Prediction method for power generation attenuation rate of photovoltaic power station, computer equipment and storage medium

Abstract

The invention discloses a photovoltaic power station power generation attenuation rate prediction method, computer equipment and a storage medium, and belongs to the technical field of new energy power generation. According to the method, the theoretical non-attenuation generating capacity of the photovoltaic power station is simulated, and the STC theoretical attenuation rate committed by a manufacturer is eliminated, so that precise separation and quantification of the additional attenuation rate caused by environmental stress such as temperature, humidity and dust in actual operation are realized, and the authenticity and comparability of attenuation rate evaluation are improved; according to the method, multi-source data such as satellite remote sensing, meteorological reanalysis and geographic information are comprehensively used, a multi-dimensional environment feature system such as climate, geography and pollution is constructed, time sequence alignment and spatial scale unification are achieved, and the comprehensiveness and precision of model input data are improved. According to the method, large-scale attenuation rate evaluation from point-to-surface-to-area is realized, data driving and physical mechanisms are considered, and the problems of poor area applicability, insufficient meteorological correction, no consideration of spatial heterogeneity and the like in the prior art are effectively solved.

Description

Technical Field

[0001] This invention belongs to the field of new energy power generation technology, and specifically relates to a method, computer equipment, and storage medium for predicting the power generation attenuation rate of photovoltaic power plants based on multi-source data fusion. Background Technology

[0002] In recent years, photovoltaic (PV) power generation systems have developed rapidly, making outstanding contributions to my country's energy security and energy conservation and emission reduction. The advancement of PV power generation systems mainly relies on the development of module technology, with crystalline silicon modules currently still dominating the market. With the large-scale application of these modules, the issue of performance degradation has become increasingly prominent. This degradation involves not only the solar cells themselves but also changes in components such as the grid, interconnects, glass, encapsulation, and backsheet. Different installation locations and climatic conditions can cause differences in degradation characteristics, and the causes and mechanisms are not yet fully understood.

[0003] Crystalline silicon photovoltaic modules are affected by environmental stress during long-term operation, and the power degradation rate directly impacts the power generation revenue of the power plant. The "Photovoltaic Manufacturing Industry Standard Conditions (2024 Edition)" issued by the Ministry of Industry and Information Technology in 2024 clearly stipulates the module efficiency degradation rate indicators: P-type crystalline silicon modules: first-year degradation rate not exceeding 2%, subsequent annual degradation rate not exceeding 0.55%, and cumulative degradation rate not exceeding 15% over 25 years; N-type crystalline silicon modules: first-year degradation rate not exceeding 1%, subsequent annual degradation rate not exceeding 0.4%, and cumulative degradation rate not exceeding 11% over 25 years; Thin-film modules: first-year degradation rate not exceeding 4%, subsequent annual degradation rate not exceeding 0.4%, and cumulative degradation rate not exceeding 14% over 25 years. Industry research shows that when the average annual degradation rate of a module exceeds 0.5%, the power generation loss over its 25-year lifespan can reach more than 10%. The current standard test condition (STC) degradation rates promised by mainstream manufacturers (such as 2% in the first year and 0.5% in each subsequent year) cannot accurately reflect the additional degradation under complex outdoor environments, leading to deviations in power generation prediction and errors in operation and maintenance decisions.

[0004] Current methods for detecting photovoltaic module degradation rates using traditional current-voltage (IV) curves require system shutdown, while electroluminescence (EL) imaging is expensive and difficult to implement for real-time monitoring of large-scale power plants, let alone simulate the overall degradation rate of a power plant within a region. Existing degradation prediction models largely rely on accelerated aging data from laboratories, failing to adequately consider the interaction between the environment and materials. For example, the synergistic effect of dust deposition and high temperatures in desert regions can result in an actual degradation rate of 1.2% / year, significantly higher than the theoretical laboratory value of 0.5% / year. Furthermore, current solutions for Potential Induced Degradation (PID) repair and cleaning maintenance are mostly reactive, lacking proactive protection measures based on environmental stress prediction.

[0005] In summary, the existing technology has the following problems: 1. The disconnect between laboratory accelerated aging models and actual outdoor degradation.

[0006] Currently, the calculation of the annual degradation rate of photovoltaic power station modules relies on laboratory standard tests such as IEC 61215. Its constant stress conditions (85 degrees Celsius / 85% relative humidity) cannot simulate outdoor dynamic environments such as freeze-thaw cycles, wet and dry alternation, dust, and ultraviolet coupling, resulting in a prediction error as high as ±0.3% / year.

[0007] 2. High data dependence and limited regional applicability.

[0008] Existing linear regression methods for annual degradation rate based on performance ratio (PR) time series mainly rely on multi-year PR series data of photovoltaic modules in individual power plants. However, most regions lack continuous and complete annual PR records for power plants, making it difficult to calculate the annual degradation rate of photovoltaic modules in power plants and even more difficult to extend to the region.

[0009] 3. Insufficient correction for meteorological and environmental factors.

[0010] Existing methods, when calculating degradation rates using power generation or PR time series, fail to adequately correct for differences in meteorological conditions across different years, such as irradiance, temperature, wind speed, precipitation, and dust storms. This leads to climate fluctuations being misjudged as component performance degradation, severely impacting calculation accuracy. Furthermore, the lack of quantitative separation of environmentally-related degradation makes it difficult to distinguish between "natural aging degradation" and "environmentally-related degradation."

[0011] 4. Regional-level assessments lack consideration of spatial heterogeneity.

