Method, device and equipment for evaluating damage and loss of photovoltaic facility after typhoon

Abstract

The invention provides a photovoltaic facility damage and loss assessment method, device and equipment after typhoon, belongs to the technical field of computer information processing, and solves the problem of accurate quantitative assessment of photovoltaic facility damage and power generation loss caused by typhoon. The method comprises the following steps: acquiring original remote sensing image data and wind speed data of a target area where the photovoltaic facility is located; preprocessing the original remote sensing image data to obtain standard remote sensing image data; determining spectral indexes of the plurality of pixels according to the standard remote sensing image data; determining a feature vector set according to the standard remote sensing image data and the spectral indexes of the plurality of pixels; inputting the feature vector set into a classification model for processing to obtain distribution data of the photovoltaic facilities in the target area; and determining photovoltaic facility damage and loss evaluation data according to the wind speed data and the distribution data of the photovoltaic facilities in the target area. According to the scheme, high-precision quantitative evaluation of the physical damage of the typhoon to the photovoltaic facility and the economic loss of power generation is realized.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and in particular to a method, apparatus and equipment for assessing damage and loss of photovoltaic facilities after a typhoon. Background Technology

[0002] With the energy structure shifting towards a low-carbon model, photovoltaic solar power generation, as a clean and efficient renewable energy technology, has been widely adopted. However, against the backdrop of climate change, the frequency and intensity of typhoons are increasing, posing a serious threat to critical infrastructure. Due to their large-area open-air installation, photovoltaic facilities are highly vulnerable to extreme weather events such as typhoons, leading to panel cracking, structural damage, and even complete failure. This not only causes direct equipment asset losses but also triggers a chain of indirect economic impacts due to power generation interruptions. Therefore, accurately and efficiently assessing the physical damage to photovoltaic facilities caused by typhoons and the resulting energy production losses has become a key technical issue for ensuring energy security and improving the climate resilience of renewable energy equipment. Summary of the Invention

[0003] This invention provides a method, apparatus, and equipment for assessing damage and loss to photovoltaic facilities after a typhoon, enabling accurate quantitative assessment of damage to photovoltaic facilities and power generation losses caused by typhoons.

[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: This invention provides a method for assessing damage and loss to photovoltaic facilities after a typhoon, including: Acquire raw remote sensing image data and wind speed data containing the target area where the photovoltaic facilities are located. The raw remote sensing image data includes raw remote sensing image data before the typhoon and raw remote sensing image data after the typhoon. The original remote sensing image data is preprocessed to obtain standard remote sensing image data, which includes multiple pixels; Based on the standard remote sensing image data, determine the spectral indices of multiple pixels; Based on the standard remote sensing image data and the spectral indices of the multiple pixels, a feature vector set is determined; The feature vector set is input into a classification model for processing to obtain the distribution data of photovoltaic facilities in the target area. Based on the wind speed data and the distribution data of the photovoltaic facilities in the target area, the damage and loss assessment data of the photovoltaic facilities are determined.

[0005] Optionally, the original remote sensing image data is preprocessed to obtain standard remote sensing image data, including: Atmospheric and geometric corrections are performed on the original remote sensing image data to obtain corrected remote sensing image data. The corrected remote sensing image data is cropped according to a preset area size to obtain standard remote sensing image data.

[0006] Optionally, based on the standard remote sensing image data, the spectral indices of multiple pixels are determined, including: The standard remote sensing image data is extracted and processed to obtain reflectance values ​​of multiple bands for multiple pixels; Based on the reflectance values ​​of the multiple bands, the spectral indices of multiple pixels are determined. The spectral indices of each pixel include vegetation index, building index, first photovoltaic index, and second photovoltaic index.

[0007] Optionally, a feature vector set is determined based on the standard remote sensing image data and the spectral indices of the plurality of pixels, including: Based on the position of the pixels in the standard remote sensing image data, the spectral indices of the multiple pixels are integrated and sorted to obtain a feature vector set. The feature vector set includes multiple feature vectors, and each feature vector includes the spectral index of the pixel at the corresponding position in the standard remote sensing image data.

[0008] Optionally, the classification model is trained through the following process: Acquire historical photovoltaic distribution vector data, land classification data, and historical facies remote sensing image data for the target area; Based on the historical photovoltaic distribution vector data, positive sample data are determined; Based on the land classification data, determine the unmarked sample data; Based on the historical phase remote sensing image data, determine the historical feature vector set; The positive sample data, the unlabeled sample data, and the historical feature vector set are input into the random forest model for training to obtain the classification model.

[0009] Optionally, based on the wind speed data and the distribution data of the photovoltaic facilities in the target area, damage and loss assessment data for the photovoltaic facilities are determined, including: Based on the distribution data of the photovoltaic facilities in the target area, determine the damage and loss assessment data of the photovoltaic facilities; and / or, based on the distribution data of the photovoltaic facilities in the target area and the wind speed data, determine the relationship data between wind level and photovoltaic facility damage rate.

[0010] Optionally, the distribution data of the photovoltaic facilities in the target area includes: distribution data of photovoltaic facilities before the typhoon and distribution data of photovoltaic facilities after the typhoon. Based on the distribution data of the photovoltaic facilities in the target area, damage and loss assessment data of the photovoltaic facilities are determined, including: The damaged area was determined based on the distribution data of photovoltaic facilities before and after the typhoon. Based on the damaged area, determine the first loss data and installed capacity loss data of the photovoltaic facilities; Based on the installed capacity loss data, the second loss data for photovoltaic facilities is determined.

[0011] Optionally, based on the wind speed data and the photovoltaic facility distribution data, the relationship between wind level and photovoltaic facility damage rate is determined, including: The standard remote sensing image data is divided into raster regions according to preset values ​​to obtain multiple raster regions; Statistical analysis was performed on the wind speed data of the multiple grid areas to determine the maximum wind speed data of the multiple grids; The damaged areas of the multiple grid regions are statistically analyzed to determine the total damaged area of ​​the multiple grid regions; Linear fitting was performed on the maximum wind speed data of the multiple grids and the damaged area of ​​the multiple grids to obtain the relationship data between wind level and photovoltaic facility damage rate.

