Carbon flux measurement and inversion method and system for Shagohuang New Energy Base

Abstract

This invention provides a method and system for carbon flux measurement and inversion in the Shagohuang new energy base. The method includes acquiring physical parameters, topographic data, and image data of the Shagohuang new energy base; establishing a three-dimensional geometric model using three-dimensional reconstruction technology; generating an LBM computational grid by spatially latticizing the model area; inputting the image data into a trained ground parameter model, outputting the ground boundary parameters and ground source term parameters for each grid cell; monitoring cloud information using an all-sky imager, and calculating the ground irradiance distribution using the Monte Carlo method in conjunction with a solar trajectory tracking model; acquiring atmospheric input boundary parameters and source term parameters; solving the LBM model to obtain the distribution of carbon dioxide concentration, velocity, temperature, and pollutant concentration; and calculating the carbon flux of the base based on the simulation results. This achieves high-precision, low-cost, and spatially continuous inversion calculation of carbon flux over a large area in the Shagohuang region.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring technology, and in particular to a method and system for measuring and retrieving carbon flux in the Shagohuang New Energy Base. Background Technology

[0002] For new energy bases built in special environments such as deserts, Gobi, and wastelands, carbon flux monitoring is of great significance for assessing the environmental impact of the project.

[0003] Traditional carbon flux monitoring methods mainly rely on direct measurements from ground-based flux towers. This approach faces several challenges in the desert region: First, the desert region is vast, and deploying flux towers across the entire area would result in excessively high construction costs; second, due to the complexity of the terrain and the difficulty of equipment maintenance, the density of flux towers is often insufficient, making it difficult to accurately reflect the spatial distribution characteristics of carbon flux over a large area and ensuring monitoring accuracy.

[0004] Therefore, how to reduce costs while ensuring monitoring accuracy has become a core technical problem that urgently needs to be solved. Summary of the Invention

[0005] This invention provides a carbon flux measurement and inversion method and system for the Shagohuang New Energy Base, which solves the defects of high carbon flux detection cost and low monitoring accuracy in the existing technology.

[0006] On the one hand, this invention provides a method for carbon flux measurement and inversion in a desert renewable energy base, which includes:

[0007] The physical parameters, terrain data, and image data of the Shagohuang New Energy Base were obtained, and a three-dimensional geometric model of the Shagohuang New Energy Base was established using three-dimensional reconstruction technology.

[0008] The region represented by the three-dimensional geometric model is spatially latticized to generate the computational grid required for the LBM model;

[0009] The image data is input into a pre-trained ground parameter model for identification, and the ground boundary parameters and ground source term parameters corresponding to each grid cell in the computational grid are output to set the ground boundary conditions and ground material source terms of the LBM model.

[0010] By monitoring cloud information with an all-sky imager and combining it with a high-precision solar trajectory tracking model, the Monte Carlo method is used to calculate the ground irradiance distribution under cloud cover conditions, so as to set the ground thermal radiation source term of the LBM model.

[0011] The input boundary parameters and input source term parameters of the atmosphere within the monitoring area are obtained in order to set the atmospheric boundary conditions and atmospheric material source terms of the LBM model.

[0012] Solve the LBM model to obtain detailed simulation results of carbon dioxide concentration distribution, flow velocity distribution, temperature distribution and pollutant concentration distribution in the Shagohuang New Energy Base, and calculate the carbon flux of the Shagohuang New Energy Base based on the detailed simulation results.

[0013] On the other hand, the present invention also provides a carbon flux measurement and inversion system for a desert renewable energy base, comprising:

[0014] The construction module is used to acquire the physical parameters, terrain data and image data of the Shagohuang New Energy Base, so as to establish a three-dimensional geometric model of the Shagohuang New Energy Base using three-dimensional reconstruction technology.

[0015] The generation module is used to perform spatial lattice processing on the region represented by the three-dimensional geometric model to generate the computational grid required for the LBM model;

[0016] The first setting module is used to input the image data into a pre-trained ground parameter model for identification, and output ground boundary parameters and ground source term parameters corresponding to each grid cell in the computational grid, so as to set the ground boundary conditions and ground material source terms of the LBM model.

[0017] The second setting module is used to monitor cloud information through an all-sky imager, combine it with a high-precision solar trajectory tracking model, and use the Monte Carlo method to calculate the ground irradiance distribution under cloud cover conditions, so as to set the ground thermal radiation source item of the LBM model.

[0018] The third setting module is used to obtain the input boundary parameters and input source term parameters of the atmosphere within the monitoring area, so as to set the atmospheric boundary conditions and atmospheric material source terms of the LBM model.

[0019] The solution module is used to solve the LBM model, obtain detailed simulation results of carbon dioxide concentration distribution, flow velocity distribution, temperature distribution and pollutant concentration distribution in the Shagohuang New Energy Base, and calculate the carbon flux of the Shagohuang New Energy Base based on the detailed simulation results.

[0020] The carbon flux measurement and inversion method and system for the desert-Gobi new energy base provided by this invention utilizes remote sensing imagery and 3D reconstruction technology to construct a detailed geometric model of the base. It automatically identifies surface characteristics and sets boundary conditions through a ground parameter model, and accurately simulates actual irradiance distribution by combining all-sky cloud monitoring and a Monte Carlo radiative transfer model. Finally, it solves the multi-physics coupled atmospheric transport process using LBM, outputting high-resolution spatial distribution results of carbon dioxide concentration, velocity, temperature, and pollutants. This achieves high-precision, low-cost, and spatially continuous inversion calculation of large-scale carbon flux in the desert-Gobi region. This invention effectively overcomes the limitations of traditional flux tower deployment, such as high cost and poor spatial representativeness, and significantly improves the accuracy and economy of carbon flux monitoring in the desert-Gobi environment. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the carbon flux measurement and inversion method for the Shagohuang New Energy Base provided in this embodiment of the invention.

[0023] Figure 2 This is a schematic diagram of the carbon flux metering and inversion system for the Shagohuang New Energy Base provided in this embodiment of the invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] In existing technologies, carbon flux monitoring in new energy bases built in special environments such as deserts, Gobi, and wastelands mainly relies on direct measurements from ground-based flux towers. These areas are vast, and deploying flux towers across the board results in excessively high construction costs. The complexity of the terrain and the difficulty of equipment maintenance limit the density of flux tower deployment, making it difficult to accurately reflect the spatial distribution characteristics of carbon flux over large areas, thus limiting monitoring accuracy. Traditional methods cannot balance cost and accuracy, necessitating an efficient and economical solution.

[0026] To address these issues, research revealed the need to move beyond reliance on purely physical monitoring equipment and instead combine numerical simulations with limited measured data. Analysis showed that carbon flux in the desert region is influenced by a combination of factors, including surface features, atmospheric flow, and thermal radiation, necessitating the construction of a high-precision three-dimensional model to reflect this complex environment. Further consideration was given to using the Lattice Boltzmann Method (LBM) to simulate flow fields and mass transport, dynamically adjusting model parameters using remote sensing data, ultimately forming a low-cost monitoring system based on simulation inversion.

[0027] Therefore, the present invention proposes the following technical solution:

[0028] Figure 1 This is a flowchart illustrating the carbon flux measurement and inversion method for the Shagohuang New Energy Base provided in this embodiment of the invention.

[0029] like Figure 1 As shown in the embodiment of the present invention, the execution subject of the carbon flux measurement and inversion method for the Shagohuang new energy base can be an electronic device, and the method mainly includes the following steps:

[0030] 101. Obtain the physical parameters, terrain data, and image data of the Shagohuang New Energy Base, and use 3D reconstruction technology to establish a 3D geometric model of the Shagohuang New Energy Base;

[0031] 102. Perform spatial lattice processing on the region represented by the three-dimensional geometric model to generate the computational grid required for the LBM model;

[0032] 103. Input the image data into a pre-trained ground parameter model for identification, and output the ground boundary parameters and ground source term parameters corresponding to each grid cell in the computational grid, so as to set the ground boundary conditions and ground material source terms of the LBM model.

[0033] 104. By monitoring cloud information through an all-sky imager and combining it with a high-precision solar trajectory tracking model, the Monte Carlo method is used to calculate the ground irradiance distribution under cloud cover conditions, so as to set the ground thermal radiation source term of the LBM model.

[0034] 105. Obtain the input boundary parameters and input source term parameters of the atmosphere within the monitoring area to set the atmospheric boundary conditions and atmospheric material source terms of the LBM model;

[0035] 106. Solve the LBM model to obtain detailed simulation results of carbon dioxide concentration distribution, flow velocity distribution, temperature distribution and pollutant concentration distribution in the Shagohuang New Energy Base, and calculate the carbon flux of the Shagohuang New Energy Base based on the detailed simulation results.

[0036] In a specific implementation process, the 3D geometric model refers to a 3D digital model that integrates terrain elevation, facility geometry, and surface cover. Specifically, it can be generated by registering UAV aerial point cloud data with satellite multispectral imagery and then using a Poisson surface reconstruction algorithm. This model is used to accurately characterize surface features such as wind turbine blade height and photovoltaic panel tilt angle. The LBM computational grid refers to a grid system that discretizes the simulation area into regular cubic units. Specifically, an octree adaptive gridding method can be used to densify the grid to meter-level resolution in densely populated areas of facilities, ensuring the accuracy of wake vortex capture. The ground parameter model refers to a deep learning-based multi-label classification model. Specifically, it can be trained using the ResNet-50 architecture. The input is a slice of remote sensing imagery or UAV aerial imagery matching the grid size, and the output is the ground roughness, thermal conductivity, and material source parameters for each grid, solving the problem of low efficiency in traditional manual calibration. Monte Carlo ray tracing algorithm refers to a probabilistic model that simulates the interaction between sunlight and clouds. Specifically, it can use the Buie solar intensity distribution model combined with cloud transmittance inversion data to calculate the irradiance of the shadow area through millions of light samples, thus eliminating the interference of dynamic cloud shading on the calculation of thermal radiation.

[0037] Specifically, the process begins by acquiring centimeter-level terrain point cloud data using a drone equipped with a lidar system. This data, combined with Sentinel-2 satellite imagery, is used to extract land cover types and construct a 3D model incorporating photovoltaic panel tilt angles and wind turbine tower heights. The simulated area is divided into adjustable-length cubic grids, with a 0.5-meter finer grid used at the photovoltaic array edges to capture turbulent boundary layer variations. Remote sensing image slices corresponding to each grid are input into a pre-trained ResNet-50 model, automatically identifying the area proportions of materials such as rock, vegetation, and concrete within the grid, and weighted to calculate the roughness parameters for each grid. Cloud distribution maps are acquired every 5 minutes using an all-sky imager, and the ratio of direct to diffuse radiation on the photovoltaic panel surface is calculated in real-time using the solar altitude angle. A Monte Carlo simulation of 100,000 light paths is employed, and the effective radiation energy received by each grid is statistically analyzed as a heat source term. Finally, atmospheric boundary conditions are coupled to solve the LBM equations, outputting hourly updated carbon dioxide concentration fields and 3D flow field data, providing spatially continuous input parameters for carbon flux calculation.