[0012] Current research on photovoltaic (PV) degradation rates mainly focuses on the level of individual power plants, calculating the annual degradation rate by fitting multi-year production performance (PR) or power generation data of a single power plant. This approach lacks a systematic estimation of the overall degradation rate of PV modules across a region, such as a province. It fails to fully integrate operational data from multiple power plants within the region, as well as spatial distribution characteristics such as meteorological and environmental factors. Furthermore, it ignores significant differences in climate conditions, air pollution levels, altitude, and topography across different geographical locations, making it difficult to accurately reflect the overall level and spatial distribution of PV module performance degradation within a region.

[0013] How to correlate changes in material properties with environmental stress in real time, establish a high-precision attenuation prediction model, and dynamically optimize operation and maintenance strategies has become a technical challenge that the industry urgently needs to solve.

[0014] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0015] The purpose of this invention is to solve the problems existing in the prior art and to provide a method for predicting the power generation attenuation rate of a photovoltaic power plant, a computer device, and a storage medium.

[0016] The first aspect of this invention provides a method for predicting the power generation degradation rate of a photovoltaic power plant, comprising the following steps: S1. Obtain historical power generation, continuous meteorological data of the photovoltaic environment, and photovoltaic module parameter data of centralized photovoltaic power stations in the target area. The centralized photovoltaic power station is referred to as a photovoltaic power station. S2. Calculate the annual environmentally-added degradation rate of the power generation of each photovoltaic power station; S3. Divide the country into different resource zones, classify each photovoltaic power station according to the resource zones, and extract the photovoltaic power station mask. The photovoltaic power station mask refers to the spatial boundary or geographical range of the photovoltaic power station. S4. Based on the photovoltaic power station mask, determine and obtain the time series data of environmental factors related to the power generation attenuation of each photovoltaic power station under the influence of environmental stress, calculate the annual average value of the data, and then align it with the data of the annual environmental additional attenuation rate. S5. Establish machine learning models, namely photovoltaic environmental additional attenuation rate prediction models, for the photovoltaic power stations after regional classification to predict the environmental additional attenuation rate of the resource area. S6. Extend the photovoltaic environmental degradation rate prediction model from the scale of some power plants to the scale of the expected area, and predict the photovoltaic environmental degradation rate of the entire expected area, i.e. the photovoltaic environmental degradation rate of the expected area. S7. The actual degradation rate of photovoltaic modules in the expected area is obtained by back-calculating the sum of the average theoretical degradation rate of photovoltaic modules in the expected area and the additional degradation rate of photovoltaic environment in the expected area.

[0017] In one embodiment of the present invention, step S2 includes: Simulated theoretical zero-attenuation power generation of the photovoltaic power station under the target geographical location and corresponding meteorological conditions: , In the above formula, : No. The photovoltaic power station mentioned above The theoretical annual power generation without degradation, in kilowatt-hours. and It is an integer; The initial nominal power of the photovoltaic modules used in the photovoltaic power station is expressed in kilowatts. : No. Total solar irradiance on the tilted surface of the sky, expressed in kilowatt-hours per square meter. The value is an integer, and the inclined surface is the inclined surface of the photovoltaic panel in the photovoltaic module relative to the ground; Reference irradiance under standard test conditions, in kilowatt-hours per square meter; The power temperature coefficient of the photovoltaic module is expressed as percentages per degree Celsius. : No. The average operating temperature of the photovoltaic modules described above is in degrees Celsius. The reference temperature under standard test conditions is taken as 25 degrees Celsius. System efficiency, including inverter efficiency, line loss, and transformer loss, is dimensionless. The total number of days in a year Calculate annual performance ratio: , In the above formula, : No. The photovoltaic power station mentioned above Annual performance ratio, dimensionless; : No. The photovoltaic power station mentioned above The actual measured annual power generation is expressed in kilowatt-hours. Calculate the actual annual decay rate: , In the above formula, : No. The photovoltaic power station mentioned above The actual annual decay rate, expressed as a percentage, for the initial year, i.e. When =1, directly use the manufacturer's promised first-year degradation rate. , Calculate the environmentally-added degradation rate caused by the aforementioned environmental factors: , In the above formula, : No. The photovoltaic power station mentioned above The environmentally-added degradation rate for the year; : No. The theoretical attenuation rate under standard annual test conditions is calculated based on the manufacturer's promised linear attenuation model: , In the above formula, This refers to the manufacturer's promised annual average degradation rate.

[0018] The environmentally-added degradation rate refers to the additional power degradation of the photovoltaic module in actual operating environment due to environmental stress exceeding the laboratory standard test conditions.

[0019] In one embodiment of the present invention, step S3 includes: According to the "Solar Energy Resource Assessment Method" (GB / T 37526-2019), the country is divided into four resource zones. Photovoltaic power stations are then regionally classified according to these four resource zones. The photovoltaic power station masks are extracted using a geographic information system platform. The four categories of resource zones include: Category I, the extremely abundant zone: areas with a GHR ≥ 1750 kWh / m²; Category II, the very abundant zone: areas with a GHR of 1400 ≤ GHR < 1750 kWh / m²; Category III, the abundant zone: areas with a GHR of 1050 ≤ GHR < 1400 kWh / m²; and Category IV, the general zone: areas with a GHR < 1050 kWh / m². The GHR refers to the total amount of solar radiation received at a horizontal surface on the Earth's surface.

