can only be triggered after power generation loss has already occurred, making preventative maintenance impossible as losses have already taken place.

[0010] In summary, most existing technologies rely on empirical values ​​or dust accumulation status to predict cleaning needs. However, dust accumulation is a dynamic physical-meteorological process, influencing various meteorological factors such as wind speed, wind direction, humidity, precipitation, and particulate matter concentration. Therefore, they cannot effectively predict future cleaning needs. Consequently, there is an urgent need for a method that can proactively, accurately, and efficiently predict the dust accumulation and cleaning needs of photovoltaic modules. Summary of the Invention

[0011] To address the aforementioned technical issues, this invention proposes a method for predicting the meteorological dust accumulation cleaning requirements of photovoltaic modules based on numerical models. This method utilizes meteorological numerical models to predict the future meteorological dust accumulation cleaning requirements of photovoltaic modules in target areas under different combined scenarios.

[0012] This invention discloses a method for predicting the demand for meteorological dust cleaning of photovoltaic modules based on numerical models, comprising the following steps:

[0013] Step S1: From the multi-model dataset of the Sixth Coupling Model Intercomparison Project, determine at least one shared socio-economic path-typical concentration path (SSP-RCP) combined scenario; for each determined combined scenario, select future climate scenario simulation data from at least two different global climate models and historical monthly precipitation from the corresponding historical simulation experiments of the models. The future climate scenario simulation data shall include at least monthly dry aerosol deposition flux, monthly wet deposition flux, and monthly precipitation.

[0014] Step S2: Select the target area. For each combined scenario, calculate the multi-model average annual dry deposition flux, annual wet deposition flux, and multi-model average annual precipitation for each grid point in the target area. Next, calculate the historical multi-model average annual precipitation for each grid point in the same target area based on the historical monthly precipitation obtained in Step S1.

[0015] Step S3: Based on the historical multi-model average annual precipitation and the annual precipitation classification threshold for each grid point, the target area is divided into arid and semi-arid regions and semi-humid and humid regions. The annual isohyet classification threshold refers to the 400 mm annual precipitation classification threshold used in Chinese geography and climatology to distinguish between semi-humid and semi-arid climate zones. Among them, the historical multi-model average annual precipitation <400 mm is arid and semi-arid region, and the historical multi-model average annual precipitation ≥400 mm is semi-humid and humid region.

[0016] Step S4: Calculate the annual effective sedimentation amount corresponding to each grid point in the target region under each combined scenario. :

[0017] ,in, For each grid point, the annual total deposition is defined as follows: α is the effective deposition coefficient, which is a real number greater than 0 and less than 1. Its value is used to correct the coupling parameter of atmospheric particulate matter from vertical deposition flux to the actual deposition amount on the tilted surface of the photovoltaic module; β is the wind erosion correction coefficient, which is used to correct the loss of dust already deposited on the photovoltaic module surface due to secondary wind erosion caused by the environmental wind field. Its value is determined according to the climate type of the target area: the wind erosion correction coefficient for arid and semi-arid areas is less than that for semi-humid and humid areas.

[0018] Step S5: Annual effective sedimentation rate based on the result obtained in step S4 Calculate the theoretical annual number of photovoltaic module cleaning operations for each grid point in the target area under each combined scenario. :

[0019] ,

[0020] The theoretical number of times the photovoltaic module is cleaned per year for each grid point, where T represents the photovoltaic module cleaning threshold. T is a positive number greater than 0, and the unit is g / m².

[0021] Furthermore, after obtaining the theoretical annual cleaning frequency of the photovoltaic modules, considering the environmental uncertainties caused by rainfall, the theoretical annual cleaning frequency of the photovoltaic modules obtained in step S5 is corrected for rainfall to obtain the actual annual cleaning frequency of the photovoltaic modules, specifically:

[0022] In arid and semi-arid regions, the actual annual cleaning frequency of photovoltaic modules is calculated using the effective rainfall monthly count method. : ,in The month with effective rainfall;

[0023] In semi-humid and humid regions, the actual annual cleaning frequency of photovoltaic modules is calculated using the rainfall reduction factor method. : ,in This is the reduction factor;

[0024] Transition zones are defined based on annual precipitation thresholds. A boundary fusion mechanism is used for these transition zones, with weighted fusion applied to the boundary areas.