[0038] This scheme achieves continuous spatiotemporal simulation of the entire region using the LBM model, and quantifies the actual impact of cloud morphology on irradiance through ray tracing algorithms. Furthermore, this scheme improves parameter identification efficiency through the fusion processing of UAV and satellite imagery.

[0039] Through the above technical solutions, this invention achieves high-precision inversion of carbon flux in desert environments, requiring only a small number of flux towers for model validation, significantly reducing monitoring costs. The refined construction of the 3D model effectively captures the impact of renewable energy facilities on the local microenvironment; Monte Carlo irradiance calculation improves simulation accuracy under cloudy weather conditions; and automated parameter identification through multi-source data fusion ensures the spatiotemporal consistency of model input, ultimately effectively reducing global carbon flux calculation errors. In other words, this invention effectively overcomes the limitations of traditional flux tower deployment, such as high cost and poor spatial representativeness, significantly improving the accuracy and economy of carbon flux monitoring in desert environments.

[0040] In some embodiments, the present invention further proposes a method for calculating the carbon flux of the Shagohuang New Energy Base based on detailed simulation results, comprising: dividing the Shagohuang New Energy Base into multiple partition types; for each partition type, calculating a first carbon flux using the gradient diffusion method based on the carbon dioxide concentration distribution and the flow velocity distribution, and calculating a second carbon flux using the eddy covariance method; performing a weighted summation of the first carbon flux and the second carbon flux to obtain the basic carbon flux corresponding to each partition type; correcting the basic carbon flux based on the correction coefficient of each partition type to obtain the corrected carbon flux of each partition type; performing an area-weighted summation of the corrected carbon fluxes of all partition types, and dividing the sum by the total area to obtain the carbon flux of the Shagohuang New Energy Base.

[0041] In a specific implementation process, the zoning type refers to different functional areas divided according to the terrain structure, vegetation coverage and the distribution characteristics of new energy facilities. It can be realized by comprehensively judging multiple sources of data, such as digital elevation model data, multispectral remote sensing images, hyperspectral remote sensing images, wind turbine coordinates, wind resource meteorological data and equipment deployment drawings.

[0042] Specifically, this embodiment provides a method for calculating terrain complexity indicators based on a Digital Elevation Model (DEM) and applying them to zoning discrimination. The DEM data can be derived from spaceborne radar or LiDAR measurements, with a preferred spatial resolution of 5 to 30 meters.

[0043] Specifically, the digital elevation model data can be preprocessed to calculate the terrain complexity index of each computational grid. Combined with a preset terrain threshold, the computational grids can be initially partitioned to obtain a first partition result. This first partition result is then used to perform a comprehensive logical judgment on each computational grid, along with partition results obtained by other methods, to determine the final partition type of each computational grid. The specific implementation process is as follows:

[0044] S11, Data Preprocessing:

[0045] Obtain the DEM data of the Shagohuang New Energy Base. First, preprocess the DEM data to eliminate data holes and outliers. The preprocessing includes smoothing and denoising using Gaussian filtering or median filtering, and filling invalid data areas using natural neighbor interpolation or Kriging interpolation.

[0046] S12, Calculate the slope value: Calculate the slope value within the specified first moving window using the maximum elevation difference method or the Horn algorithm.

[0047] Specifically, based on the preprocessed DEM data, the maximum elevation difference method or the Horn algorithm can be used to calculate the slope value pixel by pixel within a specified moving window (e.g., a 3×3 pixel window), with the unit being degrees or percentages. The slope is used to characterize the inclination of the land surface.

[0048] Calculation formula (taking the maximum elevation difference method as an example):

[0049]

[0050] in, Indicates the slope value. The elevation change rate in the east-west direction. This represents the rate of change of elevation in the north-south direction.

[0051] Slope value partitioning application: The slope values ​​of the computational grid are compared with preset thresholds to obtain partitions. For example, partitioning rules can include, but are not limited to, the following rules, which can be set according to actual needs:

[0052] If the average slope is less than 5°, the terrain of the area is flat and suitable for designation as a photovoltaic power generation zone.

[0053] If the average slope is between 5° and 15°, the area needs to be further assessed in conjunction with other indicators.

[0054] If the average slope is greater than 15°, the terrain is steep, and the construction and maintenance costs are high. It should be prioritized to be designated as a primitive desert area or ecological protection area and excluded from the core construction area of ​​the photovoltaic and wind power generation areas.

[0055] S13. Calculate Aspect Value: Calculate the aspect value within the first moving window using the preset aspect calculation formula.

[0056] Specifically, the aspect value can be calculated based on the preprocessed DEM data using the same moving window used when calculating the slope value, in degrees, from 0° to 360° (normally due north is 0°).

[0057] The formula for calculating slope aspect is as follows:

[0058]

[0059] in, This represents the slope value. The application of slope values ​​in different zones: Slope aspect is primarily used to analyze solar radiation received.

[0060] The southern slope (e.g., 135° to 225°) receives the strongest solar radiation in the mid-to-high latitude regions of the hemisphere, which has a positive impact on photovoltaic power generation efficiency and can be used as a preferred factor for photovoltaic zones in zoning.

[0061] The north slope receives less radiation, which may result in snow accumulation or differences in vegetation.

[0062] S14. Calculate Relief Amplitude: Calculate the terrain relief within the specified second moving window using a preset relief calculation formula.

[0063] Topographic relief is an index describing the difference in extreme elevation values ​​within a unit, effectively reflecting the macroscopic roughness of the terrain. Its calculation needs to be performed within a larger analysis window (e.g., 500m × 500m).

[0064] The formula for calculating undulation is:

[0065]

[0066] in, To calculate the terrain relief, and These are the maximum and minimum elevation values ​​within the analysis window, respectively.

[0067] Application of calculating terrain relief in regional divisions:

[0068] If the terrain undulation is less than 30 meters, it indicates that the terrain is flat or gentle hills, which is suitable for the construction of large-scale new energy facilities.

[0069] If the terrain undulation is between 30 and 100 meters, it indicates that the terrain is hilly, and the layout of facilities needs to be carefully assessed. It may be classified as a mixed-use area.

[0070] If the terrain relief is greater than 100 meters, it indicates that the terrain is mountainous or deeply dissected, and in principle, it should be classified as a primitive desert area or a restricted development area, because it has a complex impact on the wind field and is extremely difficult to construct.

[0071] S15. Comprehensive Judgment:

[0072] For each computational grid, the first partition result obtained from its slope, aspect, and topographic relief indices is overlaid and fused with its own data, such as image spectral data (e.g., Normalized Difference Vegetation Index (NDVI)), equipment deployment information (e.g., wind turbine foundation locations extracted from construction drawings), or other partition results. For example, different partition results can be prioritized, and corresponding fusion strategies can be set for different priorities. These fusion strategies can be configured according to actual needs. For example:

[0073] If a unit has a flat slope (<5°), small undulation (<30m), extremely low NDVI value, and is located within the planning area of ​​a photovoltaic plant, it will be ultimately determined as a photovoltaic power generation area.

[0074] If a unit has a steep slope (>10°) and large undulation (>50m), but NDVI shows natural vegetation cover, it is ultimately determined to be a primitive desert area.

[0075] If a unit has no special restrictions on its topographic indicators, but the hyperspectral imagery identifies concrete foundation features and the construction drawings indicate it as a substation, then it is ultimately determined to be an electrical facility area.

[0076] The above example is for illustrative purposes only. This example does not limit other fusion strategies, and they will not be illustrated one by one here.

[0077] Through the above embodiments, the method realizes the quantitative calculation of terrain complexity index and automated zoning discrimination, effectively improving the objectivity, accuracy and efficiency of the zoning of the Shagohuang New Energy Base.

[0078] In one specific implementation, an alternative approach is to calculate NDVI based on multispectral remote sensing imagery and classify vegetation cover levels according to preset thresholds. The multispectral remote sensing imagery data source can be from Sentinel or higher resolution satellite imagery, and must include at least the red band (approximately 0.64-0.67 μm) and the near-infrared band (approximately 0.85-0.88 μm).

[0079] Specifically, the multispectral remote sensing image can be radiometrically calibrated and atmospherically corrected, NDVI can be calculated pixel by pixel, and vegetation cover levels can be divided according to a preset vegetation index threshold. The vegetation cover levels are then used to perform preliminary partitioning of the computational grid, resulting in a second partitioning result. This result is then combined with partitioning results obtained by other methods to perform a comprehensive logical judgment on each computational grid, ultimately determining the partition type to which each computational grid belongs. The specific implementation process is as follows:

[0080] S21. Data Preprocessing and NDVI Calculation:

[0081] Acquire multispectral remote sensing images covering the Shagohuang New Energy Base. First, perform radiometric calibration and atmospheric correction on the images (using models such as FLASSH and 6S) to eliminate interference from atmospheric aerosols and water vapor on the reflectance of ground features, obtaining true surface reflectance data. Then, calculate the normalized vegetation index (NVI) value using a pre-defined formula:

[0082] ;

[0083] in, The reflectivity is in the near-infrared band. denoted as the reflectivity in the red light band. The theoretical range of NDVI is [-1, 1].

[0084] S22. Determine the vegetation cover level threshold to determine the vegetation cover level and the second partition result corresponding to the vegetation cover level based on the range of the normalized vegetation index value:

[0085] Due to the generally sparse vegetation cover and unique environment of the Shago desert region, its NDVI threshold range differs from that of typical densely vegetated areas. Based on statistical analysis of historical field survey samples and remote sensing imagery, the following vegetation cover level discrimination thresholds were set for this base:

[0086] Bare Soil / Sand Area: NDVI ≤ 0.1;

[0087] This area mainly consists of exposed sand, Gobi desert, rocks, or construction sites. The NDVI value is low, and the near-infrared reflectance is close to the red light reflectance.

[0088] Sparse vegetation area: 0.1 < NDVI ≤ 0.2;

[0089] The area is mainly composed of sparse desert herbs, mosses, or seedlings in their early growth stages. The vegetation canopy coverage is typically less than 15%.

[0090] Moderate vegetation cover: 0.2 < NDVI ≤ 0.4;

[0091] The area mainly consists of well-grown shrubs, artificially planted sand-fixing grass grids, or natural grasslands with high coverage. The vegetation canopy coverage is typically between 15% and 50%.

[0092] Dense vegetation area: NDVI > 0.4;

[0093] The area mainly consists of artificially maintained green belts, protective forests, or rare natural oases. The vegetation canopy coverage exceeds 50%.

[0094] Optional, S23, Post-processing and optimization:

[0095] Post-processing is performed on the initially calculated NDVI raster layer to eliminate noise and outliers. This can be achieved by using various filtering or morphological opening / closing operations to smooth the data, such as using a 3×3 or 5×5 pixel moving window, to eliminate misjudgments caused by individual pixel anomalies.

[0096] S24. Comprehensive Judgment and Partition Application:

[0097] The calculated vegetation cover level map is overlaid and integrated with the results of other zoning criteria (topography complexity, facility type, etc.) to participate in the final zoning type decision. For example:

[0098] If a computational grid is identified as a non-vegetation area, has flat terrain (slope < 5°), and is located within a photovoltaic planning area, then its final zoning type is a photovoltaic power generation area.