[0012] This invention also provides a device for assessing damage and loss to photovoltaic facilities after a typhoon, comprising: The acquisition module is used to acquire raw remote sensing image data and wind speed data of the target area where the photovoltaic facility is located. The raw remote sensing image data includes raw remote sensing image data before the typhoon and raw remote sensing image data after the typhoon. The processing module is used to preprocess the original remote sensing image data to obtain standard remote sensing image data, which includes multiple pixels; determine the spectral index of the multiple pixels based on the standard remote sensing image data; determine a feature vector set based on the standard remote sensing image data and the spectral index of the multiple pixels; and input the feature vector set into a classification model for processing to obtain the distribution data of photovoltaic facilities in the target area. The determination module is used to determine the damage and loss assessment data of the photovoltaic facilities based on the wind speed data and the distribution data of the photovoltaic facilities in the target area.

[0013] This invention also provides a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when run by the processor, executes the above-described method.

[0014] The technical solution of the present invention has at least the following effects: The above-described solution of the present invention acquires raw remote sensing image data and wind speed data containing the target area where photovoltaic facilities are located. The raw remote sensing image data includes raw remote sensing image data before and after the typhoon. The raw remote sensing image data is preprocessed to obtain standard remote sensing image data, which includes multiple pixels. Based on the standard remote sensing image data, the spectral indices of the multiple pixels are determined. Based on the standard remote sensing image data and the spectral indices of the multiple pixels, a feature vector set is determined. The feature vector set is input into a classification model for processing to obtain the distribution data of photovoltaic facilities in the target area. Based on the wind speed data and the distribution data of photovoltaic facilities in the target area, damage and loss assessment data for photovoltaic facilities are determined, achieving a high-precision quantitative assessment of the physical damage to photovoltaic facilities and the economic loss of power generation caused by typhoons. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method for assessing damage and loss to photovoltaic facilities after a typhoon, provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the technical route for the method of assessing damage and loss to photovoltaic facilities after a typhoon provided in an embodiment of the present invention; Figure 3 This is a structural diagram of the typhoon damage and loss assessment device for photovoltaic facilities provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of the computing device provided in an embodiment of the present invention. Detailed Implementation

[0016] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0017] like Figure 1 As shown, an embodiment of the present invention proposes a method for assessing damage and loss to photovoltaic facilities after a typhoon, including: Step 11: Obtain raw remote sensing image data and wind speed data containing the target area where the photovoltaic facility is located. The raw remote sensing image data includes raw remote sensing image data before the typhoon and raw remote sensing image data after the typhoon. Step 12: Preprocess the original remote sensing image data to obtain standard remote sensing image data, which includes multiple pixels; Step 13: Determine the spectral indices of multiple pixels based on the standard remote sensing image data; Step 14: Determine the feature vector set based on the standard remote sensing image data and the spectral indices of the multiple pixels; Step 15: Input the feature vector set into the classification model for processing to obtain the distribution data of photovoltaic facilities in the target area; Step 16: Based on the wind speed data and the distribution data of the photovoltaic facilities in the target area, determine the damage and loss assessment data of the photovoltaic facilities.

[0018] In step 11 of this embodiment, raw remote sensing image data before and after the typhoon are first collected. This data should cover the study area and include multispectral information such as visible light, near-infrared (NIR), and shortwave infrared (SWIR). Simultaneously, the maximum near-surface wind speed data within the study area during the typhoon is acquired for subsequent analysis of the relationship between wind speed and photovoltaic damage. The remote sensing image data can be obtained from multispectral satellite images acquired through platforms such as Earth Engine. Wind speed data can be obtained from typhoon wind field data provided by relevant databases.

[0019] In step 12, the acquired raw remote sensing image data before and after the typhoon are preprocessed, including radiometric correction, atmospheric correction, and geometric correction, to eliminate sensor errors, atmospheric interference, and geometric distortion, resulting in standardized remote sensing image data. The preprocessed data should contain multiple pixels, each representing a specific area of ​​the ground, with a clear spatial location and spectral information.

[0020] In step 13, using the preprocessed standard remote sensing image data, the spectral indices for each pixel are calculated, such as the vegetation index (NDVI), building index (NDBI), and a custom photovoltaic index (e.g., ...). and These spectral indices can reflect the spectral characteristics of ground features and help distinguish photovoltaic facilities from other ground features.

[0021] In step 14, the reflectance values ​​of each band in the preprocessed standard remote sensing image data are compared with the calculated spectral indices (such as NDVI, NDBI, etc.). and These parameters are combined to form the feature vector for each pixel. These feature vectors contain multi-dimensional information about the ground features and can be used for subsequent classification model training. The feature vector for each pixel includes: reflectance of each band of the multispectral satellite imagery (Sentinel-2), NDVI, NDBI, and so on. and wait.

[0022] In step 15, the feature vector set is classified to distinguish photovoltaic (PV) facilities from other ground features. By training a classification model, the distribution data of PV facilities before and after the typhoon is obtained, including information on whether each pixel belongs to a PV facility. The model is trained using positive samples (pixels with known PV facilities) and unlabeled samples. The trained model is then applied to remote sensing imagery before and after the typhoon to obtain the distribution data of PV facilities in the target area.

[0023] In step 16, by comparing the distribution data of photovoltaic (PV) facilities before and after the typhoon, the damaged PV facility areas are identified, and the damaged area is calculated. Direct economic losses are estimated based on the unit area cost of PV facilities. Simultaneously, the loss of power generation revenue due to outages is calculated based on the PV installed capacity and average daily power generation efficiency. Combining wind speed data during the typhoon and the distribution data of PV facilities in the target area, the damage to PV facilities under different wind levels is analyzed. The study area is divided into grids, the PV damage rate within each grid is calculated, and a linear function is used for fitting to obtain a model of the relationship between wind level and damage rate.

[0024] The technical solution described in this embodiment integrates multi-source remote sensing data, machine learning classifiers, and photovoltaic-specific spectral indices to achieve high-precision identification and change detection of photovoltaic facilities. Combined with wind speed field data, installed capacity estimation models, and economic value assessment models, it achieves multi-dimensional and accurate quantification of physical damage to photovoltaic facilities and energy production losses under typhoon disasters.

[0025] In an optional embodiment of the present invention, step 12, preprocessing the original remote sensing image data to obtain standard remote sensing image data, may include: Step 121: Perform atmospheric correction and geometric correction on the original remote sensing image data to obtain corrected remote sensing image data; Step 122: The corrected remote sensing image data is cropped according to a preset area size to obtain standard remote sensing image data.

[0026] In step 121 of this embodiment, the original remote sensing image data is affected by various factors such as atmospheric absorption, scattering, sensor attitude, and Earth curvature during the acquisition process, resulting in radiation distortion and geometric aberration. Therefore, atmospheric correction and geometric correction are required to eliminate these interferences and obtain more accurate ground reflection information.