[0025] ,

[0026] in, This refers to the actual number of cleaning cycles performed in the transition zone. It is a weighted fusion coefficient with a value between 0 and 1.

[0027] Furthermore, the reduction factor Determined based on piecewise linear interpolation of annual rainfall:

[0028] ,

[0029] Where p is the multi-model average annual precipitation.

[0030] Furthermore, the actual number of times the photovoltaic modules were cleaned, N, after rainfall correction. actual The equivalent deposition amount is converted, and then the square root mapping is applied to obtain the corrected score, as follows:

[0031] ,corrected=N actual ×T,

[0032] Score_corrected=sqrt( ,corrected / T ref )×100,

[0033] Where Score_corrected is the cleaning demand intensity score, with a value of [0, 100]; N actual The corresponding values ​​are based on the calculation results for different regions. , , T ref For reference sedimentation amount, sqrt() represents the square root operation.

[0034] Furthermore, based on the annual effective sedimentation rate obtained in step S4 The intensity score of cleaning demand is calculated using square root mapping. A higher score indicates a stronger cleaning demand, specifically expressed as follows:

[0035] Score = sqrt( / T ref )×100,

[0036] Wherein, Score represents the intensity of cleaning demand, with a value of [0, 100]. T represents the annual effective sedimentation volume. ref For reference sedimentation, we take 100 g / m², which corresponds to an extremely high sedimentation standard. sqrt() represents the square root operation.

[0037] Furthermore, the actual number of times the photovoltaic modules were cleaned, N, after rainfall correction. actual The equivalent deposition amount is converted, and then the square root mapping is applied to obtain the corrected score, as follows:

[0038] ,corrected=N actual ×T,

[0039] Score_corrected=sqrt( ,corrected / T ref )×100,

[0040] Among them, T ref For reference sedimentation, we take 100 g / m², which corresponds to an extremely high sedimentation standard. sqrt() represents the square root operation.

[0041] Furthermore, the shared socioeconomic path-typical concentration path combination scenario described in step S1 includes SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5.

[0042] Furthermore, in step S1, eight global climate models were selected that have performance validation literature support in the simulation of aerosol-climate interactions.

[0043] Furthermore, in step S4, the effective deposition coefficient α is determined by the geometric projection effect of the tilted installation of the photovoltaic module, the trapping and enhancement effect of the surface micro-roughness, and the deposition loss effect caused by the flow around the frame; the effective deposition coefficient is calculated through the following coupling relationship:

[0044] α = cosθ × η_r × η_e,

[0045] Where θ is the installation tilt angle of the photovoltaic module; η_r is the surface micro-roughness trapping enhancement coefficient; and η_e is the deposition loss coefficient caused by flow around the frame.

[0046] Furthermore, the data extraction process in step S2 is as follows:

[0047] Step S21: Based on the latitude and longitude boundaries of the target area, extract a subset of grid data that completely covers the region from the high-resolution climate model grid data at the global scale.

[0048] Step S22: Convert the extracted raw time series data of each grid point to a monthly time scale. Specifically, convert the raw flux unit to kg·m -2 ·s -1 Multiply by the number of months and seconds to convert to g·m -2 ·month -1 The unit for precipitation (pr) is kg·m³. -2 ·s -1 This refers to the mass flux per unit area per unit time; since the density of water is approximately 1000 kg / m³. 3 1kg·m -2 =1mm, therefore kg·m -2 ·s -1 In fact, it is equivalent to mm / s. To convert it to monthly precipitation, it needs to be multiplied by the number of seconds per month.

[0049] Step S23: For each grid point in the target area, under the same combined scenario, perform an equal-weighted ensemble average on the monthly scale data after each global climate model transformation, and output the multi-model average dry aerosol deposition flux, wet deposition flux and precipitation data for each grid point in the target area under each combined scenario; at the same time, perform an equal-weighted ensemble average on the historical precipitation data, and output the multi-model average historical precipitation data for each grid point in the target area.