[0099] If a computational grid is identified as a high vegetation cover area, it should be given priority for classification as an artificial greening area, regardless of its topography.

[0100] If a computational grid is identified as a low vegetation cover area with a topographic relief greater than 50 meters and no planned facilities, its final zoning type is primitive desert area.

[0101] If a computational grid is identified as a medium vegetation cover zone, but a booster station is built on it (identified from the construction drawings), then its final zoning type is an electrical facility zone, and the vegetation on it is considered as part of the greening within the facility zone.

[0102] Through the above embodiments, the method achieves objective and quantitative classification of vegetation cover under the special environment of desert and Gobi, providing key data support for ecological assessment of new energy bases, monitoring of greening project effects, and zoning planning.

[0103] In a specific implementation process, the hyperspectral remote sensing image can be radiometrically calibrated and atmospherically corrected. Characteristic high-reflectance peaks of photovoltaic panels in the near-infrared band can be extracted, and feature enhancement can be applied to these peaks. Photovoltaic panel pixels can be identified using a preset reflection peak threshold, and post-processing can be performed to form a preliminary range of the photovoltaic power generation area. The computational grid can then be preliminarily partitioned to obtain a third partition result. This result is used to combine with partition results obtained by other methods to perform a comprehensive logical judgment on each computational grid, ultimately determining the partition type to which each grid belongs. The hyperspectral remote sensing data source is preferably a sensor with high spectral resolution, and its band setting needs to cover at least the near-infrared region of 760 nm to 900 nm. The specific implementation process is as follows:

[0104] S31. Data Preprocessing and Spectral Library Construction:

[0105] Acquire a hyperspectral image cube covering the Shagohuang New Energy Base. First, perform radiometric calibration and atmospheric correction on the data (using advanced atmospheric correction models such as MODTRAN and FLAASH) to accurately reconstruct the surface reflectance spectral curves of the ground features.

[0106] Simultaneously, a standard spectral feature library for photovoltaic panels will be constructed. Typical reflectance spectral curves of photovoltaic panels of different materials, with different ages, and different installation angles will be obtained through field measurements (using ground-based spectrometers such as ASD) or from validated public / commercial spectral libraries, serving as reference benchmarks for identification.

[0107] S32. Feature Reflection Peak Extraction and Enhancement:

[0108] Due to the anti-reflection coating (AR coating) covering its surface and the electronic properties of semiconductor materials (such as silicon), photovoltaic panels form a very significant and stable high reflection peak in the near-infrared band (especially around 850nm). This is the core spectral fingerprint that distinguishes them from other natural features (such as soil, vegetation, and rocks).

[0109] To enhance this feature, the following operations are performed on the spectral curve of each pixel in the image:

[0110] 1. Calculate the first derivative spectrum: to eliminate the influence of background noise and differences in lighting conditions, and to accurately locate the wavelength position of the reflection peak (the point where the derivative value is zero corresponds to the peak position of the reflection peak).

[0111] 2. Calculate the normalization index (optional): An index similar to (R850 - R700) / (R850 + R700) can be constructed, where R850 and R700 represent the reflectance values ​​of the 850nm and 700nm bands, respectively, to quantify the intensity of the reflection peak.

[0112] S33. Setting the discrimination threshold and identification:

[0113] Based on the aforementioned standard spectral feature library and statistical analysis of a large number of samples, the following discrimination thresholds are set for this base:

[0114] Reflection peak intensity threshold: The reflectance value at the 850nm band is greater than 0.35.

[0115] Reflection peak morphology threshold: There must be a continuous and prominent reflection peak in the wavelength range of 760nm to 900nm, with a full width at half maximum (FWHM) within a certain range (e.g., 30-50 nm), and the spectral curve has no obvious absorption valley in this range (to distinguish it from the "red edge" feature of vegetation).

[0116] Additional exclusion criteria: The NDVI value of the pixel should be less than 0.2 to exclude the possibility of dense vegetation; at the same time, its blue light reflectance should not be too high to exclude water bodies.

[0117] The spectral curve of each pixel is compared with the aforementioned thresholds. If both the reflection peak intensity threshold and the reflection peak shape threshold are met, the pixel is preliminarily determined to be a photovoltaic panel pixel.

[0118] S34. Post-processing and object optimization:

[0119] The initially identified scattered pixels are then post-processed to form a complete photovoltaic power generation area:

[0120] 1. Morphological processing: Cluster analysis and morphological closing operation (dilation followed by erosion) are used to aggregate nearby scattered pixels into complete contiguous areas, fill internal pores, and smooth the boundaries.

[0121] 2. Spatial Constraints: Based on the photovoltaic array planning boundary extracted from the base construction drawings, the identification results are spatially constrained and verified to eliminate misidentified spots that are obviously located outside the planning area.

[0122] 3. Logical verification: If an area is identified as a photovoltaic panel, but its average slope is greater than 15°, it needs to be confirmed by manual visual interpretation to rule out misjudgment caused by steep bare rock.

[0123] S35. Comprehensive Judgment and Partition Confirmation:

[0124] The photovoltaic panel coverage areas identified through the aforementioned hyperspectral features are directly assigned the zoning type "Photovoltaic Power Generation Zone." This is one of the high-priority discrimination criteria. The boundaries of these zones are then overlaid with zoning results based on other criteria (such as topography and NDVI) to ensure the logical consistency of the zoning map.

[0125] In a specific implementation process, to achieve precise quantification and spatial division of the wake effect influence range of wind turbines within the desert-based new energy base, this embodiment provides a wake area delineation method that couples wind turbine geographical location, fluid dynamics model, and Geographic Information System (GIS) spatial analysis technology. This method not only considers geometric distance but also incorporates wind resource distribution characteristics and unique environmental factors of the desert, thereby dynamically and accurately defining the wake influence range.

[0126] Specifically, using the wind turbine coordinates, wind resource meteorological data, and digital elevation model data, a customized wake model combined with a wind speed deficit threshold can be employed to calculate the wind turbine wake influence range. This allows for preliminary partitioning of the computational grid, yielding a fourth partition result. This result is then used to perform a comprehensive logical judgment on each computational grid, combining it with partitioning results obtained from other methods, to determine the final partition type for each grid. The specific implementation process is as follows:

[0127] S41. Data Preparation and Preprocessing:

[0128] Obtain the following data from the Shagohuang New Energy Base:

[0129] Wind turbine coordinates: Extract the precise geographical coordinates (latitude and longitude), hub height, and rotor diameter of each wind turbine from the construction drawings, as-built drawings, or on-site GPS measurements provided by the wind turbine manufacturer.

[0130] Meteorological data for wind resources: Acquire long-term wind observation data for the base and analyze the prevailing wind direction (e.g., prevailing northwesterly wind) and wind speed frequency distribution. In particular, it is necessary to obtain parameters characterizing atmospheric stability, such as the Richardson number (Ri) calculated from temperature profiles or turbulence intensity data directly from the anemometer tower.

[0131] High-precision DEM data: used for subsequent analysis of the potential impact of terrain on the wake. Surface roughness maps can be generated based on the corresponding high-precision DEM data.

[0132] S42. Construct and customize the wind turbine wake model:

[0133] The wake of a wind turbine is a complex fluid dynamics phenomenon. This embodiment uses the widely validated Jensen wake model (also known as the Park model) as the core calculation engine, and customizes key parameters for the desert environment.

[0134] The core function of the Jensen model is as follows:

[0135]

[0136] in, This represents the wind speed at a distance x downwind. Indicates the incoming wind speed. This indicates the thrust coefficient of the wind turbine. This indicates the rotor diameter of the wind turbine, and x represents the downwind distance. This represents the wake attenuation coefficient, which is dynamically adjusted based on atmospheric stability. Specifically, after obtaining... Then, it can be used and The difference is used to obtain the wind speed loss.

[0137] S43, Special Adaptation to the Desert Environment – ​​Dynamic Wake Attenuation Coefficient k:

[0138] The atmospheric stratification in the Shago desert region is unstable (strong convection due to intense surface heating during the day) and the background turbulence intensity is high (due to surface roughness and thermal effects), which leads to a faster recovery speed of the wake and a significantly smaller affected area compared to the neutral stability conditions.

[0139] Adaptive design:

[0140] The wake attenuation coefficient k is a function dynamically related to atmospheric stability. Based on observations and research on the atmospheric boundary layer characteristics of the Shago Desert region, this embodiment uses the following relationship for dynamic setting:

[0141] Unstable atmospheric conditions (usually occurring during clear daytime): k = 0.05 to 0.06 (rapid wake recovery and small impact range).

[0142] Neutral atmospheric conditions (typically occurring on cloudy days or during periods of strong wind): k = 0.04 to 0.05 (standard conditions).

[0143] Stable atmospheric conditions (usually occurring during clear nighttime periods): k = 0.03 to 0.04 (slow wake recovery, large impact area).

[0144] By statistically analyzing wind frequencies under different stability conditions, a weighted average k value is assigned to each wind direction sector, or multiple stability scenarios are run directly and then the comprehensive results are synthesized.

[0145] S44. Define wake impact threshold and buffer generation:

[0146] Set a wind speed loss threshold: When the wind speed at a certain point is reduced to 90% of the incoming wind speed due to the wake effect of the upstream wind turbine (i.e., a wind speed loss of 10%), that point is considered to be within the wake region with significant influence. This threshold can be adjusted according to specific requirements for power generation efficiency or turbulence intensity.

[0147] Calculate the dynamic wake range: Taking a single wind turbine as the center, along the dominant and secondary wind directions (usually covering 16 wind direction sectors), calculate the farthest distance that can be reached when the wind speed loss reaches the wind speed loss threshold under each wind direction sector according to a customized Jensen model (using dynamic k value).

[0148] Generate wind direction sector buffers: Calculate the farthest distance that can be reached when the wind speed loss reaches the wind speed loss threshold under each wind direction sector as the dynamic wake range. The azimuth of each sector is determined by the wind direction frequency, and the radius R is determined by the farthest influence distance of the wake under that wind direction (R = x_max).

[0149] Generate a comprehensive wake influence zone: Spatially overlay the buffer zones of all wind turbine sectors to generate a comprehensive wake influence zone layer for the entire base.

[0150] S45. Spatial Overlay and Partition Discrimination:

[0151] The generated integrated wake influence zone layer is overlaid with the computational grid for analysis to obtain the fourth partition result.

[0152] For computational grids that are entirely or partially located within the integrated wake influence zone, their partition type is marked or partially marked as "wind turbine wake influence zone".

[0153] This region is a key input area for subsequent high-precision turbulence intensity calculations and carbon flux transport simulations.

[0154] S46, Wind Erosion Risk Sign:

[0155] The wake vortex of wind turbines can significantly enhance the wind speed shear force near the ground, which can exacerbate soil erosion on sandy surfaces, cause dust, and thus indirectly affect carbon flux (dust deposition) and the surrounding ecological environment.