[0020] In one embodiment of the present invention, the environmental factors are divided into: Climate category, including elements: Average annual temperature, data from China Meteorological Administration website; Annual average land surface temperature, data from MODIS sensor land surface temperature or Landsat satellite land surface temperature; Annual average surface shortwave radiation, data from FY-4B and Himawari 8 / 9 satellites; Average annual total precipitation, data from ERA5 data and Global Precipitation Observation Program data; Annual precipitation days, data sourced from China Meteorological Administration Network and ERA5 data; Snowfall, snow accumulation, and hail; data from the China Meteorological Administration website. Dust storm data from the China Meteorological Administration website. Geographic category, including elements: Slope, aspect, and elevation data are derived from Space Shuttle radar topographic mapping mission data or digital elevation model data. The shadow of the mountain was calculated by a geographic information system platform using a digital elevation model. Land cover type or land use type, data sourced from the global 30-meter land cover dataset. Pollution categories, including elements: Annual average aerosol optical thickness, data from Sentinel-5P satellite and global land evapotranspiration dataset; Dust / dust, data from China Meteorological Administration website; Ammonia gas was used as a substitute, with the data calculated using the Euclidean distance from the pasture through a geographic information system platform. Bird droppings were replaced with data calculated using a geographic information system platform based on Euclidean distances from farmland, forests, and wetlands. The data obtained in step S4 includes raster data and non-raster data. The raster data includes the data of the environmental factors other than the numerical data obtained from the China Meteorological Administration Network. The non-raster data includes the numerical data of the environmental factors obtained from the China Meteorological Administration Network.

[0021] In one embodiment of the present invention, step S5 includes: S51. Data preprocessing, including the following steps: S511. In the geographic information system platform, the raster data is first geographically corrected to unify the spatial scale, and then the photovoltaic power station mask is used to extract the required feature values.

[0022] S512, The feature data of the non-raster data is obtained by collecting and extracting the corresponding numerical data from the China Meteorological Administration Network based on the center coordinates of the photovoltaic power station.

[0023] S513. Missing feature values ​​are filled with the corresponding feature values ​​of adjacent years of the same photovoltaic power station; environmental degradation rate data other than ±3 times the standard deviation are removed; the numerical feature values ​​are normalized to obtain the photovoltaic power station dataset.

[0024] S52. Dataset partitioning, including the following steps: S521. Divide the photovoltaic power station dataset according to the four types of resource areas to ensure that the photovoltaic power station data in the same resource area are divided into the same photovoltaic power station dataset.

[0025] S522. For each of the photovoltaic power station datasets obtained in step S521, divide the data into 60% of the years as the training set, 20% of the years as the validation set, and 20% of the years as the test set.

[0026] S523. During model training, the time series data is sequentially divided into multiple consecutive time periods, and the validation set and test set are selected alternately multiple times in each time period to evaluate the model's performance in different time periods.

[0027] S52. Model training optimization and result evaluation.

[0028] In one embodiment of the present invention, step S6 includes: The expected region is assigned to a corresponding resource zone according to the division of the resource zone, and a machine learning model is trained for different resource zones. The photovoltaic device mask of the expected area is extracted from high-resolution remote sensing data using a semantic segmentation method. The photovoltaic device mask of the expected area refers to the spatial boundary or geographical range of all photovoltaic devices in an expected area. After replacing the photovoltaic power station mask with the photovoltaic equipment mask in the expected area, the time series data of environmental factors in the expected area are obtained according to step S4, and the data are calculated and processed into an annual average value, which includes the grid data and non-grid data. According to step S51, after preprocessing the annual average data of environmental factors in the expected area, the obtained feature values ​​are input into the machine learning model trained for the corresponding resource area to predict the photovoltaic environmental degradation rate of the expected area.

[0029] In one embodiment of the present invention, step S7 includes calculation: , In the above formula, : No. The expected actual photovoltaic degradation rate in the region for the year; : No. The expected regional photovoltaic environmental degradation rate for the year; : No. The average theoretical degradation rate of photovoltaic modules in the expected region for the year is the average of the degradation rates promised by different manufacturers.

[0030] In one embodiment of the present invention, step S1 includes: Obtain historical power generation data for each of the photovoltaic power stations, wherein the photovoltaic power stations are of the same model as the photovoltaic modules they use; Meteorological data is acquired through the photovoltaic environmental monitoring instruments equipped in the photovoltaic power station. The meteorological data includes continuous temperature, humidity, irradiance, precipitation, and wind speed of the power station environment. Obtain the technical parameters of the photovoltaic modules used in the photovoltaic power station. The technical parameters include: nominal power under standard test conditions, maximum conversion efficiency, power tolerance, first-year power degradation, and 2-30-year power degradation.

[0031] A second aspect of the present invention provides a computer device including a memory, a first processor, and a first computer program stored in the memory and executable on the first processor, wherein the first computer program, when executed by the first processor, implements the above-described method for predicting the power generation attenuation rate of a photovoltaic power plant.

[0032] A third aspect of the present invention provides a computer-readable storage medium for storing a second computer program, which can be executed by at least one second processor to perform the above-described method for predicting the power generation attenuation rate of a photovoltaic power plant.

[0033] Compared with the prior art, the technical effects achieved by the present invention are as follows: 1. Precisely separate environmental attenuation.

[0034] By simulating the theoretical zero-attenuation power generation of a photovoltaic power plant and eliminating the theoretical attenuation rate under the standard test conditions (STC) promised by the manufacturer, the additional attenuation rate caused by environmental stresses such as temperature, humidity, and dust in actual operation is accurately separated and quantified, thereby improving the authenticity and comparability of attenuation rate assessment.

[0035] 2. Fusion of multi-source remote sensing and meteorological data.