[0050] Step S24: Calculate the corresponding multi-model average annual dry deposition flux, annual wet deposition flux, and annual precipitation based on the multi-model average monthly dry deposition flux, monthly wet deposition flux, and monthly precipitation data obtained in step S23; calculate the historical multi-model average annual precipitation data based on the multi-model historical average monthly precipitation data obtained in step S23.

[0051] The present invention also discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for predicting the meteorological dust accumulation cleaning requirements of photovoltaic modules based on numerical models.

[0052] The beneficial effect of this invention is that, compared with the prior art, it can predict the future demand for meteorological dust removal of photovoltaic modules in different combinations of scenarios through meteorological numerical models, providing an effective basis for subsequent dust removal of photovoltaic modules, as detailed below:

[0053] 1) This invention addresses the specific physical process of dust accumulation in photovoltaic modules, overcoming the shortcomings of existing fixed-period or simple threshold methods that rely on historical data and lack universality for different installation tilt angles, surface materials, and array layouts. By constructing a numerical model based on the principles of atmospheric physics, surface science, and fluid mechanics, it achieves mechanism-driven dust accumulation prediction that fits the physical characteristics of photovoltaic modules and their environmental response.

[0054] 2) This invention precisely quantifies the dynamic balance between dust deposition on photovoltaic glass covers, dust loss due to frame-side turbulent winds, and dust washing by rainwater by coupling an effective deposition coefficient, wind erosion correction coefficient, and a rainwater washing model specifically designed for photovoltaic modules. This model can directly extrapolate changes in the light transmittance of the module panel, thereby accurately predicting the trend of power generation efficiency degradation and the optimal cleaning window, providing a core basis for photovoltaic power plant power generation prediction and preventive maintenance decisions.

[0055] 3) This invention combines the actual annual cleaning frequency of photovoltaic modules with the cleaning demand intensity score to directly serve the goal of ensuring the highest power generation revenue of the modules with the lowest operation and maintenance cost, thereby completely solving the problem of over-cleaning or under-cleaning of photovoltaic modules caused by traditional methods. Attached Figure Description

[0056] Figure 1 This is a flowchart of the prediction method of the present invention; Detailed Implementation

[0057] The climate projection data used in this invention comes from the Sixth Coupled Model Intercomparison Project (CMIP6), specifically retrieved and obtained through the data portal of the Earth System Grid Consortium (ESGF). This portal is a distributed data network that allows users to locate the required dataset based on multiple search dimensions such as model name, experimental scenario, variables, and frequency, and can download files in batches using a generated wget script. To cover the uncertainties of future socioeconomic development and radiative forcing, this invention selects four representative shared socioeconomic pathway (SSP) and representative concentration pathway (RCP) combination scenarios from the CMIP6 scenario model intercomparison project.

[0058] like Figure 1 As shown, the present invention provides a method for predicting the meteorological dust accumulation cleaning demand of photovoltaic modules based on numerical models, comprising the following steps:

[0059] Step S1: From the multi-model dataset of the Sixth Coupling Model Intercomparison Project (CPIP), four shared socio-economic path-typical concentration path combination scenarios (SSP-RCP) are identified. Specifically, SSP1-2.6 represents the sustainable development path, corresponding to the low greenhouse gas emission scenario; SSP2-4.5 represents the "middle way" development model, which belongs to the medium emission scenario; SSP3-7.0 represents the regional competition path, which corresponds to the high emission scenario; and SSP5-8.5 represents the fossil fuel-driven rapid development path, which belongs to the extremely high emission scenario.

[0060] For each identified scenario combination, this invention selects future climate scenario simulation data from independent, literature-verified global climate models in the field of aerosol-climate interaction simulation. The selected data must include at least three variables crucial for ash deposition prediction: monthly dry aerosol deposition flux, monthly wet deposition flux, and monthly precipitation.