[0156] Adaptive design:

[0157] Based on the delineation of the wake zone, a surface soil type map (which can be interpreted from remote sensing imagery or geological maps) is overlaid. If the wake influence area overlaps with the loose sandy soil area, the overlapping area is further identified as a "high wind erosion risk area" and highlighted in the final zoning map to provide a precise target area for the source region of aeolian particulate matter.

[0158] In a specific implementation process, to ensure the objectivity, accuracy, and engineering practicality of the Shagohuang New Energy Base zoning, this embodiment provides a method for spatially constraining, verifying, and refining the zoning results by comprehensively utilizing the coordinate information of equipment deployment drawings, remote sensing image recognition results, and Geographic Information System (GIS) spatial analysis technology. This method ensures a high degree of consistency between the zoning boundaries and the actual engineering facilities.

[0159] Specifically, the coordinates of facility elements in the equipment deployment drawings are transformed to a coordinate system consistent with the multi-source data, generating a facility element vector layer. The computational grid is then preliminarily partitioned to obtain a fifth partition result. This result is used to further integrate partition results obtained by other methods to perform a comprehensive logical judgment on each computational grid, ultimately determining the partition type to which each computational grid belongs. The specific implementation process is as follows:

[0160] S51. Data Preparation and Coordinate System I:

[0161] Obtain the equipment deployment design drawings or as-built drawings of the Shagohuang New Energy Base. The drawings are usually in CAD format (such as .dwg or .dxf) and contain the spatial range, boundaries, and attribute information of various facilities.

[0162] Data Extraction: Parse and extract the vector boundary coordinates of the following key facility elements from the drawings:

[0163] Photovoltaic array area: The precise boundary range of each photovoltaic array.

[0164] Wind turbine: Coordinates of the center point of the wind turbine tower and the area of ​​its foundation.

[0165] Electrical facilities: the boundary area of ​​the substation, the collection station, and the box-type transformer.

[0166] Road: The centerline or boundary line of a road under construction or maintenance.

[0167] Other structures: office area, fence, etc.

[0168] Coordinate transformation and projection: The original coordinate system in the drawings (which may be a local coordinate system or a construction coordinate system) is precisely transformed using common control points to unify it under the same geodetic coordinate system (such as WGS-84) and projected coordinate system (such as UTM) as the remote sensing imagery and DEM data. This is a prerequisite for all spatial overlay analysis.

[0169] S52, Construct the authoritative vector layer for facility elements:

[0170] The converted facility elements are then used to generate corresponding vector layers (such as polygonal ".shp" files) in GIS software.

[0171] Photovoltaic array area (PV_Design.shp);

[0172] Wind turbine foundation scope (Turbine_Design.shp);

[0173] Substation Scope (Substation_Design.shp);

[0174] Road surface area (Road_Design.shp);

[0175] These vector layers constitute the "authoritative spatial benchmark" for the partition.

[0176] S53, Spatial Constraints and Forced Assignment – ​​Highest Priority Judgment:

[0177] The aforementioned authoritative vector layer is overlaid with the preliminary zoning results obtained through remote sensing interpretation (OverlayAnalysis), and the following forced assignment logic is executed to ensure that the drawing information has the highest priority:

[0178] Direct assignment area:

[0179] All computational grids within the PV_Design.shp area, regardless of their spectral or topographical features, must be forcibly classified as "photovoltaic power generation zone" in their final partitioning type.

[0180] All computational meshes within the Turbine_Design.shp area must be forcibly divided into "wind power generation zones".

[0181] All computational meshes within the Substation_Design.shp or Road_Design.shp area must be forcibly divided into "Electrical Facilities Zone" or "Road Zone".

[0182] This step ensures that the zoning results are completely consistent with the engineering design, avoiding any possible errors in remote sensing identification.

[0183] S54. Spatial verification and conflict handling:

[0184] For areas outside the authoritative vector layer range, remote sensing interpretation results are primarily relied upon. However, for areas where the interpretation results are adjacent to the drawing information, a consistency check is performed.

[0185] False Negative Check: If remote sensing identifies a region as a "primitive desert area," but this area is largely covered by PV_Design.shp, it indicates that the remote sensing missed detecting the photovoltaic array. The system should record this conflict and correct it based on the drawings.

[0186] False Positive Check: If remote sensing interprets an area as a "photovoltaic power generation zone," but this area is completely outside the scope of any design drawings, it indicates that remote sensing may have misidentified bare rock, water accumulation, etc., as photovoltaic panels. The system should record this conflict and initiate manual interactive review or re-identify the area by combining other evidence (such as time-series imagery).

[0187] S55. Definition and marking of "Facility Adjacent Area":

[0188] Special considerations: The Shagohuang New Energy Base has a unique microenvironment and human activities surrounding the facilities, which need to be distinguished.

[0189] Generate a buffer of a certain width (e.g., 10-20 meters) around Substation_Design.shp, Road_Design.shp, etc.

[0190] Grid cells located within the buffer zone but not included in the core facility area are designated as "facility proximity zones." These areas may be subject to frequent human disturbance and have complex surface types (such as compacted land and temporary stockpiles), and their carbon flux characteristics may differ significantly from those of the distant "primitive desert areas." This refined grading improves the granularity of the zoning and the accuracy of the model.

[0191] S56. Generate the final authoritative partition map:

[0192] In summary, the above steps are as follows:

[0193] The vector region generated by the device deployment coordinates is used as the core framework for forced assignment.

[0194] In addition to this framework, rules based on remote sensing spectroscopy, topography, and other factors are applied for supplementary discrimination.

[0195] Perform a logical consistency check on all areas (e.g., wind turbine foundations should not appear in the photovoltaic area) to generate the final digital zoning map that fully corresponds to the on-site project.

[0196] Through the above embodiments, the method closely integrates abstract remote sensing image features with real-world engineering coordinates, forming a dual-drive zoning process that combines "top-down" (drawing constraints) and "bottom-up" (remote sensing inversion). This maximizes the accuracy, reliability, and engineering application value of the zoning results, providing a data foundation for subsequent accurate carbon flux measurement.

[0197] It should be noted that the specific steps for combining the results of each partition with other partition results to determine the final partition type of each computing grid are only illustrative examples. The integration logic can be set according to actual needs, and will not be illustrated in detail here.

[0198] In a specific implementation process, the gradient diffusion method refers to a method for calculating vertical mass transport based on turbulent diffusion theory. Specifically, it is achieved by analyzing the product of the carbon dioxide concentration gradient and the eddy diffusion coefficient, used to quantify the contribution of the vertical diffusion process to carbon flux. The eddy correlation method refers to a method for calculating mass flux based on the characteristics of atmospheric turbulent fluctuations. Specifically, it is achieved by calculating the correlation between vertical wind speed fluctuations and concentration fluctuations using covariance, used to capture the instantaneous exchange characteristics of turbulence. Weighted summation refers to the comprehensive processing of calculation results from different methods. Specifically, it can be achieved by using a dynamic weight allocation strategy combined with an error analysis model, used to balance the systematic errors of different methods. The correction coefficient refers to an adjustment parameter reflecting the environmental influence of a specific zone, used to eliminate calculation biases caused by differences in land surface types.

[0199] Specifically, the new energy base is first divided into several functional zones, including a photovoltaic power generation zone, a wind power generation zone, an electrical facility zone, a road zone, a pristine desert zone, and an artificial greening zone. Each zone is defined based on spatial attributes such as vegetation index thresholds and facility types. For each zone, carbon flux calculations are performed simultaneously using both the gradient diffusion method and the eddy covariance method. The gradient diffusion method obtains the flux value by multiplying the concentration gradients at three vertical height levels by the eddy diffusion coefficient, while the eddy covariance method calculates the flux value using the covariance of high-frequency pulsating data. The results of the two methods are weighted according to the coefficient of determination to form a fused baseline flux value. Based on this, a correction coefficient is applied to the underlying surface characteristics of each zone, and finally, the overall carbon flux is calculated using area weighting.

[0200] In a specific implementation process, the flux correction logic for each zone is as follows: the flux in the original desert zone is dynamically adjusted based on the soil respiration correction coefficient; the flux in the photovoltaic power generation zone introduces the shading correction coefficient; the flux in the wind power generation zone is superimposed with the thermal disturbance term; the flux in the artificial greening zone is superimposed with the photosynthetic absorption flux; the flux in the electrical facility zone is superimposed with the inverter emission term; and the flux in the road zone can be superimposed with the vehicle emission term.

[0201] Among them, the flux correction coefficient for primary desert areas refers to a dynamic adjustment factor based on changes in soil temperature and moisture content, specifically achieved using a linear combination formula of temperature and moisture content, reflecting the nonlinear response of soil respiration to environmental changes. The shading rate of photovoltaic panels refers to the proportion of the ground obstructed by the photovoltaic array, specifically calculated through UAV image analysis or irradiance intensity comparison, used to quantify the shading effect of photovoltaic facilities on the ground carbon exchange process. The tip velocity of wind turbine blades refers to the linear velocity at the end of the wind turbine blades, specifically calculated by multiplying the rotational speed by the blade length, used to characterize the degree of influence of wind turbine operation on the surrounding turbulence intensity. Photosynthetically active radiation refers to the solar radiation energy available for vegetation photosynthesis, specifically measured by spectral sensors or converted using solar irradiance models, used to quantify the contribution of vegetation photosynthesis to carbon flux absorption. The leaf area index refers to the total area of ​​vegetation leaves per unit surface area, specifically achieved through multispectral image inversion or lidar scanning, used to reflect the impact of vegetation canopy structure on photosynthetic efficiency. Inverter emissions refer to the carbon dioxide emissions generated by the operation of power facilities. These emissions can be calculated using a power and emission factor model to quantify the additional impact of equipment operation on carbon flux. Road vehicle emission factors refer to the carbon dioxide emissions generated per unit of traffic flow or per unit of mileage. These emissions can be calculated by matching vehicle type statistics with an emission factor database to quantify the additional impact of road vehicles on carbon flux.

[0202] Specifically, this method divides the area into multiple zones, including photovoltaic (PV) power generation, wind power generation, electrical facilities, roads, pristine desert, and artificial greening, and establishes differentiated correction models for each zone based on underlying surface characteristics and facility interference. Flux correction in the pristine desert zone considers the synergistic effect of soil temperature and moisture content on respiration, dynamically adjusting the base flux value using a linear combination formula. Flux correction in the PV array zone quantitatively assesses the shading effect of PV panels based on shading rate, reducing the base flux value through a proportionality coefficient. Flux correction in the wind turbine periphery zone combines blade tip velocity to calculate turbulence enhancement effects and superimposes a constant thermal disturbance term to reflect the heat emissions from wind turbine operation. Flux correction in the vegetation zone introduces a product term of photosynthetically active radiation and leaf area index to characterize the carbon dioxide absorption capacity of vegetation photosynthesis. Flux correction in the facility zone uses a temperature-dependent coefficient to adjust the base flux, while superimposing an inverter emission term inversely proportional to the facility area. Flux correction in the road zone can be superimposed with vehicle emission terms. All correction parameters are derived from multi-field data output by LBM simulation, achieving refined calculation of zoned carbon flux through formula coupling.