[0027] Atmospheric correction aims to eliminate the influence of the atmosphere on solar radiation and ground-reflected radiation, restoring the true reflectivity of ground features. Commonly used atmospheric correction methods include physical model-based methods (such as the 6S model and the MODTRAN model) and empirical line-based methods. Atmospheric correction can eliminate the absorption of the spectrum by gases such as water vapor, carbon dioxide, and ozone, as well as the scattering of light by aerosols, thereby improving the radiometric quality of images.

[0028] The purpose of geometric correction is to eliminate image geometric distortions caused by factors such as sensor attitude, Earth curvature, and terrain undulations, ensuring that pixels in the image correspond to their actual locations on the ground. Commonly used geometric correction methods include polynomial correction and collinearity equation correction. Through geometric correction, it can be ensured that each pixel in the image accurately corresponds to its specific location in the geographic coordinate system, providing an accurate spatial basis for subsequent analysis and processing.

[0029] After atmospheric and geometric correction, the corrected remote sensing image data is obtained. These data are more accurate in terms of radiometrics and geometry, and can more realistically reflect the actual situation on the ground.

[0030] In step 122, the corrected remote sensing image data typically covers a large geographical area, while actual analysis may only require focusing on a specific region. Therefore, the image needs to be cropped according to a preset area size to obtain standard remote sensing image data that meets the analysis requirements. The boundaries of the cropping area are determined based on the extent and shape of the study area. This can be achieved using Geographic Information System (GIS) software or remote sensing image processing software. The cropped area should accurately cover the study area, avoiding omissions or the inclusion of unnecessary information. Image processing software or scripts are used to crop the corrected remote sensing image data. During the cropping process, the integrity and continuity of the image should be ensured to avoid edge effects or information loss caused by cropping.

[0031] After cropping, standard remote sensing image data was obtained. This data perfectly matches the study area in terms of spatial extent and has consistent radiometric and geometric characteristics, providing an accurate data foundation for subsequent extraction of photovoltaic facilities and damage assessment.

[0032] In an optional embodiment of the present invention, step 13, determining the spectral indices of multiple pixels based on the standard remote sensing image data, may include: Step 131: Extract and process the standard remote sensing image data to obtain reflectance values ​​of multiple bands for multiple pixels; Step 132: Determine the spectral index of multiple pixels based on the reflectance values ​​of the multiple bands. The spectral index of each pixel includes a vegetation index, a building index, a first photovoltaic index, and a second photovoltaic index.

[0033] In step 131 of this embodiment, standard remote sensing image data (such as Sentinel-2 imagery) contains multiple spectral bands (such as visible light, near-infrared, shortwave infrared, etc.), and each band records the intensity of light reflected from the Earth's surface at a specific wavelength. To analyze the spectral characteristics of photovoltaic facilities, the reflectance value of each pixel in each band needs to be extracted first.

[0034] (1) Band selection: Based on the spectral characteristics of the photovoltaic facility, select bands that are sensitive to vegetation, buildings, and photovoltaic modules. For example: Visible light bands (Blue, Green, Red): used to distinguish between vegetation and artificial features.

[0035] Near-infrared (NIR) band: Vegetation has high reflectivity in this band, while photovoltaic modules and buildings have low reflectivity.

[0036] Shortwave infrared (SWIR): Used to distinguish between dry and wet surfaces, and to help identify photovoltaic modules.

[0037] (2) Reflectance extraction: For each pixel of the standard remote sensing image, extract its reflectance value in the selected bands (such as B2-Blue, B3-Green, B4-Red, B8-NIR, B11 / B12-SWIR). These values ​​are expressed in dimensionless form (0-1) and reflect the surface's ability to reflect light.

[0038] Multiple band reflectance data for each pixel are obtained to form a feature vector (e.g., [B2, B3, B4, B8, B11, B12]), which provides a basis for subsequent spectral index calculations.

[0039] In step 132, the spectral index is a mathematical indicator constructed by combining reflectance values ​​from different bands, used to highlight the spectral characteristics of specific land features. The following spectral indices are calculated for photovoltaic facilities, vegetation, and buildings: (1) Vegetation Index (NDVI), the NDVI value ranges from -1 to 1. Vegetation-covered areas have higher NDVI values ​​(>0.5), while artificial features (such as photovoltaic modules) have lower NDVI values ​​(<0.2).

[0040] (2) Building Index (NDBI), the NDBI value ranges from -1 to 1. The NDBI value is higher in built areas (>0), and lower in vegetation and water bodies (<0).

[0041] (3) First Photovoltaic Index ( The calculation formula is: ; B11 is the SWIR1 band, and B9 is the near-infrared band. Photovoltaic modules have higher reflectivity in the SWIR band and lower reflectivity in the near-infrared band. This characteristic can be highlighted.

[0042] (4) Second photovoltaic index ( The calculation formula is: ; Among them, B12 is the SWIR2 band, and B8A is the narrow near-infrared band; It can further enhance the distinction between photovoltaic modules and surrounding objects, and is especially suitable for high-reflectivity photovoltaic panels.

[0043] For each pixel, the extracted band reflectance value is substituted into the above formula to calculate NDVI, NDBI, and NDBI. and Four spectral indices. These indices will serve as feature inputs to the classification model, distinguishing photovoltaic facilities from other ground features.

[0044] In an optional embodiment of the present invention, step 14, determining the feature vector set based on the standard remote sensing image data and the spectral indices of the plurality of pixels, may include: Step 141: Based on the position of the pixels in the standard remote sensing image data, the spectral indices of the multiple pixels are integrated and sorted to obtain a feature vector set. The feature vector set includes multiple feature vectors, and each feature vector includes the spectral index of the pixel at the corresponding position in the standard remote sensing image data.

[0045] In step 141 of this embodiment, the feature vector set is the core input to the subsequent classification model, and its construction must ensure that the spectral index of each pixel strictly corresponds to its spatial location. This step achieves the spatial integration of spectral indices through the following operations: (1) Spatial location matching: Each pixel in standard remote sensing image data has a unique row and column number (or geographic coordinates) corresponding to a specific location on the Earth's surface. The spectral indices (NDVI, NDBI, ...) obtained from step 132... and This requires matching each pixel's location to its corresponding pixel in the image. For example, the spectral index of pixel (row 100, column 200) is [NDVI=0.3, NDBI=0.2, ...]. , If ], then the initial form of the feature vector of this pixel is (100, 200, [0.3, 0.2, 0.7, 0.6]).

[0046] (2) Spectral index integration: Multiple spectral indices of each pixel are combined in a fixed order to form a one-dimensional feature vector. A typical order is: [NDVI, NDBI, ... , The order of priority is: vegetation index first, then building index, and finally photovoltaic-related index. For example, the feature vector of the above pixel is simplified to [0.3, 0.2, 0.7, 0.6].