[0061] To obtain a highly reliable dataset and to adhere to the principle of optimal simulation performance of different climate variables in different models, this invention adopts a multi-model, multivariate coupled data acquisition strategy, with the specific sources as follows:

[0062] Data sources for aerosol deposition fluxes: The following eight global climate models, which have demonstrated excellent performance in simulating aerosol processes, were selected as the sources for obtaining dry and wet aerosol deposition fluxes:

[0063] CanESM5, CanESM5-1, CESM2-WACCM, GFDL-ESM4, INM-CM4-8, INM-CM5-0, MIROC6, MRI-ESM2-0.

[0064] Precipitation data sources: The following eight global climate models, which have a recognized advantage in precipitation simulation, were selected as the sources of precipitation data: BCC-CSM2-MR, CAS-ESM2-0, CMCC-ESM2, CMCC-CM2-SR5, GFDL-ESM4, INM-CM4-8, MRI-ESM2-0, and TaiESM1.

[0065] For the selected global climate models, corresponding historical monthly precipitation values ​​were selected from historical simulation experiments.

[0066] Step S2: Select the target area. For each combined scenario, calculate the multi-model average annual dry aerosol deposition flux, annual wet deposition flux, and multi-model average annual precipitation for each grid point within the target area. Next, based on the historical precipitation obtained in Step S1, calculate the historical multi-model average annual precipitation for each grid point within the same target area. The specific extraction process is as follows:

[0067] Step S21: Based on the latitude and longitude boundaries of the target area, extract a subset of grid data that fully covers the region from the high-resolution climate model grid data at the global scale. This focuses massive amounts of global data on the target area, significantly reducing the amount of data required for subsequent calculations and laying a spatial foundation for regional average calculations.

[0068] Step S22: Convert the extracted raw time series data of each grid point to a monthly time scale. Specifically, convert the raw flux unit to kg·m -2 ·s -1 Multiply by the number of months and seconds to convert to g·m -2 ·month -1 The unit for precipitation (pr) is kg·m³. -2 ·s -1 This refers to the mass flux per unit area per unit time; since the density of water is approximately 1000 kg / m³. 3 1kg·m -2 =1mm, therefore kg·m -2 ·s -1 In fact, it is equivalent to mm / s. To convert it into monthly precipitation, it needs to be multiplied by the number of seconds per month to eliminate the possible phase deviation between different climate models on the daily scale.

[0069] Step S23: For each grid point in the target area, under the same combined scenario, perform average calculation on the monthly scale data after each global climate model conversion, and output the multi-model average dry aerosol deposition flux, wet deposition flux and precipitation data of each grid point in the target area under each combined scenario; at the same time, perform average calculation on the historical precipitation data, and output the multi-model average historical precipitation data of each grid point in the target area.

[0070] Step S24: Calculate the corresponding multi-model average annual dry deposition flux, annual wet deposition flux, and annual precipitation based on the multi-model average monthly dry deposition flux, monthly wet deposition flux, and monthly precipitation data obtained in step S23; calculate the historical multi-model average annual precipitation data based on the multi-model historical average monthly precipitation data obtained in step S23.

[0071] Step S3: Divide the target area into arid and semi-arid regions and semi-humid and humid regions. Each grid point with a multi-model average historical precipitation of <400mm is classified as an arid and semi-arid region; each grid point with a multi-model average historical precipitation of ≥400mm is classified as a semi-humid and humid region.

[0072] Step S4: Calculate the annual effective sedimentation amount corresponding to each grid point in the target region under each combined scenario. :

[0073] ,

[0074] in, The annual total deposition at each grid point, and the effective deposition coefficient α, are jointly determined by the geometric projection effect of the tilted installation of the photovoltaic module, the trapping and enhancement effect of the surface micro-roughness, and the deposition loss effect caused by the flow around the frame; the effective deposition coefficient is calculated through the following coupling relationship:

[0075] α = cosθ × η_r × η_e,

[0076] Wherein, θ is the installation tilt angle of the photovoltaic module, ranging from 30° to 40°, with a corresponding cos(θ) value of 0.77 to 0.87; η_r is the surface micro-roughness capture enhancement coefficient, ranging from 1.10 to 1.20; η_e is the deposition loss coefficient caused by flow around the frame, ranging from 0.90 to 0.95; the effective deposition coefficient α ranges from 0.76 to 0.99; considering the diversity of actual photovoltaic power plants, the frame design of modules from different manufacturers, the difference in optimal tilt angle at different geographical latitudes, and the regional particulate matter characteristics with different particle size distributions, this invention takes α=0.70 as a conservative estimate;

[0077] A wind erosion correction factor β is applied to correct for the secondary wind erosion loss caused by the environmental wind field, which affects the dust already deposited on the photovoltaic module surface. Its value is determined according to the climate type of the target area.

[0078] In the arid and semi-arid regions of Northwest China: β=0.10; high wind speeds but low relative humidity, resulting in strong dust adhesion.

[0079] Eastern semi-humid and humid regions: β=0.15; high humidity makes dust condense easily but also makes it easier to be dispersed by the wind.

[0080] Step S5: Annual effective sedimentation rate based on the result obtained in step S4 Calculate the theoretical annual number of photovoltaic module cleaning operations for each grid point in the target area under each combined scenario. :

[0081] ,

[0082] The theoretical number of annual cleaning cycles for the photovoltaic module corresponding to each grid point, with T=10g / m² as the photovoltaic module cleaning threshold;

[0083] This threshold was determined based on the following scientific criteria:

[0084] Photovoltaic module power generation efficiency loss experimental data: In the arid climate of Bahawalpur region, Pakistan; after 6 weeks of exposure, the dust density was 10.254 g / m², resulting in a 25.42% decrease in output power, with a daily average dust accumulation rate of 0.244 g / m².

[0085] Studies have shown that the impact of dust accumulation on photovoltaic performance is most significant in the initial stage. An initial 10 g / m² of dust reduces photovoltaic output power by 34%, and at low dust mass density (<30 g / m²), the conversion efficiency decreases by an average of 3.4% for every 10 g / m² increase. Multiple studies have indicated that when the dust accumulation reaches the range of 8-12 g / m², the power generation efficiency loss is typically substantial, constituting a critical point for economical cleaning.

[0086] After obtaining the annual effective deposition rate, this invention further utilizes square root mapping to calculate a cleaning demand intensity score. A higher score indicates a stronger cleaning demand, specifically expressed as follows:

[0087] Score = sqrt( / T ref )×100,

[0088] Wherein, Score represents the intensity of cleaning demand, with a value of [0, 100]. T represents the annual effective sedimentation volume. ref For reference sedimentation amount, 100 g / m² is taken, corresponding to the extreme high sedimentation standard. sqrt() represents the square root operation.

[0089] The following scoring requirement correspondence table is used in this invention:

[0090] Table 1 - Relationship between Scoring Requirements

[0091] Rating range Cleaning requirement level 0-20 points Extremely low demand 20-40 points low demand 40-60 points medium demand 60-80 points High demand 80-100 points Extremely high demand

[0092] Meanwhile, considering the environmental uncertainties caused by precipitation, this invention further corrects the theoretical annual cleaning frequency of photovoltaic modules for precipitation, obtaining the actual annual cleaning frequency of photovoltaic modules, specifically:

[0093] Mode A, for arid and semi-arid regions, utilizes the effective rainfall monthly count method:

[0094] (1) Monthly scanning of the multi-year average monthly rainfall series;

[0095] (2) Months with monthly rainfall ≥ 50 mm are marked as effective rainfall months, and the count is 1. ;

[0096] (3) Actual number of cleanings per year calculate:

[0097] .

[0098] Mode B, for semi-humid and humid areas, uses the rainfall reduction factor method:

[0099] Reduction factor Determined by piecewise linear interpolation of the multi-mode average annual precipitation P:

[0100] ,

[0101] Actual number of cleanings per year: ,

[0102] A transition zone is defined based on an annual precipitation line threshold of 400 mm. A boundary fusion mechanism is used for the transition zone, and a weighted fusion is designed for the boundary area. In this embodiment, the transition zone is selected when the multi-mode average annual precipitation P is in the range of 350-450 mm.