[0203] This scheme effectively addresses the adaptability issue in carbon flux calculation for densely populated areas by employing a multi-attribute gridding partitioning mechanism, combining the complementary advantages of the gradient diffusion method and the eddy covariance method, and introducing dynamic correction coefficients. Compared to the eddy covariance method, which is easily affected by the intermittent nature of turbulence, or the gradient diffusion method, which struggles to capture transient transport, this scheme achieves lower root mean square errors for both methods through a weighted fusion mechanism.

[0204] Through the above technical solutions, this invention achieves refined modeling of complex underlying surface conditions within new energy bases, overcoming the calculation bias of traditional methods in densely populated areas. A zonal correction mechanism effectively distinguishes the interaction between natural carbon cycling and artificial facilities, keeping the calculation error of carbon flux under special conditions such as photovoltaic array shading effects and wind turbine wake disturbances within 5%. Uncertainty analysis using the Monte Carlo method quantifies the sensitivity of different correction coefficients to the final result, providing data support for the optimized deployment of carbon flux monitoring networks.

[0205] In some embodiments, the present invention further proposes a method for calculating the first carbon flux using the gradient diffusion method, including: calculating the vertical concentration gradient of carbon dioxide and the eddy diffusion coefficient based on turbulent diffusion theory, combined with carbon dioxide concentration distribution and velocity distribution; multiplying the vertical concentration gradient of carbon dioxide and the eddy diffusion coefficient and taking the negative value to obtain the first carbon flux. The calculation process of the vertical concentration gradient of carbon dioxide includes selecting multiple turbulently stable vertical height layers from a preset height range, where the height difference between each vertical height layer is a preset difference; extracting the average carbon dioxide concentration of each vertical height layer; calculating the carbon dioxide concentration difference between two adjacent vertical height layers; and averaging all carbon dioxide concentration differences to obtain the vertical concentration gradient of carbon dioxide. The calculation process of the eddy diffusion coefficient includes: selecting a first velocity at a data point at a first height and a second velocity at a data point at a second height based on the velocity distribution; determining the height difference between the second height and the first height, and the velocity difference between the first velocity and the second velocity; and multiplying the height difference, velocity difference, and von Kármán constant to obtain the eddy diffusion coefficient.

[0206] The carbon dioxide vertical concentration gradient refers to the difference in carbon dioxide concentration between different vertical height layers. Specifically, it can be achieved by selecting multiple turbulent and stable vertical height layers within a preset height range, extracting the average carbon dioxide concentration of each vertical height layer, and calculating the carbon dioxide concentration difference between two adjacent vertical height layers; the average of all carbon dioxide concentration differences is then taken to obtain the carbon dioxide vertical concentration gradient. The height difference of each vertical height layer is a preset difference value. This carbon dioxide vertical concentration gradient reflects the diffusion trend of carbon dioxide in the vertical direction and is a core parameter for calculating carbon flux using the gradient diffusion method.

[0207] The eddy diffusion coefficient is a parameter characterizing the ability of turbulent motion to diffuse matter. Specifically, based on the velocity distribution, a first velocity at a first altitude and a second velocity at a second altitude are selected; the height difference between the second and first altitudes, and the velocity difference between the first and second velocities are determined; the eddy diffusion coefficient is obtained by multiplying the height difference, the velocity difference, and the von Kármán constant. This eddy diffusion coefficient reflects the contribution of atmospheric turbulence to the vertical transport of carbon dioxide and directly affects the accuracy of carbon flux calculations.

[0208] Specifically, in the implementation process, firstly, multiple turbulently stable vertical height layers are selected from the LBM simulation data. For example, three height layers are selected within the range of 10 to 30 meters, with a height difference of 10 meters between adjacent layers. After extracting the average carbon dioxide concentration of each layer, the concentration difference between adjacent layers is calculated, and the average value is taken as the vertical concentration gradient. Simultaneously, based on the velocity distribution data of the same region, wind speed values ​​at two different heights are selected, and the height difference and wind speed difference are calculated. The eddy diffusion coefficient is then calculated using the von Kármán constant. For example, when the first height is 10 meters and the second height is 20 meters, the measured wind speeds are 5 m / s and 6 m / s, respectively. The eddy diffusion coefficient is then calculated using the formula. Finally, the vertical concentration gradient is multiplied by the eddy diffusion coefficient and the result is negative to obtain the first carbon flux.

[0209] This method effectively reduces measurement bias caused by local turbulent fluctuations by selecting multiple turbulent stability height layers and calculating the average concentration difference. Furthermore, the method of calculating the eddy diffusion coefficient based on the difference in wind speed at two heights more accurately reflects the actual turbulent diffusion capacity than traditional single-point wind speed estimation.

[0210] Through the above technical solution, this invention can improve the calculation accuracy of the vertical concentration gradient of carbon dioxide and the eddy diffusion coefficient, thereby ensuring the reliability of the first carbon flux data obtained by the gradient diffusion method. This method, through multi-height layer data averaging and two-point wind speed calculation, effectively suppresses the influence of random errors from single-point measurements on the final result, providing more accurate basic input parameters for subsequent carbon flux fusion calculations.

[0211] In some embodiments, the present invention further proposes to differentiate the settings based on each partition type and meteorological interference factors to obtain a probability distribution model for the eddy diffusion coefficient, the vertical concentration gradient of carbon dioxide, and the correction coefficient; wherein, the meteorological interference factors include at least aerosol optical interference caused by strong winds and sandstorms, and intense near-surface thermal turbulence caused by strong solar radiation; Monte Carlo random perturbation iteration is performed based on the differentiated probability distribution model to obtain multiple candidate carbon fluxes; after sorting the multiple candidate carbon fluxes, the top n candidate carbon fluxes and the bottom m candidate carbon fluxes are extracted, and the minimum candidate carbon flux and the maximum candidate carbon flux among the remaining candidate carbon fluxes constitute the confidence interval of the carbon flux of the Shagohuang New Energy Base.

[0212] The differentiated setting refers to using different distribution methods for different zoning types and adjusting the standard deviation of each zoning type for different meteorological interference factors. For example, for the eddy diffusion coefficient, a normal distribution can be used in unobstructed areas such as road areas, while a skewed distribution can be used in areas such as photovoltaic power generation areas and wind power generation areas where turbulence distribution is uneven due to equipment obstruction. When strong solar radiation is detected, the standard deviation of each zoning type can be increased, while when there is no strong radiation, the standard deviation of each zoning type can be left unchanged. The vertical concentration gradient of carbon dioxide and the correction coefficient can also be differentiated based on similar principles and actual needs. Examples will not be given here.

[0213] The Monte Carlo method is a numerical simulation method based on probability and statistics. It employs random sampling and repeated calculations to simulate carbon flux distribution under different scenarios by generating a large number of random parameter combinations. The eddy diffusion coefficient is a parameter characterizing the ability of atmospheric turbulence to diffuse matter. It is calculated using wind speed gradients and the von Kármán constant and is used to quantify the vertical transport efficiency of carbon dioxide. The vertical carbon dioxide concentration gradient is the rate of change of carbon dioxide concentration per unit height. It is calculated by the average concentration difference across multiple height layers and reflects the spatial distribution characteristics of carbon dioxide. The correction coefficient is an adjustment factor set for the surface characteristics of different functional zones. It is dynamically calculated based on soil temperature, moisture content, and equipment operating parameters and is used to compensate for biases in the basic carbon flux model in specific regions. The confidence interval is a statistically significant range representing the reliability of the results. It is obtained by removing extreme values ​​and retaining the region with the highest intermediate probability density and is used to quantify the uncertainty of carbon flux calculation results.

[0214] Specifically, the Monte Carlo method involves determining the perturbation range for key parameters. For example, the eddy diffusion coefficient is allowed to fluctuate by ±15%, the vertical carbon dioxide concentration gradient by ±10%, and the correction coefficient by ±8%. A computer program performs 1000 independent iterations, with parameter values ​​randomly selected from the pre-defined perturbation range and the carbon flux recalculated in each iteration. After sorting all iteration results by value, the top 2.5% and bottom 2.5% of extreme data are removed, and the interval between the minimum and maximum values ​​in the remaining 95% of data is the confidence interval. This process effectively reflects the impact of model parameter errors on the final result.

[0215] Compared to existing technologies, traditional carbon flux assessment methods typically rely on calculations from a single model, neglecting the impact of parameter uncertainties on the results. Existing technologies lack error propagation analysis methods for multi-field coupled models in desert and Gobi regions, making it impossible to quantify the reliability boundaries of monitoring results. This proposed solution introduces parameter perturbation mechanisms and probabilistic statistical methods to construct a comprehensive assessment system covering model errors and measurement errors.

[0216] Through the above technical solution, this invention can accurately quantify the uncertainty range of carbon flux calculation in the Shagohuang new energy base, providing decision-makers with statistically significant confidence interval data. This method effectively solves the problem of insufficient reliability of traditional single-model calculation results, especially under complex underlying surfaces and multi-field coupling conditions, significantly improving the credibility and engineering applicability of carbon flux monitoring results.

[0217] In some embodiments, the present invention further proposes a method for calculating a second carbon flux using the eddy covariance method, comprising: extracting high-frequency data from multiple data points within a preset time period based on atmospheric turbulence fluctuation characteristics; the sampling time interval of the high-frequency data being a preset time interval; the high-frequency data including high-frequency vertical velocity extracted from the velocity distribution and carbon dioxide concentration extracted from the carbon dioxide concentration distribution, with a sampling frequency of not less than 10 Hz; performing quality labeling and preprocessing on the high-frequency raw data sequence corresponding to the high-frequency data, including at least removing abnormal values ​​of carbon dioxide concentration signals caused by aerosol concentration anomalies due to strong winds and sandstorms in the Gobi Desert environment; subtracting the high-frequency vertical velocity of each data point in the sequence from the mean of the vertical velocity to obtain the fluctuation value of the high-frequency vertical velocity of each data point, and subtracting the carbon dioxide concentration of each data point from the mean of the carbon dioxide concentration of all data points to obtain the fluctuation value of the carbon dioxide concentration of each data point; and calculating the covariance as the second carbon flux based on the fluctuation value of the high-frequency vertical velocity of each data point and the fluctuation value of the carbon dioxide concentration of each data point.

[0218] High-frequency data refers to sampling data with short sampling intervals. Pulsation values ​​refer to the instantaneous deviation of a physical quantity from its average value, which can be calculated using a moving average of time-series data, such as a 30-minute moving average window, to eliminate background trends and extract turbulent fluctuation components. Covariance refers to the statistical correlation between the fluctuation components of two variables, which can be achieved by averaging the product of time-synchronized vertical wind speed and concentration fluctuations, for example, by summing the products of 18,000 consecutive sampling points and taking the average, to quantify the contribution of vertical turbulent transport to carbon flux.