[0036] By comprehensively utilizing multi-source data such as satellite remote sensing, meteorological reanalysis, and geographic information, a multi-dimensional environmental feature system encompassing climate, geography, and pollution is constructed. Furthermore, time series alignment and spatial scale unification are achieved, thereby enhancing the comprehensiveness and accuracy of the model's input data.

[0037] 3. Strong regional partitioning modeling and promotion capabilities.

[0038] Regional sub-models are established according to solar energy resource zones, and ensemble learning strategies such as Stacking and Bagging are used to improve prediction performance. This not only ensures the generalization ability of the model, but also extends some power plant data to larger regions such as the provincial level, so as to achieve overall prediction of regional photovoltaic degradation rate.

[0039] 4. Remote sensing automatically extracts the distribution of photovoltaic equipment.

[0040] By leveraging deep learning segmentation algorithms such as DeepLab and YOLOv8, the distribution masks of photovoltaic equipment can be automatically identified, supporting the large-scale batch extraction of power plant environmental features and enabling large-scale degradation rate assessment from point-to-power plant to area-to-province.

[0041] 5. Improved prediction accuracy and applicability.

[0042] It takes into account both data-driven and physical mechanisms, and can be applied to power plants with complete data, as well as to areas lacking long-term monitoring data. It effectively solves the problems of poor regional applicability, insufficient meteorological correction, and lack of consideration of spatial heterogeneity in existing technologies. Attached Figure Description

[0043] Figure 1 This is a flowchart of a method for predicting the power generation attenuation rate of a photovoltaic power plant according to an embodiment of the present invention. Detailed Implementation

[0044] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0045] The technical solution of the present invention is illustrated below through specific embodiments. It should be understood that the one or more steps mentioned in the present invention do not preclude the existence of other methods and steps before or after the combined steps, or that other methods and steps may be inserted between these explicitly mentioned steps. It should also be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Unless otherwise stated, the numbering of each method step is only for the purpose of identifying each method step, and not for limiting the order of each method or limiting the scope of the present invention. Changes or adjustments to their relative relationships, without substantial changes to the technical content, can also be considered as within the scope of the present invention.

[0046] The raw materials and instruments used in the examples are not subject to any specific restrictions on their source; they can be purchased from the market or prepared according to conventional methods known to those skilled in the art.

[0047] Figure 1 This is a flowchart of the method for predicting the power generation attenuation rate of a photovoltaic power station provided by the present invention. Figure 1 As shown, the method for predicting the power generation attenuation rate of a photovoltaic power plant according to a preferred embodiment of the present invention includes the following steps.

[0048] S1. Consult centralized photovoltaic (PV) power plants in the target area to obtain historical power generation, continuous meteorological data of the PV environment, and PV module parameter data, etc., of these centralized PV power plants. Specifically, this includes the following steps: S11. Obtain the annual power generation data of each photovoltaic power station, wherein the photovoltaic power station is of the same model as the photovoltaic modules it uses; S12. Obtain continuous meteorological data such as temperature, humidity, irradiance, precipitation, and wind speed of the power station environment through the photovoltaic environmental monitoring instruments equipped in the photovoltaic power station. S13. Obtain the technical parameters of the photovoltaic modules used in the photovoltaic power station. The technical parameters include: STC nominal power, maximum conversion efficiency, power tolerance, first-year power degradation, and 2-30-year power degradation. S2. Calculate the annual environmentally-added degradation rate of power generation for each photovoltaic power station. This involves using the geographical location of each photovoltaic power station, the technical parameters of the photovoltaic modules, and annual meteorological data such as irradiance and temperature of the power station environment. Based on physical performance modeling methods from engineering, simulations are performed using physical laws and module characteristics to simulate the theoretical zero-degradation power generation of the photovoltaic power station under specific geographical locations (target locations) and meteorological conditions. The actual power generation is then compared with the theoretical power generation for the corresponding year to calculate the annual actual degradation rate. Further, the manufacturer's promised STC theoretical degradation rate (e.g., 2% in the first year and 0.5% in subsequent years) is deducted, and then the additional degradation rate purely caused by environmental factors is separated out. Specifically, this includes the following steps: S21. Theoretical, non-attenuated power generation of a simulated photovoltaic power station under target geographical location and corresponding meteorological conditions: , In the above formula, : No. The first photovoltaic power station The theoretical annual power generation without degradation is expressed in kilowatt-hours (kWh). and It is an integer; The initial nominal power of the photovoltaic modules used in a photovoltaic power station, expressed in kilowatts (kW). : No. Total solar irradiance on the tilted surface of the sky, expressed in kilowatt-hours per square meter (kWh / m²). 2 ), The value is an integer, where the inclined surface is the inclined surface of the photovoltaic panel in the photovoltaic module relative to the ground; Reference irradiance at STC, in kilowatt-hours per square meter (kWh / m²) 2 ); The power temperature coefficient of photovoltaic modules used in photovoltaic power plants, expressed as percentages per degree Celsius (% / ℃). : No. The average operating temperature of the photovoltaic modules described above is in degrees Celsius (°C). The reference temperature under STC is taken as 25℃; System efficiency, including inverter efficiency, line loss, transformer loss, etc., is dimensionless. Total number of days in a year.