[0047] (3) Sorting and Structuring: Sort the feature vectors of all pixels according to the row and column order of the image (e.g., from left to right, from top to bottom) to generate a regularized feature vector set. The feature vector set has a matrix structure, where: Row: corresponds to a pixel in the image (e.g., 1000 rows represent 1000 pixels). Column: Corresponding spectral index type (e.g., 4 columns represent 4 indices).

[0048] Example snippet: [0.3, 0.2, 0.7, 0.6] [0.4, 0.1, 0.6, 0.5] [0.2, 0.3, 0.8, 0.7] (4) Quality check: Verify the correspondence between feature vectors and image positions to avoid data misalignment due to coordinate offset or indexing errors.

[0049] Check if the spectral index range is reasonable (e.g., NDVI should be between -1 and 1), and eliminate outliers.

[0050] This embodiment can achieve the following: 1) Spatial consistency: Ensure that feature vectors are strictly bound to the location on the ground, providing spatial context for subsequent classification.

[0051] 2) Feature standardization: By combining spectral indices in a fixed order, the model can uniformly process image data from different regions.

[0052] 3) Efficiency optimization: The structured matrix form can accelerate the model training and inference process.

[0053] The final feature vector set is generated in the following form: V ={ v 1, v 2, ..., v n}, where the feature vector set V The i element v i =[NDVI i NDBI i NSPI 1i NSPI 2i ], NDVI i NDBI i NSPI1i NSPI 2i For the first i The spectral index of each pixel; Optionally, the classification model is trained through the following process: Step 151: Obtain historical photovoltaic distribution vector data, land classification data, and historical facies remote sensing image data for the target area; Step 152: Determine positive sample data based on the historical photovoltaic distribution vector data; Step 153: Determine unmarked sample data based on the land classification data; Step 154: Determine the historical feature vector set based on the historical phase remote sensing image data; Step 155: Input the positive sample data, the unlabeled sample data, and the historical feature vector set into the random forest model for training to obtain the classification model.

[0054] In step 151 of this embodiment, training the classification model requires three types of historical data as input: (1) Historical photovoltaic distribution vector data: that is, the vector map layer of photovoltaic power plant construction in historical periods (such as Shapefile format), which includes information such as the boundary coordinates, area and installation time of photovoltaic arrays. This data serves as a positive sample source for labeling photovoltaic facilities.

[0055] (2) Land classification data: that is, historical land use classification maps (such as the GLC-FCS300 dataset), which are labeled with land types such as vegetation, water bodies, buildings, and bare land. This data is used to filter unlabeled samples (i.e. areas that are not explicitly labeled as photovoltaic but may contain photovoltaics).

[0056] (3) Historical phase remote sensing image data: i.e., multispectral remote sensing images (such as Sentinel-2 or Landsat images) that are time-matched with historical photovoltaic distribution data. The spatial resolution and band settings of the images to be classified should be consistent with those of the current images to be classified to ensure feature transferability. This data is used to extract historical feature vector sets as input features for model training.

[0057] The above data is preprocessed, including the following steps: (1) Unify the coordinates of vector data and classification data to ensure spatial alignment with remote sensing images.

[0058] (2) Perform radiometric and atmospheric correction on remote sensing images to eliminate sensor errors and environmental interference.

[0059] In step 152, positive samples are pixels explicitly labeled as photovoltaic facilities, and their determination process is as follows: (1) Spatial intersection analysis: Overlay the historical photovoltaic distribution vector layer with the remote sensing image and extract all pixels within the vector boundary. For example, if the vector boundary of a photovoltaic power station covers pixels in rows 100-120 and columns 200-220 of the image, these pixels are marked as positive samples.

[0060] (2) Sample balance processing: If the number of positive samples is too small, the sampling range inside the photovoltaic array is expanded by using the sliding window method, or the adjacent pixels are appropriately expanded. If the number of positive samples is too large, random sampling is used to avoid class imbalance.

[0061] (3) Label assignment: For the selected pixel, mark the category label as "photovoltaic" (e.g., 1) and record its spatial coordinates.

[0062] The final positive sample dataset can be represented as: ; in, For positive sample datasets, ( x i , y i ) represents the pixel coordinates. N is the number of positive samples, and label is the class label.

[0063] In step 153, unlabeled samples are pixels that are not explicitly labeled as photovoltaic but may contain photovoltaic or background features. The determination process is as follows: (1) Exclude positive sample areas: Remove areas that overlap with historical photovoltaic distribution vectors from land classification data to avoid duplicate sampling.

[0064] (2) Land use classification screening: Prioritize land use areas that may be confused with photovoltaics as unlabeled samples, for example: 1) Building land: Rooftop photovoltaics may be classified as buildings; 2) Bare land / sandy land: Similar to the spectral characteristics of photovoltaic modules; 3) Farmland: Agricultural greenhouses may be mistakenly detected as photovoltaics.

[0065] (3) Random sampling: Randomly select pixels within the screened land use areas to ensure that unmarked samples cover diverse scenarios.

[0066] (4) Label processing: Unlabeled samples are not given a clear category label (or are labeled as “unknown”), and only their spatial coordinates and features are retained.

[0067] The final unlabeled sample dataset is generated in the following format: ; in, For pixel coordinates, MThe number of unlabeled samples, and M ≫ N , This is an unlabeled sample dataset; In step 154, the construction of the historical feature vector set is the same as in step 14, but based on historical remote sensing image data. The specific process is as follows: (1) Band reflectance extraction: For each pixel of the historical remote sensing image, extract the reflectance value of the multispectral bands (such as Blue, Green, Red, NIR, SWIR).

[0068] (2) Spectral index calculation: Calculate the spectral indices of corresponding pixels in positive and unlabeled samples, including: vegetation index (NDVI), building index (NDBI), and photovoltaic index (NDBI). , ).

[0069] (3) Feature vector integration: The spectral indices of each pixel are combined in a fixed order to form a feature vector.

[0070] In step 155, a semi-supervised random forest (SSRF) framework is used to train the model by combining labeled and unlabeled data: (1) Model initialization: Set the parameters for the random forest: number of trees (e.g., 500), maximum depth (e.g., 20), and feature subset size.

[0071] (2) Supervised learning portion: Initial training was performed using positive sample data (labeled 1) and negative samples (such as building and vegetation pixels, labeled 0) randomly selected from land classification data.