[0219] Specifically, in the Shagohuang New Energy Base, the vertical transport of carbon dioxide caused by atmospheric turbulence exhibits significant spatiotemporal variations. High-frequency sampling equipment, such as ultrasonic anemometers and open-circuit infrared gas analyzers installed on flux towers, is deployed to continuously collect vertical wind speed and carbon dioxide concentration data at 20Hz sampling intervals. For data within each 30-minute time window, the 30-minute average of the vertical wind speed is first calculated. Subtracting this average from each instantaneous measurement yields a vertical wind speed fluctuation sequence. Simultaneously, the 30-minute average of the carbon dioxide concentration is calculated, generating a concentration fluctuation sequence. The two fluctuation sequences are then multiplied point-by-point and summed, then divided by the total number of data points to obtain the covariance value. This covariance directly characterizes the net vertical flux caused by turbulent motion; for example, it contributes positively when updrafts carry high concentrations of carbon dioxide and negatively when downdrafts carry low concentrations. Statistical averaging eliminates random errors and accurately reflects the carbon flux.

[0220] Through the above technical solution, this invention can accurately quantify the instantaneous contribution of turbulent motion to carbon flux, avoiding systematic biases caused by low-frequency sampling. For example, in the photovoltaic array region, due to the enhanced turbulence effect caused by the photovoltaic panels, the carbon flux accuracy of traditional methods is poor. However, this method captures the enhanced turbulence characteristics through high-frequency data, making the calculation results closer to the real physical process. At the same time, this method complements the gradient diffusion method, and by weighted fusion, it considers both molecular diffusion and turbulent transport mechanisms, significantly improving the overall accuracy of carbon flux inversion.

[0221] In some embodiments, the image data is a multi-temporal remote sensing image sequence covering the Shagohuang New Energy Base; the ground parameter model is a spatiotemporal sequence prediction model built based on the Transformer architecture. This invention further proposes inputting the image data into a pre-trained ground parameter model for identification, outputting ground boundary parameters and ground source term parameters corresponding to each grid cell in the computational grid. Specifically, this includes: processing the multi-temporal remote sensing image sequence to extract the dynamic change characteristics of surface cover over time, and outputting the ground boundary parameters and ground source term parameters corresponding to each computational grid.

[0222] In a specific implementation, a Transformer can include an encoder, a decoder, a self-attention mechanism, etc.

[0223] The encoder receives multi-temporal remote sensing image sequences (such as time series composed of daily images covering the Shagohuang New Energy Base), uses a multi-head self-attention mechanism to mine the temporal dependencies between images at different time steps (such as the correlation of changes in land cover over consecutive days), and interactively extracts spatial features within a single image (such as the spatial distribution of vegetation and facilities in different areas), encoding the image sequence into a feature representation containing spatiotemporal dynamic information.

[0224] Decoder: Based on the features output by the encoder, the decoder generates ground parameters corresponding to the computational grid through a self-attention mechanism (including cross-attention, focusing on key features of the encoder). For static ground boundary parameters (such as average ground roughness and average thermal conductivity), the decoder directly outputs the parameter values ​​for each grid. For dynamic ground source parameters (such as the time series of CO2 uptake / release rates), the decoder further learns their temporal variation patterns and outputs parameter sequences for continuous time steps (such as the daily CO2 release rate for the next 7 days).

[0225] Supervised learning can be used for training: historical multi-temporal remote sensing image sequences are used as input, while ground boundary parameters (measured or high-confidence simulation values, such as average blackbody emissivity, ground roughness, etc.) and ground source parameters (measured data of CO2 absorption / release rate, sensible heat flux, etc.) for the corresponding time period and computational grid are collected as labels. By defining a loss function (such as mean squared error for continuous parameters, and cross-entropy for parameters with categorical properties), the error between the model output and the label is minimized, thereby optimizing the parameters of the Transformer encoder, decoder, and self-attention mechanism.

[0226] In a specific implementation process, taking the prediction of ground CO2 emission rate at the Shagohuang New Energy Base as an example:

[0227] Input: A multi-temporal remote sensing image sequence covering the base once a day for the past 30 days (capturing the dynamic changes of surface vegetation, soil, new energy facilities, etc. over 30 days).

[0228] Output: The time series of ground CO2 emission rate for each computational grid at the base over the next 7 days (i.e., the dynamic pattern of output over time), along with the corresponding ground boundary parameters (such as average ground roughness and average blackbody emissivity).

[0229] Results demonstration: A line graph of "time versus CO2 release rate" is plotted to compare the results of the Transformer model with those of traditional methods (such as prediction models based on physical equations), thus intuitively demonstrating its superior performance in spatiotemporal sequence prediction compared to traditional methods.

[0230] In a specific implementation, the ground source term parameters are dynamic parameters that change over time, and the model simultaneously outputs the time variation pattern of these parameters; the ground boundary parameters include, but are not limited to, the average ground roughness value, average thermal conductivity, and average blackbody emissivity of the computational grid; the ground source term parameters include, but are not limited to, the average carbon dioxide absorption / release rate source term, oxygen absorption / release rate source term, dust particle release rate source term, sensible heat flux source term, latent heat flux source term, methane absorption / release rate source term, nitrous oxide absorption / release rate source term, water evaporation / transpiration rate source term, volatile organic compound release rate source term, anthropogenic heat source term, and equipment volatile organic compound release rate source term.

[0231] Among them, the source item of dust particle emission rate is directly related to the "strong wind and sand" characteristics of the desert area. Dust not only affects the ecology, but also indirectly affects the climate by changing radiation transmission and snow and ice albedo.

[0232] Sensible heat flux source term: Heat exchange between the Earth's surface and the atmosphere caused by temperature differences. It strongly affects the stability of the atmospheric boundary layer and the development of turbulence, and directly restricts the eddy diffusion coefficient.

[0233] Latent heat flux source term (evaporation / transpiration source term): The energy consumed by water evaporation or plant transpiration. It is a key parameter connecting the water cycle and the energy cycle. Even in arid regions, evaporation processes following limited vegetation and brief rainfall are very important.

[0234] Methane absorption / release rate source term: Although the Shago desert region is not a major source or sink of methane, nearby livestock farming or wetlands may contribute. Thorough consideration is needed.

[0235] Nitrous oxide uptake / release rate source term: Again, from soil microbial processes.

[0236] Volatile organic compound (VOC) release sources: Vegetation (especially certain shrubs) releases VOCs, which participate in atmospheric chemical reactions.

[0237] Water evaporation / transpiration rate source term: another form of latent heat flux, quantifying water loss.

[0238] Dew condensation rate source term: In arid regions, dew that condenses at night is an important water source for some plants and microorganisms.

[0239] Source terms directly related to new energy facilities include:

[0240] Anthropogenic heat sources: Heat emitted from power facilities such as photovoltaic inverters and transformer substations during operation forms a local, stable artificial heat source, affecting the local microclimate.

[0241] Sources of volatile organic compound (VOC) release from equipment: trace amounts of gases that may be released from materials such as transformer insulating oil and cable sheaths at high temperatures.

[0242] In a specific implementation process, to achieve high-precision supervised training of the ground parameter model, it is necessary to obtain high-reliability ground truth data as sample labels. This embodiment, tailored to the environmental characteristics of the Shagohuang New Energy Base, provides a process for constructing a sample label library that combines multi-source fusion, air-ground collaboration, and mechanism-driven approaches.

[0243] S61, Labels for obtaining ground boundary parameters:

[0244] The ground boundary parameters are mainly static parameters characterizing the inherent physical properties of the underlying surface, and are obtained in the following ways:

[0245] Average surface roughness length (z0):

[0246] Main method: Airborne LiDAR scanning. High-precision digital elevation model (DEM) and digital surface model (DSM) of the base are acquired through flight operations. The roughness length value of each computational grid is calculated based on the standard deviation of the elevation data using the micro-topography method. This method can obtain full-coverage, high-precision spatial distribution data.

[0247] Auxiliary method: For typical underlying surface types (such as smooth photovoltaic panels, Gobi Desert, and shrublands), a miniature drag anemometer array is laid out. The roughness is directly calculated by measuring the vertical wind speed profile and used to verify the remote sensing inversion results.

[0248] Average thermal conductivity and average blackbody emissivity (ε):

[0249] Main methods: field sampling and laboratory measurement. Typical surface materials (such as sand, gravel, photovoltaic glass, and concrete) were sampled in different zones (photovoltaic zone, desert zone, etc.) within the base. The thermal conductivity was measured in the laboratory using the thermal probe method, and the emissivity spectral lines were measured using Fourier transform infrared spectroscopy (FTIR) and the integral value was calculated to obtain the blackbody emissivity.

[0250] Spatial extension: Laboratory measurements are used as prior knowledge and cross-validated with emissivity results retrieved from hyperspectral remote sensing images. Finally, each computational grid is assigned a weighted average parameter value based on the proportion of its land cover type.

[0251] S62. Obtain the label for the ground source term parameter:

[0252] The ground source term parameters are dynamic parameters characterizing the flux of energy and matter exchange, and their acquisition depends on high-frequency observations.

[0253] Sensible heat flux (H), latent heat flux (LE), carbon dioxide (CO2), methane (CH4) flux:

[0254] The gold standard: Eddy Covariance (EC) system. Multiple EC systems are deployed in representative underlying surface areas within the site (such as concentrated photovoltaic areas, pristine desert areas, and green areas). This system synchronously measures three-dimensional wind speed, temperature, water vapor, and CO2 concentration fluctuations at high frequencies (10-20Hz), directly calculating the true values ​​of sensible heat, latent heat, CO2, and CH4 fluxes at half-hour or hourly averages. This is the international standard method for flux observation.

[0255] Spatial representativeness: By deploying mobile flux observation systems (such as those installed on vehicles) to conduct short-term round-trip observations at different locations, the spatial coverage of fixed-site observations is compensated for, and the spatial extrapolation capability of the model is validated.

[0256] Dust particle emission rate source items:

[0257] Main methods: A large-scale weighing automatic dust collection cylinder network and laser dust monitors. A dense network of dust collection cylinders is deployed within the site to periodically collect and weigh the settled dust material, converting it into time-integrated flux. Simultaneously, laser dust monitors are used to monitor real-time changes in PM10 / PM2.5 concentrations in the air. Combined with wind speed and direction data, the local dust release flux is obtained through a reverse diffusion model.

[0258] Control experiment: A wind tunnel experimental device was set up in a typical sand source area to erode the surface soil with different moisture contents and directly measure its sand-lifting flux curve to provide mechanistic label data for the model.

[0259] Anthropogenic heat source and equipment VOCs emission rate source:

[0260] Anthropogenic heat source item: Heat flux plate sensors are deployed around power facilities such as substations and transformer substations to directly measure the heat flux emitted from their surfaces into the environment. Simultaneously, real-time power, load rate, and operating efficiency of the equipment are obtained from the Supervisory Control and Data Acquisition (SCADA) system, and their theoretical heat dissipation is calculated using the heat balance equation and cross-checked with the measured values.

[0261] VOCs emission rate from equipment: Portable gas chromatography-mass spectrometry (GC-MS) was used to sample and analyze the air around operating equipment such as transformers and cable trenches to qualitatively and quantitatively identify the types and concentrations of VOCs. Combined with computational fluid dynamics (CFD) simulations, the source intensity emission rate was calculated.