[0049] S22. Calculate the annual performance ratio. The annual performance ratio is a core indicator that eliminates the impact of meteorological fluctuations and measures the operating status of photovoltaic power plants. , In the above formula, : No. The first photovoltaic power station Annual performance ratio, dimensionless; : No. The first photovoltaic power station The actual measured annual power generation is expressed in kilowatt-hours (kWh). S23. Calculate the annual actual degradation rate, which represents the annual decline rate of the overall performance of the photovoltaic power station (including: module aging, system efficiency): , In the above formula, : No. The first photovoltaic power station The actual annual decay rate, in percentage (%); For the starting year, i.e. When the value is 1, there is no data from the previous year for comparison, so the manufacturer's promised first-year degradation rate is used directly. .

[0050] The so-called annual actual degradation rate refers to the percentage decrease in power generation performance of a photovoltaic power station due to the combined effects of component aging, system efficiency loss, and environmental stress in adjacent years. The calculation of the annual actual degradation rate eliminates the influence of annual fluctuations in sunlight, temperature, and other meteorological factors by comparing the actual power generation with the theoretical power generation without degradation.

[0051] S24. Subtract the natural aging portion of the photovoltaic module itself from the total degradation, i.e., the actual annual degradation rate, to obtain the target variable, i.e., the manufacturer's STC theoretical degradation rate, and extract the environmentally-added degradation rate caused by environmental factors: , In the above formula, : No. The first photovoltaic power station Annual environmental degradation rate; : No. The annual STC theoretical attenuation rate is calculated based on the linear attenuation model promised by the manufacturer: , In the above formula, For example, the manufacturer's promised annualized rate of degradation. =0.5% / year.

[0052] The environmentally-added degradation rate refers to the additional power degradation of photovoltaic modules in actual operating environments due to environmental stresses exceeding the laboratory STC; the STC theoretical degradation rate refers to the degradation rate of photovoltaic modules at STC (25℃, 1000W / m²). 2 Under irradiance and atmospheric quality (1.5 spectral density), the manufacturer's promised annual power degradation rate.

[0053] S3. According to the "Solar Energy Resource Assessment Method" (Standard No.: GB / T 37526-2019), the country is divided into four resource zones. Photovoltaic power stations are then regionally classified according to these four resource zones. The masks for each photovoltaic power station are extracted using a Geographic Information System (GIS) platform. The photovoltaic power station mask refers to the spatial boundary or geographical area where a photovoltaic power station is located, and the data for the photovoltaic power station mask is stored as a vector file. The four resource zones include: Class I areas, i.e., extremely abundant zones: GHR ≥ 1750 kWh / (m²) 2 • Areas with average (a, where a represents the average) mainly include most of Tibet, southern Xinjiang, and western Qinghai, Gansu, and Inner Mongolia; Category II areas are the very rich zones: the GHR indicator meets the following criteria: ≥1400≤GHR<1750kWh / (m³). 2 Regions classified as a) mainly include most of Xinjiang, eastern Qinghai and Gansu, Ningxia, Shaanxi, Shanxi, Hebei, northeastern Shandong, eastern Inner Mongolia, southwestern Northeast China, Yunnan, and northern Sichuan; Region III, also known as the abundance zone, is defined as having a GHR (Gross Flow Rate) of 1050 ≤ GHR < 1400 kWh / (m³). 2 •a) The main regions include Heilongjiang, Jilin, Liaoning, Anhui, Jiangxi, southern Shaanxi, northeastern Inner Mongolia, Henan, Shandong, Jiangsu, Zhejiang, Hubei, Hunan, Fujian, Guangdong, Guangxi, eastern Hainan, Sichuan, Guizhou, southeastern Tibet, and Taiwan; Class IV regions, i.e., the general zone, are defined as having a GHR < 1050 kWh / (m³). 2•a) The main regions include Chongqing, central Sichuan, northern Guizhou, and northwestern Hunan. The GHR (Gross Horizontal Irradiance) index refers to the total amount of solar radiation received on the horizontal plane of the Earth's surface, which includes both direct normal irradiance (DNI) and diffuse horizontal irradiance (DHI).

[0054] S4. Based on the photovoltaic power station mask, determine and obtain the time series data of environmental factors related to the power generation attenuation of each photovoltaic power station under the influence of environmental stress, calculate the annual average value of the data, and then align it with the data of the annual environmental additional attenuation rate.

[0055] The environmental factors mentioned above are divided into several categories: S4a, Climate category, including elements: Average annual temperature, data from China Meteorological Administration website; Annual average land surface temperature (LST), data from MODIS sensor LST or Landsat satellite LST; Annual average surface shortwave radiation, data from FY-4B and Himawari 8 / 9 satellites; Total annual precipitation (TPW), data from ERA5 and Global Precipitation Observation Programme (GPM) data; Annual precipitation days, data sourced from China Meteorological Administration Network and ERA5 data; Snowfall, snow accumulation, and hail; data from the China Meteorological Administration website. Dust storm data from the China Meteorological Administration website; S4b, Geographic category, including elements: Slope, aspect, and elevation data are derived from Space Shuttle Radar Topography Mission (SRTM) data or Digital Elevation Model (DEM) data. Mountain shadow data was calculated by the GIS platform using a DEM. Land cover type or land use type, data from the Global 30-meter Land Cover Dataset (GLAD LandCover). S4c, Pollution category, including elements: Annual average aerosol optical thickness (AOD), data from Sentinel-5P satellite and global land actual evapotranspiration dataset (MERRA-2). Dust / dust, data from China Meteorological Administration website; Ammonia (NH3) was used as a substitute, with the data calculated using the Euclidean distance from the pasture through a GIS platform. Bird droppings were replaced by Euclidean distances from farmland, forests, and wetlands calculated using a GIS platform.