[0103] To avoid spatial discontinuities caused by method switching, a weighted fusion coefficient is designed for the boundary region:

[0104] ,

[0105] ,

[0106] Further, the actual number of cleaning times N after precipitation correction will be... actual The score is obtained by converting the amount of sediment to an equivalent amount and then applying the square root mapping. The calculation process is as follows:

[0107] ,corrected=N actual ×T

[0108] Score_corrected=sqrt( ,corrected / T ref)×100;

[0109] Where Score_corrected is the cleaning demand intensity score, with a value of [0, 100]; N actual The corresponding values ​​are based on the calculation results for different regions. , , The calculation results for arid and semi-arid regions correspond to The calculation results for the semi-humid and humid regions correspond to The calculation results of the transition zone correspond to ;T ref For reference sedimentation amount, take 100 g / m², and sqrt() represents the square root operation.

[0110] This invention applies model data to photovoltaic cleaning demand forecasting, providing long-term forecasting capabilities under different combined scenarios. It overcomes the limitations of existing methods that rely on historical observation data and provides a basis for decision-making in response to climate change risks. Combined with a dual-model rainfall correction system, it adapts to the characteristics of different climate zones through differentiated rainfall correction algorithms, achieving unified modeling across multiple climate zones from arid to humid. It proposes a cleaning demand intensity scoring system (0-100 points) based on square root mapping, which improves the scoring resolution of low-deposition areas by about 30% compared to linear mapping, making the scoring distribution more in line with actual operation and maintenance decision-making needs, and avoiding the "polarization" phenomenon caused by traditional linear methods.

[0111] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the specific embodiments described above. The specific embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for predicting the demand for meteorological dust cleaning of photovoltaic modules based on numerical models, characterized in that, Includes the following steps: Step S1: From the multi-model dataset of the Sixth Coupled Model Intercomparison Project, determine at least one shared socio-economic path-typical concentration path (SSP-RCP) combined scenario; for each determined combined scenario, select future climate scenario simulation data from at least two different global climate models and historical monthly precipitation from the historical simulation experiments corresponding to the models. The future climate scenario simulation data shall include at least monthly dry aerosol deposition flux, monthly wet deposition flux, and monthly precipitation. Step S2: Select the target area. For each combined scenario, calculate the multi-model average annual dry deposition flux, annual wet deposition flux, and multi-model average annual precipitation for each grid point in the target area. Then, calculate the historical multi-model average annual precipitation for each grid point in the same target area based on the historical monthly precipitation obtained in Step S1. Step S3: Based on the historical multi-model average annual precipitation of each grid point and the annual precipitation classification threshold, the target area is divided into arid and semi-arid regions and semi-humid and humid regions. Step S4: Calculate the annual effective sedimentation amount corresponding to each grid point in the target region under each combined scenario. : ,in, The total annual deposition for each grid point is α, which is the effective deposition coefficient. α is a real number greater than 0 and less than 1. Its value is a coupling parameter used to correct the conversion of atmospheric particulate matter from vertical deposition flux to the actual deposition amount on the tilted surface of the photovoltaic module. β is the wind erosion correction factor, which is a real number greater than 0 and less than 1; it is used to correct the loss caused by secondary wind erosion of dust already deposited on the photovoltaic module surface due to the action of environmental wind field; the wind erosion correction factor in arid and semi-arid regions is less than that in semi-humid and humid regions. Step S5: Annual effective sedimentation rate based on the result obtained in step S4 Calculate the theoretical annual number of photovoltaic module cleaning operations for each grid point in the target area under each combined scenario. : ,in, The theoretical number of times the photovoltaic module is cleaned per year for each grid point, where T represents the photovoltaic module cleaning threshold. T is a positive number greater than 0, and the unit is g / m². The theoretical annual cleaning frequency of the photovoltaic module obtained in step S5 is corrected for rainfall to obtain the actual annual cleaning frequency of the photovoltaic module, specifically: In arid and semi-arid regions, the actual annual cleaning frequency of photovoltaic modules is calculated using the effective rainfall monthly count method. : ,in The month with effective rainfall; In semi-humid and humid regions, the actual annual cleaning frequency of photovoltaic modules is calculated using the rainfall reduction factor method. : ,in This is the reduction factor; Transition zones are defined based on annual precipitation thresholds. A boundary fusion mechanism is used for these transition zones, with weighted fusion applied to the boundary areas. , in, This refers to the actual number of cleaning cycles performed in the transition zone. This is the weighted fusion coefficient, which takes a value between 0 and 1.