[0262] Trace gas fluxes such as nitrous oxide (N2O) and volatile organic compounds (VOCs):

[0263] Main method: Static chamber-gas chromatography. Sealed chambers are placed in typical locations within the base (such as after irrigation of green belts or near livestock farms). Gas is periodically extracted from the chambers, and the concentration changes of gases such as N2O and VOCs are analyzed using a high-precision gas chromatograph to calculate their flux. Although this method cannot provide continuous observation, it can obtain high-precision data at key locations and during key time periods.

[0264] S63. Construct a sample label database:

[0265] All high-confidence data of the ground boundary parameter labels and ground source parameter labels, obtained through field measurements, remote sensing inversion, and model calculations, are time-synchronized and spatially gridded. These data are then rigorously matched with the input data of the corresponding multi-temporal remote sensing image sequences to construct a sample label database for training and validating the ground parameter model. The construction of this database ensures the scientific rigor and effectiveness of the supervised learning process and is the foundation for the Transformer model to achieve high-precision predictions.

[0266] In some embodiments, the present invention further proposes to monitor cloud information using an all-sky imager, combine it with a high-precision solar trajectory tracking model, and use the Monte Carlo method to calculate the ground irradiance distribution under cloud cover conditions. This includes: continuously acquiring sky images of the monitoring area by deploying an all-sky imager; preprocessing the sky images to obtain cloud information, the preprocessing including eliminating image blurring and color distortion caused by dust aerosols based on the environmental characteristics of the desert region, and separating cloud and dust information; based on the cloud information, using image processing and inversion algorithms to determine the specific location and shape of the clouds in space; introducing a typical aerosol optical thickness model of the desert region, which, together with the cloud optical characteristics, serves as the input parameter for Monte Carlo ray tracing; using the solar trajectory tracking method, calculating the real-time position of the sun in the sky based on geographical location, date, and time; using the Buie distribution model as the shape model of the sun to describe the characteristics of sunlight with a non-uniform light intensity distribution; and using the Monte Carlo method to perform ray tracing simulation to calculate the solar irradiance distribution at different locations on the ground of the desert new energy base under cloud cover conditions.

[0267] Among these, an all-sky imager refers to a device capable of continuously capturing sky images with a hemispherical field of view, specifically using a CCD camera equipped with a fisheye lens, acquiring dynamic data on cloud cover changes through timed shooting. Cloud information inversion refers to extracting cloud height, thickness, and spatial distribution parameters from all-sky images based on image processing algorithms, specifically using threshold segmentation combined with a radiative transfer model, to determine the extent to which clouds obstruct solar radiation. Solar trajectory tracking is a mathematical model that accurately calculates the solar azimuth and elevation angles based on geographic coordinates, date, and time, specifically using the SOLPOS algorithm, to simulate the real-time movement trajectory of the sun in the sky. The Buie distribution model is a mathematical model describing the non-uniform distribution of solar intensity on the solar surface, specifically using a ring Gaussian function to characterize the light intensity attenuation characteristics, to more realistically simulate the intensity distribution of direct solar radiation. Monte Carlo ray tracing involves randomly sampling ray paths and statistically analyzing the probability distribution of their interactions with clouds, specifically using reverse ray tracing combined with importance sampling, to calculate the irradiance at various points on the ground under cloud obstruction.

[0268] Specifically, at the Shagohuang New Energy Base, an all-sky imager acquires sky images at a frequency of once per minute. These images are preprocessed to obtain cloud information. The preprocessing includes eliminating image blurring and color distortion caused by dust aerosols based on the environmental characteristics of Shagohuang, and separating cloud and dust information. Cloud regions are identified through grayscale thresholding and morphological processing. The optical thickness and spatial location of the clouds are inverted using the atmospheric radiative transfer equation. A typical aerosol optical thickness model for the Shagohuang region is introduced, which, along with the cloud optical characteristics, serves as the input parameters for Monte Carlo ray tracing. The solar trajectory tracking model calculates the real-time azimuth and altitude angles of the sun based on the base's latitude, longitude, date, and UTC time, determining its projection position in the celestial coordinate system. The Buie distribution model treats the sun as a non-uniform light source composed of a central bright spot and a surrounding attenuating halo, using a radial distribution function to simulate its intensity gradient changes. In the Monte Carlo ray tracing simulation, the obstruction and scattering of clouds can be simulated, as well as the absorption and scattering effects of dust aerosols on solar radiation, and the distribution of solar irradiance intensity at different locations on the ground of the Shagohuang New Energy Base under the presence of clouds can be calculated.

[0269] In some specific implementations, the all-sky imager can be installed atop a flux tower at the center of the base, at a height of 10 meters to avoid obstruction from ground obstacles. The cloud inversion algorithm can employ a semantic segmentation model based on a convolutional neural network, with training data including all-sky images of the desert region in different seasons and cloud height data measured by synchronous lidar. Declination correction and atmospheric refraction compensation can be incorporated into the solar trajectory tracking calculation to improve the accuracy of solar position calculation to within 0.1 degrees. The parameters of the Buie distribution model can be adjusted according to the season; a narrower halo width is used in summer to match strong solar illumination conditions, while the halo range is increased in winter to reflect the enhanced diffuse light characteristics.

[0270] This scheme uses an all-sky imager to dynamically monitor cloud morphology, combines the Buie model to accurately describe the characteristics of solar intensity distribution, and then uses the Monte Carlo method to simulate the interaction between light and clouds, which can more realistically reproduce the spatial distribution differences of irradiance under complex meteorological conditions.

[0271] Through the above technical solution, this invention solves the problem of insufficient accuracy in calculating ground irradiance intensity in the desert region caused by rapid cloud movement. By integrating dynamic cloud monitoring with a non-uniform light source model, the accuracy of setting the ground thermal radiation source term in the LBM model is significantly improved. This solution can effectively capture the local shading effect caused by cloud obstruction in the photovoltaic array area, providing high-precision irradiance input data for subsequent carbon flux simulation, while adapting to the variable meteorological conditions in the desert region and ensuring the reliability of all-weather simulation.

[0272] Based on the same general inventive concept, this invention also protects a carbon flux measurement and inversion system for the Shagohuang New Energy Base. The carbon flux measurement and inversion system for the Shagohuang New Energy Base provided by this invention will be described below. The carbon flux measurement and inversion system for the Shagohuang New Energy Base described below can be referred to in correspondence with the carbon flux measurement and inversion method for the Shagohuang New Energy Base described above.

[0273] Figure 2 This is a schematic diagram of the carbon flux metering and inversion system for the Shagohuang new energy base provided in this embodiment of the invention, as shown below. Figure 2 As shown, the carbon flux measurement and inversion system of the Shagohuang New Energy Base in this embodiment includes a construction module 21, a generation module 22, a first setting module 23, a second setting module 24, a third setting module 25, and a solution module 26.

[0274] Among them, the construction module 21 is used to acquire the physical parameters, terrain data and image data of the Shagohuang New Energy Base, so as to establish a three-dimensional geometric model of the Shagohuang New Energy Base using three-dimensional reconstruction technology.

[0275] The generation module 22 is used to perform spatial lattice processing on the region represented by the three-dimensional geometric model to generate the computational grid required for the LBM model;

[0276] The first setting module 23 is used to input the image data into a pre-trained ground parameter model for identification, and output ground boundary parameters and ground source term parameters corresponding to each grid cell in the computational grid, so as to set the ground boundary conditions and ground material source terms of the LBM model.

[0277] The second setting module 24 is used to monitor cloud information through an all-sky imager, combine it with a high-precision solar trajectory tracking model, and use the Monte Carlo method to calculate the ground irradiance distribution under cloud cover conditions, so as to set the ground thermal radiation source item of the LBM model.

[0278] The third setting module 25 is used to obtain the input boundary parameters and input source term parameters of the atmosphere in the monitoring area, so as to set the atmospheric boundary conditions and atmospheric material source terms of the LBM model.

[0279] The solver module 26 is used to solve the LBM model, obtain detailed simulation results of carbon dioxide concentration distribution, flow velocity distribution, temperature distribution and pollutant concentration distribution in the Shagohuang New Energy Base, and calculate the carbon flux of the Shagohuang New Energy Base based on the detailed simulation results.

[0280] It should be noted that all relevant information that may be involved in the various embodiments of the present invention is processed in strict accordance with the requirements of laws and regulations, following the principles of legality, legitimacy, and necessity, based on the reasonable purpose of the business scenario, and is information that users actively provide or that is generated as a result of using the product / service, as well as information that is obtained with the user's authorization.

[0281] The information processed by this invention may vary depending on the specific product / service scenario and should be based on the specific scenario in which the user uses the product / service. This may involve user account information, device information, or other related information. This invention will handle the relevant information and its processing with the utmost diligence.

[0282] This invention places great importance on the security of related information and has adopted reasonable and feasible security protection measures that comply with industry standards to protect related information and prevent unauthorized access, public disclosure, use, modification, damage or loss of related information.

[0283] The device embodiments described above are merely illustrative. 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0284] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0285] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for carbon flux measurement and inversion in a desert renewable energy base, characterized in that, include: The physical parameters, terrain data, and image data of the Shagohuang New Energy Base were obtained, and a three-dimensional geometric model of the Shagohuang New Energy Base was established using three-dimensional reconstruction technology. The region represented by the three-dimensional geometric model is spatially latticized to generate the computational grid required for the LBM model; The image data is input into a pre-trained ground parameter model for identification, and the model outputs ground boundary parameters and ground source term parameters corresponding to each grid cell in the computational grid, in order to set the ground boundary conditions and ground material source terms of the LBM model. The image data is a multi-temporal remote sensing image sequence covering the Shagohuang New Energy Base; the ground parameter model is a spatiotemporal sequence prediction model built based on the Transformer architecture. Inputting the image data into the pre-trained ground parameter model for identification and outputting ground boundary parameters and ground source term parameters corresponding to each grid cell in the computational grid includes: processing the multi-temporal remote sensing image sequence to extract the dynamic change characteristics of surface cover over time, and outputting the ground boundary parameters and ground source term parameters corresponding to each computational grid. Ground source term parameters; wherein, the ground source term parameters are dynamic parameters that change over time, and the model simultaneously outputs the time variation pattern of these parameters; wherein, the construction process of the sample label library of the ground parameter model includes: obtaining labels for ground boundary parameters; the ground boundary parameters are static parameters characterizing the inherent physical properties of the underlying surface; obtaining labels for ground source term parameters, the ground source term parameters being dynamic parameters characterizing energy and mass exchange fluxes, whose acquisition depends on high-frequency observations; constructing a sample label database: the data with a confidence level greater than a preset level obtained from the labels of the ground boundary parameters and the labels of the ground source term parameters through field measurements, remote sensing inversion, and model calculations are time-synchronized and spatially gridded, and matched with the input data of the corresponding multi-temporal remote sensing image sequences to form a sample label database for training and validation of the ground parameter model; By monitoring cloud information using an all-sky imager and combining it with a high-precision solar trajectory tracking model, the Monte Carlo method is used to calculate the ground irradiance distribution under cloud cover conditions, in order to set the ground thermal radiation source term of the LBM model. Specifically, the process of monitoring cloud information using an all-sky imager and combining it with a high-precision solar trajectory tracking model to calculate the ground irradiance distribution under cloud cover conditions includes: continuously acquiring sky images of the monitored area by deploying the all-sky imager; preprocessing the sky images to obtain cloud information, the preprocessing including, based on the characteristics of the desert environment, eliminating image blurring and color distortion caused by dust aerosols, and separating cloud and dust information; based on the cloud information... The system uses image processing and inversion algorithms to determine the specific location and shape of the cloud layer in space. A typical aerosol optical thickness model of the Shagohuang region is introduced, which, along with the cloud layer's optical characteristics, serves as the input parameters for Monte Carlo ray tracing. The solar trajectory tracking method is used to calculate the real-time position of the sun in the sky based on geographical location, date, and time. The Buie distribution model is used as the shape model of the sun, describing the characteristics of sunlight with a non-uniform intensity distribution. The Monte Carlo method is used to simulate ray tracing, simulating cloud obstruction and scattering, as well as the absorption and scattering effects of dust aerosols on solar radiation, to calculate the solar irradiance distribution at different locations on the ground of the Shagohuang new energy base in the presence of clouds. The input boundary parameters and input source term parameters of the atmosphere within the monitoring area are obtained in order to set the atmospheric boundary conditions and atmospheric material source terms of the LBM model. Solve the LBM model to obtain detailed simulation results of carbon dioxide concentration distribution, flow velocity distribution, temperature distribution and pollutant concentration distribution in the Shagohuang New Energy Base, and calculate the carbon flux of the Shagohuang New Energy Base based on the detailed simulation results.