[0072] The number of negative samples needs to be balanced with the number of positive samples (e.g., a 1:1 ratio).

[0073] (3) Semi-supervised learning part: Introduce unlabeled samples and expand the training set through "self-training" or "label propagation" mechanisms: The model makes preliminary predictions on unlabeled samples and adds high-confidence predictions (such as probability > 0.9) to the training set.

[0074] Iteratively optimize the model parameters until convergence.

[0075] (4) Feature importance assessment: During training, the feature importance of each spectral index is calculated (e.g., based on the decrease in the Gini index), and the features that contribute the most to classification are selected.

[0076] The final trained classification model can be represented as: ; in, v Given a feature vector as input, the output is a predicted label for either photovoltaic (1) or non-photovoltaic (0).

[0077] In an optional embodiment of the present invention, step 16, determining the damage and loss assessment data of the photovoltaic facilities based on the wind speed data and the distribution data of the photovoltaic facilities in the target area, may include: Step 161: Determine the damaged area based on the pre-typhoon photovoltaic facility distribution data and the post-typhoon photovoltaic facility distribution data; Step 162: Determine the first loss data and the installed capacity loss data based on the damaged area; Step 163: Determine the second loss data based on the installed capacity loss data.

[0078] In step 161 of this embodiment, the damaged area is a core indicator for measuring the degree of physical damage to photovoltaic facilities, and its calculation relies on high-precision distribution data before and after the typhoon. The specific process is as follows: (1) Data alignment 1) Spatial alignment: Unify the vector data of photovoltaic facilities before and after the typhoon (such as Shapefile format) to the same coordinate system (such as WGS84) to ensure spatial comparability.

[0079] 2) Time matching: Confirm the time interval between the two periods of data to exclude changes in facilities caused by natural aging or non-typhoon factors.

[0080] (2) Damaged area identification Spatial difference analysis: Using GIS tools (such as ArcGIS), a "subtraction" operation is performed to identify areas of photovoltaic arrays that disappeared or were severely deformed after a typhoon. For example: Before the typhoon, a power station's coverage area was 1000m. 2 Only 300m left after the typhoon 2 The damaged area is 700m². 2 .

[0081] Morphological verification: Combined with the standard dimensions of photovoltaic modules (such as 1.6m×1.0m monocrystalline silicon panels), verify the geometric rationality of the damaged area and eliminate misjudgments (such as temporary shading).

[0082] (3) Classification and statistics of damaged area The damaged area is summarized by facility type (e.g., ground-mounted power station, rooftop distributed power station) or degree of damage (completely damaged, partially damaged).

[0083] In step 162, the damaged area needs to be further converted into economic indicators, including direct repair costs (first loss data) and power generation capacity loss (installed capacity loss data). The specific process is as follows: (1) Calculation of first loss data (direct repair cost) The first loss data refers to direct losses, specifically the equipment replacement or repair costs incurred due to physical damage to photovoltaic panels. Its calculation is based on the total damaged area obtained from the classification results, multiplied by the cost per unit area. The formula is: ; in, The total amount of direct economic losses. This represents the total area of ​​photovoltaic damage. C The cost of constructing or replacing photovoltaic facilities per unit area is a preset economic parameter. (2) The estimated installed capacity loss is as follows: ; in, This represents the installed capacity loss; GTI represents the total solar irradiance at the optimal tilt angle, which can be obtained from the global solar atlas. Module efficiency represents the energy conversion and system loss of photovoltaic modules, and can be preset to a fixed value of 0.15; PVOUT represents the photovoltaic power generation potential, which represents the long-term average daily power generation of a 1kWp photovoltaic system, and can be obtained through the global solar energy atlas; ILR represents the inverter load ratio, and can be preset to a fixed value of 0.8. In step 163, the second loss data is indirect economic loss, which refers to the loss of power generation revenue caused by damage or shutdown of photovoltaic facilities. It mainly reflects the long-term revenue loss caused by the loss of power generation capacity, and the calculation formula is: ; in, This represents the total indirect economic losses, i.e., the loss of electricity generation revenue. T This is the total downtime; The amount of power generation loss within a time unit t (usually a day) is calculated based on the lost installed capacity and the local average daily power generation efficiency. The on-grid electricity price is the price charged within a time unit t.

[0084] In an optional embodiment of the present invention, step 16, determining the damage and loss assessment data of the photovoltaic facilities based on the wind speed data and the distribution data of the photovoltaic facilities in the target area, may further include: Step 164: The standard remote sensing image data is divided into raster regions according to preset values ​​to obtain multiple raster regions; Step 165: Statistically analyze the wind speed data of the multiple grid areas to determine the maximum wind speed data of the multiple grids; Step 166: Calculate the damaged area of ​​the multiple grid regions to determine the damaged area of ​​multiple grids; Step 167: Perform linear fitting on the maximum wind speed data of the multiple grids and the damaged area of ​​the multiple grids to obtain the relationship data between wind level and photovoltaic facility damage rate.

[0085] In step 164 of this embodiment, grid partitioning discretizes continuous geographic space into regular grid cells to achieve spatial alignment between wind speed data and photovoltaic facility damage data. The specific process is as follows: (1) Grid size setting Based on the scale of the photovoltaic power station and the required research precision, the grid side length is preset (e.g., 500m×500m).

[0086] (2) Remote sensing image preprocessing Geometric correction: Eliminates distortions in remote sensing imagery (such as deformations caused by changes in satellite attitude) to ensure that the raster matches the actual geographic coordinates.

[0087] Projection conversion: Unify the image to Gauss-Kruger projection or Web Mercator projection to avoid spatial calculation errors.

[0088] (3) Raster generation and numbering Use GIS tools (such as ArcGIS's "Create Fishnet" tool) to generate a regular raster grid and assign a unique ID to each raster.

[0089] Through the above process, the entire target area is divided into a uniform 500m × 500m grid. This grid size is consistent with the original resolution of the wind speed data, ensuring that each grid is assigned a representative wind speed value.

[0090] In step 165, wind speed data is a core indicator reflecting typhoon intensity, and the maximum wind speed at the grid level needs to be obtained through multi-source data fusion and spatial interpolation. The specific process is as follows: (1) Wind speed data source: hourly wind speed records (such as 10-minute average wind speed) obtained from meteorological departments during typhoons from wind measurement towers.

[0091] (2) Extraction of maximum wind speed of grid: For each grid, the maximum wind speed during the typhoon period (e.g., 24 hours) is counted and used as the representative wind speed of the grid.