[0056] The data obtained in step S4 includes raster data and non-raster data. The raster data includes the data of the above-mentioned environmental factors other than the numerical data obtained from the China Meteorological Administration Network. The non-raster data includes the numerical data obtained from the China Meteorological Administration Network.

[0057] S5. Establish machine learning models, i.e., photovoltaic environmental degradation rate prediction models, for photovoltaic power plants after regional classification to predict the environmental degradation rate of each resource area. The models include, but are not limited to, Random Forest (RF), Gradient Boosting Tree (XGBoost, LightGBM), etc.; perform ensemble learning (Stacking / Bagging) on ​​the models (denoted as sub-models) in the four different resource areas to improve overall prediction performance and reduce overfitting and bias. Specifically, this includes the following steps: S51. Data preprocessing, including the following steps: S511. In the GIS platform, first perform geographic correction on the raster data to unify the spatial scale, such as 1 kilometer (km) resolution, and then use the photovoltaic power station mask to extract the required feature data.

[0058] S512. The feature data of non-raster data is obtained by collecting and extracting the corresponding numerical data from the China Meteorological Administration Network based on the center coordinates of the photovoltaic power station.

[0059] S513. Missing feature values ​​are filled with the corresponding feature values ​​of adjacent years of the same photovoltaic power station; environmental degradation rate data other than ±3 times the standard deviation are removed; the numerical feature values ​​are normalized to obtain the photovoltaic power station dataset.

[0060] This step identifies the photovoltaic power plant as the target unit for the prediction model of the photovoltaic environmental degradation rate.

[0061] S52. Dataset partitioning, including the following steps: S521. Divide the photovoltaic power station dataset according to the four resource zones in step S3 to ensure that the photovoltaic power station data in the same resource zone is divided into the same dataset.

[0062] S522. For each dataset obtained in step S521, divide the data into 60% of the years as the training set, 20% of the years as the validation set, and 20% of the years as the test set.

[0063] S523. During model training, the time series data is sequentially divided into multiple consecutive time periods (also known as sliding windows), and the validation set and test set are selected alternately multiple times (more than 1 time period) in each time period to evaluate the model's performance in different time periods.

[0064] S53. Model training optimization and result evaluation, including the following steps: S531. Use machine learning models such as Random Forest (RF) and Gradient Boosting Tree (XGBoost, LightGBM) to train models for the four types of resource regions respectively, and use ensemble learning (Stacking / Bagging) methods to improve the model prediction performance in each resource region.

[0065] S532. Use Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Coefficient of Determination (R²). 2 To evaluate the performance and accuracy of the model.

[0066] S533. Use SHAP to interpret the basis of model predictions, improving the interpretability and reliability of the model. If a feature has little information or is noisy, it can be deleted to reduce model complexity and the risk of overfitting; or the power plant sample distribution can be increased for further analysis.

[0067] S6. Extend the prediction target unit of the above photovoltaic environmental degradation rate prediction model from the power plant to the expected area (e.g., provincial level), and predict the overall photovoltaic environmental degradation rate of the expected area. Note that the scale of the expected area is larger than the scale of the power plant; that is, relatively speaking, the scale of the expected area is large-scale, and the scale of the power plant is small-scale. Specifically, this includes the following steps: S61. The expected region is assigned to the corresponding resource zone according to the division of resource zones, and a machine learning model is trained for the different resource zones mentioned above.

[0068] S62. Using semantic segmentation methods such as DeepLab or instance segmentation methods such as YOLOv8, extract photovoltaic device masks for the expected region from high-resolution remote sensing data from sources such as Google Earth and Jilin-1 satellite. The photovoltaic device mask for the expected region refers to the spatial boundary or geographical range of all photovoltaic devices in an expected region. The data of the photovoltaic device mask for the expected region is stored as a vector file.

[0069] S63. After replacing the photovoltaic power station mask with the photovoltaic equipment mask in the expected area, obtain the time series data of environmental factors in the expected area according to step S4, and calculate and process it into an annual average value, which includes the grid data and non-grid data.

[0070] S64. Following step S51, the annual average environmental factor data of the expected area obtained in step S63 is preprocessed, and the obtained feature values ​​are input into the trained machine learning model of the corresponding resource area to predict the photovoltaic environmental degradation rate of the expected area.

[0071] S7. The actual photovoltaic degradation rate of the expected area is obtained by back-calculating the sum of the average theoretical degradation rate of photovoltaic modules in the expected area (the average of the degradation rates of photovoltaic modules promised by multiple different manufacturers) and the additional degradation rate of photovoltaic environment in the expected area.

[0072] , In the above formula, : No. Annual expected regional photovoltaic actual degradation rate; : No. Annual expected regional photovoltaic environmental degradation rate; : No. The expected average theoretical degradation rate of regional photovoltaic modules in the current year.

[0073] The present invention also provides a computer device, which includes a memory, a first processor, and a first computer program stored in the memory and executable on the first processor. When the first computer program is executed by the first processor, it implements the above-described method for predicting the power generation attenuation rate of a photovoltaic power plant.

[0074] The present invention also provides a computer-readable storage medium for storing a second computer program, which can be executed by at least one second processor to enable the at least one second processor to perform the above-described method for predicting the power generation attenuation rate of a photovoltaic power plant.

[0075] This invention achieves: 1. Separation and quantification of environmentally-added attenuation rate.

[0076] For the first time, the additional attenuation rate caused by environmental stresses such as temperature, humidity, and dust in the actual operating environment is separated and quantified.

[0077] 2. Construction of environmental features by fusing multi-source remote sensing and meteorological data.