2. The method for predicting the meteorological dust accumulation cleaning demand of photovoltaic modules based on numerical models according to claim 1, characterized in that, The reduction factor Determined based on piecewise linear interpolation of annual rainfall: ， Where p is the multi-model average annual precipitation.

3. The method for predicting the meteorological dust accumulation cleaning demand of photovoltaic modules based on numerical models according to claim 1 or 2, characterized in that, The actual number of times the photovoltaic modules were cleaned after rainfall correction, N. actual The equivalent deposition amount is converted, and then the square root mapping is applied to obtain the corrected score, as follows: ,corrected = N actual × T, Score_corrected=sqrt( ,corrected / T ref )×100, Where Score_corrected is the cleaning demand intensity score, with a value of [0, 100]; N actual The corresponding values ​​are based on the calculation results for different regions. , , ;T ref For reference sedimentation amount, sqrt() represents the square root operation.

4. The method for predicting the meteorological dust accumulation cleaning demand of photovoltaic modules based on numerical models according to claim 1, characterized in that, Annual effective sedimentation rate obtained in step S4 The cleaning demand intensity score is calculated using the square root mapping, specifically expressed as follows: Score=sqrt( / T ref )×100, Wherein, Score represents the intensity of cleaning demand, with a value of [0, 100]. T represents the annual effective sedimentation volume. ref For reference sedimentation amount, sqrt() represents the square root operation.

5. The method for predicting the meteorological dust accumulation cleaning demand of photovoltaic modules based on numerical models according to claim 1, characterized in that, The shared socioeconomic pathway-typical concentration pathway SSP-RCP combination scenario mentioned in step S1 includes SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.

5.

6. The method for predicting the demand for meteorological dust cleaning of photovoltaic modules based on numerical models according to claim 1, characterized in that, The global climate models selected in step S1 are eight models supported by literature on the performance validation of aerosol-climate interaction simulation.

7. The method for predicting the meteorological dust accumulation cleaning demand of photovoltaic modules based on numerical models according to claim 1, characterized in that, In step S4, the effective deposition coefficient α is calculated through the following coupling relationship: α = cosθ × η_r × η_e, Where θ is the installation tilt angle of the photovoltaic module, η_r is the surface micro-roughness capture enhancement coefficient, and η_e is the deposition loss coefficient caused by flow around the frame.

8. The method for predicting the meteorological dust accumulation cleaning demand of photovoltaic modules based on numerical models according to claim 1, characterized in that, The data extraction process in step S2 is as follows: Step S21: Based on the latitude and longitude boundaries of the target area, extract a subset of grid data corresponding to the spatial range from the global grid data; Step S22: Convert the original time series data of each grid point into a monthly time scale unit; Step S23: For each grid point in the target area, under the same combined scenario, perform an equal-weighted ensemble average on the monthly scale data after each global climate model conversion, and output the multi-model average monthly dry aerosol deposition flux, monthly wet deposition flux and monthly precipitation data for each grid point in the target area under each combined scenario; at the same time, perform an equal-weighted ensemble average on the historical monthly precipitation data, and output the multi-model historical average monthly precipitation data for each grid point in the target area. Step S24: Calculate the corresponding multi-model average annual dry deposition flux, annual wet deposition flux, and annual precipitation based on the multi-model average monthly dry deposition flux, monthly wet deposition flux, and monthly precipitation data obtained in step S23; calculate the historical multi-model average annual precipitation data based on the multi-model historical average monthly precipitation data obtained in step S23.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the prediction method as described in any one of claims 1-8.

Images