[0092] In step 166, the damaged area is a direct indicator of the extent of wind damage to photovoltaic facilities, and it needs to be obtained through multi-temporal remote sensing image comparison and grid-level statistics. The specific process is as follows: (1) Calculate the photovoltaic damage rate at the grid scale: Perform spatial overlay analysis on the damaged photovoltaic patches and the 500-meter grid created in the previous step. For each grid, calculate the ratio of the photovoltaic damage area inside it to the total photovoltaic area before the typhoon in that grid to obtain a grid layer with rich attributes, in which each grid contains its damage rate (a value between 0 and 1) and the corresponding maximum wind speed value.

[0093] (2) Calculate the damage rate at the scale of photovoltaic power station: realize the assessment from the perspective of facility management, and combine the scattered individual photovoltaic panel patches into a complete photovoltaic power station.

[0094] 1) Power plant identification: Based on the photovoltaic distribution map before the typhoon, the nearest neighbor merging rule is used (for example, merging photovoltaic patches within a certain distance threshold into the same power plant) to form independent photovoltaic power plant polygons.

[0095] 2) Damage calculation: For each photovoltaic power station polygon, calculate the ratio of its internal damaged area to the total area of ​​the power station to obtain the power station damage rate.

[0096] Obtaining a list of damage rates for each photovoltaic power station can be used to assess the severity of damage to different power stations.

[0097] In step 167, a quantitative relationship model between wind speed and damage rate is established through statistical regression analysis, providing a basis for wind disaster early warning and facility reinforcement. The specific process is as follows: (1) Wind speed to wind level: Convert the maximum wind speed value (unit: m / s) in each grid to the corresponding wind level (e.g., level 6 to level 13).

[0098] (2) Data matching: At this point, each 500-meter grid has two key attributes: wind level and photovoltaic damage rate.

[0099] (3) Linear regression analysis: with wind level as the independent variable ( X ), with photovoltaic damage rate as the dependent variable ( Y A linear model is fitted using the least squares method for all valid grid data: Y = aX + b ; in, a The regression coefficient is the slope of the fitted linear function, with a value of 0.03 (i.e., 3%), indicating that for every increase of one wind level, the average photovoltaic damage rate increases by 3%. b The regression intercept represents the theoretical zero-wind-speed damage rate, which is typically close to 0.

[0100] (4) Model validation: using the coefficient of determination R 2(0.63 in this example) is used to quantify the goodness of fit of the linear model, that is, to what extent changes in wind speed can explain changes in damage rate, R0. 2 A score >0.7 indicates high model reliability.

[0101] Residual analysis: Check whether the residuals are randomly distributed to avoid systematic biases (such as underestimating the damage rate in high wind speed areas).

[0102] like Figure 2 As shown in the embodiments of the present invention, a specific embodiment of the method for assessing damage and loss to photovoltaic facilities after a typhoon is as follows: Step 1, Data preparation and preprocessing; (1) Data Collection: Acquire Sentinel-2 multispectral images of the target area from a pre-set platform. Time periods are divided into: 1) Before the typhoon: First time T1 (no clouds or few clouds).

[0103] 2) After the typhoon: T2 as soon as possible (after the typhoon has passed, choose cloudless images if possible).

[0104] From the global photovoltaic dataset, 50 photovoltaic facility polygons within the target area are randomly selected as positive samples.

[0105] From the land cover data, stratified random sampling was conducted on five main land types (vegetation, farmland, artificial surface, water body, and bare land) within the target area, with 100 sample points drawn from each type, for a total of 500 points as unlabeled samples.

[0106] Download the GTI (tilted irradiance) and PVOUT (photovoltaic potential) raster data for the target area from the Global Solar Atlas.

[0107] Obtain near-surface maximum wind speed raster data (500-meter resolution) for the target area during a typhoon.

[0108] Collect information on photovoltaic construction costs (assumed to be RMB 200 / square meter) and feed-in tariffs (assumed to be RMB 0.4 / kWh) for the target area.

[0109] (2) Data preprocessing: Atmospheric correction and cloud masking were applied to the Sentinel-2 imagery, and the images were uniformly cropped to the boundary of the target area.

[0110] Calculate the feature vector for each pixel, including: reflectance values ​​for Sentinel-2 bands B2, B3, B4, B8, B11, and B12, as well as NDVI, NDBI, and two custom photovoltaic indices. , .

[0111] Step 2, Photovoltaic Extraction and Classification; (1) Model training: The 500 unlabeled samples and 50 positive samples were combined and randomly divided into a training set (412 samples) and a validation set (138 samples) in a 3:1 ratio.

[0112] Use the training set to train a PUL-RF classifier on a machine learning platform (such as Python's Scikit-learn library). Set the number of decision trees to 390.

[0113] Calculate the adjustment factor c using the validation set: Calculated c =0.25.

[0114] (2) Classification and application: The trained PUL-RF model was applied to the pre-processed feature images of the target area before and after the typhoon.

[0115] The model outputs a probability that each cell is labeled as a positive sample. Then through the formula Calibration yields the probability that each pixel truly belongs to a photovoltaic cell. .

[0116] To reduce randomness, this process is repeated 10 times (with resampling and training each time), and the probability values ​​of the 10 results are averaged to generate the final probability map.

[0117] (3) Binarization of results: Set the probability threshold to 0.5; classify pixels with a probability greater than or equal to 0.5 as "photovoltaic" and assign a value of 1; classify the rest as "non-photovoltaic" and assign a value of 0.

[0118] The final results are the photovoltaic distribution map of the target area before and after the typhoon (both are binary maps).

[0119] Step 3, accuracy evaluation; By visually interpreting high-resolution images, 20 real photovoltaic power station polygons were manually drawn in the target area as verification samples.

[0120] The post-typhoon photovoltaic distribution map extracted from the model is overlaid with the artificial sample for calculation: Calculate the crossover ratio : ; The calculation in this embodiment is IoU =84.5%.

[0121] Calculate the average accuracy: Plot the PR curve by varying the threshold and calculate the area under the curve. In this embodiment, the average accuracy is 90.1%.

[0122] Step 4, Loss Assessment; (1) Direct physical damage and economic loss: The photovoltaic distribution maps before and after the typhoon were overlaid to detect changes, and the total area of ​​pixels in the target area that changed from "photovoltaic" to "non-photovoltaic" was calculated. The damaged area was N = 0.72 km². 2 .

[0123] Calculate direct economic losses: L direct = N × C =720000×200=144000000 (yuan).

[0124] (2) Indirect power generation losses: 1) Estimate the lost installed capacity: ; Where area = 0.72km 2 GTI and PVOUT obtain the average value of the target region from the corresponding dataset (GTI = 1400 kWh / m). 2 PVOUT = 4.0 kWh / kWp η =0.15, IILR =0.8.