[0078] By comprehensively utilizing multi-source data such as satellite remote sensing (MODIS, Sentinel-5P, FY-4B, etc.), meteorological reanalysis (ERA5, GPM), and geographic information (DEM, land use), a multi-dimensional environmental feature system affecting the degradation of photovoltaic modules, including climate, geography, and pollution, is constructed, and annual-scale temporal alignment and spatial matching are performed.

[0079] 3. Resource zone modeling and regional promotion mechanism.

[0080] According to the solar energy resource zoning standard GB / T 37526-2019, power plants across the country are divided into four resource zones. Machine learning models (RF, XGBoost, LightGBM) are trained in each resource zone, and the prediction performance is improved by stacking / bagging integration. The models are then extended to the overall degradation rate prediction of regions (such as provinces).

[0081] 4. Back-calculation of the actual degradation rate of regional photovoltaic power.

[0082] The actual degradation rate of photovoltaic modules in the region is calculated, enabling the application of measured data from some power plants to the overall degradation rate of the region.

[0083] The foregoing description of specific exemplary embodiments of the invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it will be apparent that many changes and variations can be made in accordance with the foregoing teachings. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize various different exemplary embodiments of the invention, as well as various different choices and variations. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A method for predicting the power generation degradation rate of a photovoltaic power plant, characterized in that, Including the following steps: S1. Obtain historical power generation, continuous meteorological data of the photovoltaic environment, and photovoltaic module parameter data of centralized photovoltaic power stations in the target area. The centralized photovoltaic power station is referred to as a photovoltaic power station. S2. Calculate the annual environmentally-added degradation rate of the power generation of each photovoltaic power station; S3. Divide the country into different resource zones, classify each photovoltaic power station according to the resource zones, and extract the photovoltaic power station mask. The photovoltaic power station mask refers to the spatial boundary or geographical range of the photovoltaic power station. S4. Based on the photovoltaic power station mask, determine and obtain the time series data of environmental factors related to the power generation attenuation of each photovoltaic power station under the influence of environmental stress, calculate the annual average value of the data, and then align it with the data of the annual environmental additional attenuation rate. S5. Establish machine learning models, namely photovoltaic environmental additional attenuation rate prediction models, for the photovoltaic power stations after regional classification to predict the environmental additional attenuation rate of the resource area. S6. Extend the photovoltaic environmental degradation rate prediction model from the scale of some power plants to the scale of the expected area, and predict the photovoltaic environmental degradation rate of the entire expected area, i.e. the photovoltaic environmental degradation rate of the expected area. S7. The actual degradation rate of photovoltaic modules in the expected area is obtained by back-calculating the sum of the average theoretical degradation rate of photovoltaic modules in the expected area and the additional degradation rate of photovoltaic environment in the expected area.

2. The method for predicting the power generation attenuation rate of a photovoltaic power plant according to claim 1, characterized in that, Step S2 includes: Simulated theoretical zero-attenuation power generation of the photovoltaic power station under the target geographical location and corresponding meteorological conditions: ， In the above formula, : No. The photovoltaic power station mentioned above The theoretical annual power generation without degradation, in kilowatt-hours. and It is an integer; The initial nominal power of the photovoltaic modules used in the photovoltaic power station is expressed in kilowatts. : No. Total solar irradiance on the tilted surface of the sky, expressed in kilowatt-hours per square meter. The value is an integer, and the inclined surface is the inclined surface of the photovoltaic panel in the photovoltaic module relative to the ground; Reference irradiance under standard test conditions, in kilowatt-hours per square meter; The power temperature coefficient of the photovoltaic module is expressed as percentages per degree Celsius. : No. The average operating temperature of the photovoltaic modules described above is in degrees Celsius. The reference temperature under standard test conditions is taken as 25 degrees Celsius. System efficiency, including inverter efficiency, line loss, and transformer loss, is dimensionless. The total number of days in a year Calculate annual performance ratio: , In the above formula, : No. The photovoltaic power station mentioned above Annual performance ratio, dimensionless; : No. The photovoltaic power station mentioned above The actual measured annual power generation is expressed in kilowatt-hours. Calculate the actual annual decay rate: ， In the above formula, : No. The photovoltaic power station mentioned above The actual annual decay rate, expressed as a percentage, for the initial year, i.e. When =1, directly use the manufacturer's promised first-year degradation rate. , Calculate the environmentally-added degradation rate caused by the aforementioned environmental factors: ， In the above formula, : No. The photovoltaic power station mentioned above The environmental-added degradation rate for the year; : No. The theoretical attenuation rate under standard annual test conditions is calculated based on the manufacturer's promised linear attenuation model: ， In the above formula, This refers to the manufacturer's promised annual average degradation rate. The environmentally-added degradation rate refers to the additional power degradation of the photovoltaic module in actual operating environment due to environmental stress exceeding the laboratory standard test conditions.

3. The method for predicting the power generation attenuation rate of a photovoltaic power plant according to claim 2, characterized in that, Step S3 includes: According to the "Solar Energy Resource Assessment Method" (GB / T 37526-2019), the country is divided into four resource zones. Photovoltaic power stations are then regionally classified according to these four resource zones. The photovoltaic power station masks are extracted using a geographic information system platform. The four categories of resource zones include: Category I, the extremely abundant zone: areas with a GHR ≥ 1750 kWh / m²; Category II, the very abundant zone: areas with a GHR of 1400 ≤ GHR < 1750 kWh / m²; Category III, the abundant zone: areas with a GHR of 1050 ≤ GHR < 1400 kWh / m²; and Category IV, the general zone: areas with a GHR < 1050 kWh / m². The GHR refers to the total amount of solar radiation received at a horizontal surface on the Earth's surface.