[0125] gencap=720000×1400×0.15÷4.0÷0.8=47.25 (MWp); 2) Calculate the loss of power generation revenue: The assessment period is set from the time of the typhoon until January 7, 2025, a total of 124 days.

[0126] Average daily power generation loss: ; Indirect losses: ; Step 5: Correlation analysis between wind level and damage rate The target area is divided into 500-meter grids.

[0127] Within each grid, calculate the photovoltaic damage rate (i.e., damaged area / total area before the typhoon).

[0128] The damage rate of each grid is correlated with the maximum wind level within it (converted from wind speed data).

[0129] Linear regression analysis was performed, and the wind level ( X ) and damage rate ( YThere is a significant positive correlation between them, and the fitted equation is: Y =0.03 X -0.1, R 2 =0.61.

[0130] Conclusion: In the target area, for every increase of one wind level, the damage rate of photovoltaic facilities increases by an average of 3%.

[0131] The method for assessing damage and loss of photovoltaic facilities after typhoons proposed in this invention integrates multi-source remote sensing data, machine learning classifiers, and photovoltaic-specific spectral indices for high-precision identification and change detection of photovoltaic facilities. By combining wind speed field data, installed capacity estimation models, and economic value assessment models, it achieves multi-dimensional and accurate quantification of physical damage to photovoltaic facilities under typhoon disasters. It can automatically identify the spatial distribution and damage of photovoltaic facilities before and after typhoons, establish a quantitative linear relationship between wind level and photovoltaic damage rate, provide dynamic monitoring capabilities for post-disaster recovery based on remote sensing and machine learning, and form a replicable integrated assessment framework, providing technical support for climate resilience planning of renewable energy facilities.

[0132] like Figure 3 As shown, this embodiment of the invention also provides a device 30 for assessing damage and loss to photovoltaic facilities after a typhoon, comprising: The acquisition module 31 is used to acquire raw remote sensing image data and wind speed data of the target area where the photovoltaic facility is located. The raw remote sensing image data includes raw remote sensing image data before the typhoon and raw remote sensing image data after the typhoon. Processing module 32 is used to preprocess the original remote sensing image data to obtain standard remote sensing image data, which includes multiple pixels; determine the spectral index of the multiple pixels based on the standard remote sensing image data; determine a feature vector set based on the standard remote sensing image data and the spectral index of the multiple pixels; and input the feature vector set into a classification model for processing to obtain the distribution data of photovoltaic facilities in the target area. The determination module 33 is used to determine the damage and loss assessment data of the photovoltaic facilities based on the wind speed data and the distribution data of the photovoltaic facilities in the target area.

[0133] Optionally, processing module 32 is specifically used for: Atmospheric and geometric corrections are performed on the original remote sensing image data to obtain corrected remote sensing image data. The corrected remote sensing image data is cropped according to a preset area size to obtain standard remote sensing image data.

[0134] Optionally, the processing module 32 is also specifically used for: The standard remote sensing image data is extracted and processed to obtain reflectance values ​​of multiple bands for multiple pixels; Based on the reflectance values ​​of the multiple bands, the spectral indices of multiple pixels are determined, and the spectral indices of each pixel include vegetation index, building index, first photovoltaic index and second photovoltaic index.

[0135] Optionally, the processing module 32 is also specifically used for: Based on the position of the pixels in the standard remote sensing image data, the spectral indices of the multiple pixels are integrated and sorted to obtain a feature vector set. The feature vector set includes multiple feature vectors, and each feature vector includes the spectral index of the pixel at the corresponding position in the standard remote sensing image data.

[0136] Optionally, the classification model is trained through the following process: Acquire historical photovoltaic distribution vector data, land classification data, and historical facies remote sensing image data for the target area; Based on the historical photovoltaic distribution vector data, positive sample data are determined; Based on the land classification data, determine the unmarked sample data; Based on the historical phase remote sensing image data, determine the historical feature vector set; The positive sample data, the unlabeled sample data, and the historical feature vector set are input into the random forest model for training to obtain the classification model.

[0137] Optionally, module 33 is specifically used for: Based on the distribution data of the photovoltaic facilities in the target area, determine the damage and loss assessment data of the photovoltaic facilities; and / or, based on the distribution data of the photovoltaic facilities in the target area and the wind speed data, determine the relationship data between wind level and photovoltaic facility damage rate.

[0138] Optionally, the distribution data of the photovoltaic facilities in the target area includes: distribution data of photovoltaic facilities before the typhoon and distribution data of photovoltaic facilities after the typhoon. Based on the distribution data of the photovoltaic facilities in the target area, damage and loss assessment data of the photovoltaic facilities are determined, including: The damaged area was determined based on the distribution data of photovoltaic facilities before and after the typhoon. Based on the damaged area, determine the first loss data and installed capacity loss data of the photovoltaic facilities; Based on the installed capacity loss data, the second loss data for photovoltaic facilities is determined.

[0139] Optionally, based on the wind speed data and the photovoltaic facility distribution data, the relationship between wind level and photovoltaic facility damage rate is determined, including: The standard remote sensing image data is divided into raster regions according to preset values ​​to obtain multiple raster regions; Statistical analysis was performed on the wind speed data of the multiple grid areas to determine the maximum wind speed data of the multiple grids; The damaged areas of the multiple grid regions are statistically analyzed to determine the total damaged area of ​​the multiple grid regions; Linear fitting was performed on the maximum wind speed data of the multiple grids and the damaged area of ​​the multiple grids to obtain the relationship data between wind level and photovoltaic facility damage rate.

[0140] It should be noted that this device is a device corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0141] like Figure 4 As shown, this embodiment of the invention also provides a computing device 40, including a processor 41, a memory 42, and a program or instructions stored in the memory 42 and executable on the processor 41. When the program or instructions are executed by the processor 41, they implement the various processes of the above-described embodiment of the method for assessing damage and loss to photovoltaic facilities after a typhoon, and achieve the same technical effect. To avoid repetition, they will not be described again here. It should be noted that the computing device in this embodiment of the invention includes the aforementioned mobile electronic devices and non-mobile electronic devices.

[0142] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0143] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0144] In the embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0145] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0146] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0147] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0148] Furthermore, it should be noted that in the apparatus and method of the present invention, it is obvious that the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent solutions of the present invention. Moreover, the steps performing the above series of processes can naturally be executed in the order described, but are not necessarily required to be executed in chronological order; some steps can be executed in parallel or independently of each other. Those skilled in the art will understand that all or any step or component of the method and apparatus of the present invention can be implemented in any computing device (including processors, storage media, etc.) or network of computing devices, in hardware, firmware, software, or a combination thereof. This is something that those skilled in the art can achieve by using their basic programming skills after reading the description of the present invention.

[0149] Therefore, the object of the present invention can also be achieved by running a program or a set of programs on any computing device. The computing device can be a known general-purpose device. Therefore, the object of the present invention can also be achieved simply by providing a program product containing program code for implementing the method or apparatus. That is, such a program product also constitutes the present invention, and the storage medium storing such a program product also constitutes the present invention. Obviously, the storage medium can be any known storage medium or any storage medium developed in the future. It should also be noted that in the apparatus and method of the present invention, it is obvious that the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent to the present invention. Furthermore, the steps for performing the above series of processes can naturally be performed in the order described, but are not necessarily required to be performed in chronological order. Some steps can be performed in parallel or independently of each other.

[0150] The above are preferred embodiments of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for assessing damage and loss to photovoltaic facilities after a typhoon, characterized in that, include: Acquire raw remote sensing image data and wind speed data containing the target area where the photovoltaic facilities are located. The raw remote sensing image data includes raw remote sensing image data before the typhoon and raw remote sensing image data after the typhoon. The original remote sensing image data is preprocessed to obtain standard remote sensing image data, which includes multiple pixels; Based on the standard remote sensing image data, determine the spectral indices of multiple pixels; Based on the standard remote sensing image data and the spectral indices of the multiple pixels, a feature vector set is determined; The feature vector set is input into a classification model for processing to obtain the distribution data of photovoltaic facilities in the target area. Based on the wind speed data and the distribution data of the photovoltaic facilities in the target area, the damage and loss assessment data of the photovoltaic facilities are determined.

2. The method for assessing damage and loss to photovoltaic facilities after a typhoon, as described in claim 1, is characterized in that... The original remote sensing image data is preprocessed to obtain standard remote sensing image data, including: Atmospheric and geometric corrections are performed on the original remote sensing image data to obtain corrected remote sensing image data. The corrected remote sensing image data is cropped according to a preset area size to obtain standard remote sensing image data.

3. The method for assessing damage and loss to photovoltaic facilities after a typhoon, as described in claim 1, is characterized in that... Based on the standard remote sensing image data, determine the spectral indices of multiple pixels, including: The standard remote sensing image data is extracted and processed to obtain reflectance values ​​of multiple bands for multiple pixels; Based on the reflectance values ​​of the multiple bands, the spectral indices of multiple pixels are determined. The spectral indices of each pixel include vegetation index, building index, first photovoltaic index, and second photovoltaic index.

4. The method for assessing damage and loss to photovoltaic facilities after a typhoon, as described in claim 1, is characterized in that... Based on the standard remote sensing image data and the spectral indices of the multiple pixels, a feature vector set is determined, including: Based on the position of the pixels in the standard remote sensing image data, the spectral indices of the multiple pixels are integrated and sorted to obtain a feature vector set. The feature vector set includes multiple feature vectors, and each feature vector includes the spectral index of the pixel at the corresponding position in the standard remote sensing image data.

5. The method for assessing damage and loss to photovoltaic facilities after a typhoon, as described in claim 1, is characterized in that... The classification model is trained through the following process: Acquire historical photovoltaic distribution vector data, land classification data, and historical facies remote sensing image data for the target area; Based on the historical photovoltaic distribution vector data, positive sample data are determined; Based on the land classification data, determine the unmarked sample data; Based on the historical phase remote sensing image data, determine the historical feature vector set; The positive sample data, the unlabeled sample data, and the historical feature vector set are input into the random forest model for training to obtain the classification model.

6. The method for assessing damage and loss to photovoltaic facilities after a typhoon, as described in claim 1, is characterized in that... Based on the wind speed data and the distribution data of the photovoltaic facilities in the target area, damage and loss assessment data for the photovoltaic facilities are determined, including: Based on the distribution data of the photovoltaic facilities in the target area, determine the damage and loss assessment data of the photovoltaic facilities; and / or, based on the distribution data of the photovoltaic facilities in the target area and the wind speed data, determine the relationship data between wind level and photovoltaic facility damage rate.

7. The method for assessing damage and loss to photovoltaic facilities after a typhoon, as described in claim 6, is characterized in that... The distribution data of the photovoltaic facilities in the target area includes: distribution data of photovoltaic facilities before the typhoon and distribution data of photovoltaic facilities after the typhoon. Based on the distribution data of the photovoltaic facilities in the target area, damage and loss assessment data of the photovoltaic facilities are determined, including: The damaged area was determined based on the distribution data of photovoltaic facilities before and after the typhoon. Based on the damaged area, determine the first loss data and installed capacity loss data of the photovoltaic facilities; Based on the installed capacity loss data, the second loss data for photovoltaic facilities is determined.

8. The method for assessing damage and loss to photovoltaic facilities after a typhoon, as described in claim 6, is characterized in that... Based on the wind speed data and the photovoltaic facility distribution data, the relationship between wind level and photovoltaic facility damage rate is determined, including: The standard remote sensing image data is divided into raster regions according to preset values ​​to obtain multiple raster regions; Statistical analysis was performed on the wind speed data of the multiple grid areas to determine the maximum wind speed data of the multiple grids; The damaged areas of the multiple grid regions are statistically analyzed to determine the total damaged area of ​​the multiple grid regions; Linear fitting was performed on the maximum wind speed data of the multiple grids and the damaged area of ​​the multiple grids to obtain the relationship data between wind level and photovoltaic facility damage rate.

9. A device for assessing damage and loss of photovoltaic facilities after a typhoon, characterized in that, include: The acquisition module is used to acquire raw remote sensing image data and wind speed data of the target area where the photovoltaic facility is located. The raw remote sensing image data includes raw remote sensing image data before the typhoon and raw remote sensing image data after the typhoon. The processing module is used to preprocess the original remote sensing image data to obtain standard remote sensing image data, which includes multiple pixels; determine the spectral index of the multiple pixels based on the standard remote sensing image data; determine a feature vector set based on the standard remote sensing image data and the spectral index of the multiple pixels; and input the feature vector set into a classification model for processing to obtain the distribution data of photovoltaic facilities in the target area. The determination module is used to determine the damage and loss assessment data of the photovoltaic facilities based on the wind speed data and the distribution data of the photovoltaic facilities in the target area.

10. A computing device, characterized in that, include: A processor, a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method as described in any one of claims 1 to 8.

Images