4. The method for predicting the power generation attenuation rate of a photovoltaic power plant according to claim 3, characterized in that, The environmental factors are categorized as follows: Climate category, including elements: Average annual temperature, data from China Meteorological Administration website; Annual average land surface temperature, data from MODIS sensor land surface temperature or Landsat satellite land surface temperature; Annual average surface shortwave radiation, data from FY-4B and Himawari 8 / 9 satellites; Average annual total precipitation, data from ERA5 data and Global Precipitation Observation Program data; Annual precipitation days, data sourced from China Meteorological Administration Network and ERA5 data; Snowfall, snow accumulation, and hail; data from the China Meteorological Administration website. Dust storm data from the China Meteorological Administration website. Geographic category, including elements: Slope, aspect, and elevation data are derived from Space Shuttle radar topographic mapping mission data or digital elevation model data. The shadow of the mountain was calculated by a geographic information system platform using a digital elevation model. Land cover type or land use type, data sourced from the global 30-meter land cover dataset. Pollution categories, including elements: Annual average aerosol optical thickness, data from Sentinel-5P satellite and global land evapotranspiration dataset; Dust / dust, data from China Meteorological Administration website; Ammonia gas was used as a substitute, with the data calculated using the Euclidean distance from the pasture through a geographic information system platform. Bird droppings were replaced with data calculated using a geographic information system platform based on Euclidean distances from farmland, forests, and wetlands. The data obtained in step S4 includes raster data and non-raster data. The raster data includes the data of the environmental factors other than the numerical data obtained from the China Meteorological Administration Network. The non-raster data includes the numerical data of the environmental factors obtained from the China Meteorological Administration Network.

5. The method for predicting the power generation attenuation rate of a photovoltaic power station according to claim 4, characterized in that, Step S5 includes: S51. Data preprocessing, including the following steps: S511. In the geographic information system platform, the raster data is first geographically corrected to unify the spatial scale, and then the photovoltaic power station mask is used to extract the required feature values. S512, The feature data of the non-raster data is obtained by collecting and extracting the corresponding numerical data from the China Meteorological Administration Network based on the center coordinates of the photovoltaic power station. S513. Missing feature values ​​are filled with the corresponding feature values ​​of adjacent years of the same photovoltaic power station; environmental degradation rate data other than ±3 times the standard deviation are removed; the numerical feature values ​​are normalized to obtain the photovoltaic power station dataset. S52. Dataset partitioning, including the following steps: S521. Divide the photovoltaic power station dataset according to the four types of resource areas to ensure that the photovoltaic power station data in the same resource area are divided into the same photovoltaic power station dataset. S522. For each of the photovoltaic power station datasets obtained in step S521, divide the data into 60% of the years as the training set, 20% of the years as the validation set, and 20% of the years as the test set. S523. During model training, the time series data is sequentially divided into multiple consecutive time periods, and the validation set and test set are selected alternately multiple times in each time period to evaluate the model's performance in different time periods. S52. Model training optimization and result evaluation.

6. The method for predicting the power generation attenuation rate of a photovoltaic power plant according to claim 5, characterized in that, Step S6 includes: The expected region is assigned to a corresponding resource zone according to the division of the resource zone, and a machine learning model is trained for different resource zones. The photovoltaic device mask of the expected area is extracted from high-resolution remote sensing data using a semantic segmentation method. The photovoltaic device mask of the expected area refers to the spatial boundary or geographical range of all photovoltaic devices in an expected area. After replacing the photovoltaic power station mask with the photovoltaic equipment mask in the expected area, the time series data of environmental factors in the expected area are obtained according to step S4, and the data are calculated and processed into an annual average value, which includes the grid data and non-grid data. According to step S51, after preprocessing the annual average data of environmental factors in the expected area, the obtained feature values ​​are input into the machine learning model trained for the corresponding resource area to predict the photovoltaic environmental degradation rate of the expected area.

7. The method for predicting the power generation attenuation rate of a photovoltaic power plant according to claim 6, characterized in that, Step S7 includes calculation: ， In the above formula, : No. The expected actual photovoltaic degradation rate in the region for the year; : No. The expected regional photovoltaic environmental degradation rate for the year; : No. The average theoretical degradation rate of photovoltaic modules in the expected region for the year is the average of the degradation rates promised by different manufacturers.

8. The method for predicting the power generation attenuation rate of a photovoltaic power plant according to any one of claims 1-7, characterized in that, Step S1 includes: Obtain historical power generation data for each of the photovoltaic power stations, wherein the photovoltaic power stations are of the same model as the photovoltaic modules they use; Meteorological data is acquired through the photovoltaic environmental monitoring instruments equipped in the photovoltaic power station. The meteorological data includes continuous temperature, humidity, irradiance, precipitation, and wind speed of the power station environment. Obtain the technical parameters of the photovoltaic modules used in the photovoltaic power station. The technical parameters include: nominal power under standard test conditions, maximum conversion efficiency, power tolerance, first-year power degradation, and 2-30-year power degradation.

9. A computer device, characterized in that, The method includes a memory, a first processor, and a first computer program stored in the memory and executable on the first processor. When the first computer program is executed by the first processor, it implements the method for predicting the power generation attenuation rate of a photovoltaic power plant as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a second computer program, which can be executed by at least one second processor to enable the at least one second processor to perform the method for predicting the power generation attenuation rate of a photovoltaic power plant as described in any one of claims 1-8.

Images