2. The carbon flux measurement and inversion method for the Shagohuang new energy base according to claim 1, characterized in that, Based on the detailed simulation results, the carbon flux of the Shagohuang new energy base is calculated, including: The Shagohuang New Energy Base was divided into several zones, resulting in multiple zone types. For each partition type, the first carbon flux is calculated using the gradient diffusion method based on the carbon dioxide concentration distribution and the flow velocity distribution, and the second carbon flux is calculated using the eddy covariance method. The first carbon flux and the second carbon flux are weighted and summed to obtain the basic carbon flux corresponding to each partition type; Based on the correction factor for each partition type, the basic carbon flux is corrected to obtain the corrected carbon flux for each partition type; The carbon flux of the Shagohuang New Energy Base is obtained by summing the area-weighted values ​​of the modified carbon flux for all partition types and then dividing the sum by the total area.

3. The carbon flux measurement and inversion method for the Shagohuang new energy base according to claim 2, characterized in that, The Shagohuang New Energy Base was divided into several zones, resulting in multiple zone types, including: Acquire multi-source data covering the Shagohuang New Energy Base; wherein, the multi-source data includes at least one of digital elevation model data, multispectral remote sensing imagery, hyperspectral remote sensing imagery, wind turbine coordinates and wind resource meteorological data, and equipment deployment drawings; The digital elevation model data is preprocessed, the terrain complexity index of each computational grid is calculated, and the computational grid is initially partitioned based on a preset terrain threshold to obtain the first partition result. Radiometric calibration and atmospheric correction are performed on the multispectral remote sensing image. The normalized vegetation index value is calculated pixel by pixel. The vegetation cover level is divided according to the preset vegetation cover level. The calculation grid is initially partitioned based on the vegetation cover level to obtain the second partition result. Radiometric calibration and atmospheric correction are performed on the hyperspectral remote sensing image. Characteristic high reflectance peaks of photovoltaic panels in the near-infrared band are extracted. Feature enhancement is performed on the characteristic high reflectance peaks. Photovoltaic panel pixels are identified by combining the preset reflectance peak threshold. The preliminary range of photovoltaic power generation area is formed by post-processing. The computational grid is preliminarily partitioned to obtain the third partition result. Using the wind turbine coordinates, the wind resource meteorological data, and the digital elevation model data, a customized wake model combined with a wind speed loss threshold is used to calculate the wind turbine wake influence range, and the computational grid is initially partitioned to obtain the fourth partition result. The coordinates of the facility elements in the equipment deployment drawings are transformed to a coordinate system consistent with the multi-source data, a facility element vector layer is generated, and the computational grid is initially partitioned to obtain the fifth partition result; Based on the results of the first to the fifth partitioning, a comprehensive logical judgment is performed on each computing grid to determine the final partition type of each computing grid.

4. The carbon flux measurement and inversion method for the Shagohuang new energy base according to claim 3, characterized in that, Calculate the terrain complexity index for each computational grid, and perform preliminary partitioning of the computational grid based on a preset terrain threshold to obtain the first partitioning result, including: The slope value is calculated within a specified first moving window using the maximum elevation difference method or the Horn algorithm; the aspect value is calculated within the first moving window using a preset aspect calculation formula; the terrain undulation is calculated within a specified second moving window using a preset undulation calculation formula; and the first partition result is determined based on the slope value, the aspect value, and the terrain undulation. The multispectral remote sensing image is radiometrically calibrated and atmospherically corrected. The normalized vegetation index (NDI) is calculated pixel-by-pixel. Vegetation cover levels are then defined based on preset vegetation cover thresholds. Combined with these vegetation cover levels, the computational grid is preliminarily partitioned to obtain a second partitioning result, including: Using a preset normalized vegetation index (NVI) calculation formula, the NVI value is calculated, and based on the range to which the NVI value belongs, the vegetation cover level and the second zoning result corresponding to the vegetation cover level are determined.

5. The carbon flux measurement and inversion method for the Shagohuang new energy base according to claim 3, characterized in that, Using the wind turbine coordinates, the wind resource meteorological data, and the digital elevation model data, a customized wake model combined with a wind speed deficit threshold is employed to calculate the wind turbine wake influence range. The computational grid is then preliminarily partitioned to obtain the fourth partition result, including: Centered on each wind turbine, along the dominant and secondary wind directions, the farthest distance that can be reached when the wind speed loss reaches the wind speed loss threshold is calculated according to the customized wake model for each wind direction sector as the dynamic wake range. Based on the dynamic wake range, a corresponding wind direction sector buffer is generated for each wind turbine; Spatially overlay all wind direction sector buffer zones of all wind turbines to generate a comprehensive wake influence zone layer for the entire base; The results of the fourth partition are obtained by overlaying the comprehensive wake influence zone layer with the computational grid. The function corresponding to the customized wake model is: in, This represents the wind speed at a distance x downwind. Indicates the incoming wind speed. This indicates the thrust coefficient of the wind turbine. This represents the rotor diameter of the wind turbine, and x represents the downwind distance. This represents the wake attenuation coefficient, which is dynamically adjusted based on atmospheric stability.

6. The carbon flux measurement and inversion method for the Shagohuang new energy base according to claim 3, characterized in that, Also includes: Based on the differentiated settings for each zone type and meteorological interference factors, the probability distribution models of the eddy diffusion coefficient, the vertical concentration gradient of carbon dioxide, and the correction coefficient are obtained; wherein, the meteorological interference factors include at least aerosol optical interference caused by strong winds and sandstorms, and intense near-surface thermal turbulence caused by strong solar radiation; Monte Carlo random perturbation iteration is performed based on the probability distribution model with the aforementioned differential settings to obtain multiple candidate carbon fluxes; After sorting the multiple candidate carbon fluxes, the top n and bottom m candidate carbon fluxes are removed. The minimum and maximum candidate carbon fluxes among the remaining candidate carbon fluxes constitute the confidence interval of the carbon flux of the Shagohuang New Energy Base.

7. The carbon flux measurement and inversion method for the Shagohuang New Energy Base according to any one of claims 2-6, characterized in that, The second carbon flux was calculated using the eddy covariance method, including: Based on the atmospheric turbulence fluctuation characteristics, high-frequency data from multiple data points within a preset time period are extracted; wherein, the sampling time interval of the high-frequency data is a preset time interval, and the high-frequency data includes high-frequency vertical flow velocity extracted from the flow velocity distribution, and carbon dioxide concentration extracted from the carbon dioxide concentration distribution; The high-frequency data sequence is quality marked and preprocessed, including at least removing abnormal values ​​of carbon dioxide concentration signal caused by aerosol concentration abnormalities due to strong winds and sandstorms in the desert environment. The high-frequency vertical velocity of each data point in the sequence is subtracted from the mean vertical velocity to obtain the pulsation value of the high-frequency vertical velocity of each data point. The carbon dioxide concentration of each data point is subtracted from the mean carbon dioxide concentration of all data points to obtain the pulsation value of the carbon dioxide concentration of each data point. The covariance is calculated as the second carbon flux based on the high-frequency vertical velocity fluctuation value and the carbon dioxide concentration fluctuation value of each data point.

8. A carbon flux measurement and inversion system for a desert-based new energy base, characterized in that, include: The construction module is used to acquire the physical parameters, terrain data and image data of the Shagohuang New Energy Base, so as to establish a three-dimensional geometric model of the Shagohuang New Energy Base using three-dimensional reconstruction technology. The generation module is used to perform spatial lattice processing on the region represented by the three-dimensional geometric model to generate the computational grid required for the LBM model; The first setting module is used to input the image data into a pre-trained ground parameter model for identification, and output ground boundary parameters and ground source term parameters corresponding to each grid cell in the computational grid, so as to set the ground boundary conditions and ground material source terms of the LBM model; wherein, the image data is a multi-temporal remote sensing image sequence covering the Shagohuang New Energy Base; the ground parameter model is a spatiotemporal sequence prediction model built based on the Transformer architecture; inputting the image data into the pre-trained ground parameter model for identification and outputting ground boundary parameters and ground source term parameters corresponding to each grid cell in the computational grid includes: processing the multi-temporal remote sensing image sequence, extracting the dynamic change characteristics of surface cover over time in each location, and outputting the ground boundary parameters and ground source term parameters corresponding to each computational grid. Boundary parameters and surface source term parameters; wherein, the surface source term parameters are dynamic parameters that change over time, and the model simultaneously outputs the time variation pattern of these parameters; wherein, the construction process of the sample label library of the surface parameter model includes: obtaining labels for the surface boundary parameters; the surface boundary parameters are static parameters characterizing the inherent physical properties of the underlying surface; obtaining labels for the surface source term parameters, which are dynamic parameters characterizing energy and mass exchange fluxes, and whose acquisition depends on high-frequency observations; constructing a sample label database: the data with a confidence level greater than a preset level obtained from the labels of the surface boundary parameters, remote sensing inversion, and model calculations, are time-synchronized and spatially gridded, and matched with the input data of the corresponding multi-temporal remote sensing image sequences to form a sample label database for training and validation of the surface parameter model; The second setting module is used to monitor cloud information using an all-sky imager, combine it with a high-precision solar trajectory tracking model, and calculate the ground irradiance distribution under cloud cover conditions using the Monte Carlo method, in order to set the ground thermal radiation source term of the LBM model. Specifically, monitoring cloud information using an all-sky imager and calculating the ground irradiance distribution under cloud cover conditions using the Monte Carlo method includes: continuously acquiring sky images of the monitoring area by deploying the all-sky imager; preprocessing the sky images to obtain cloud information, the preprocessing including, based on the characteristics of the desert environment, eliminating image blurring and color distortion caused by sand and dust aerosols, and separating cloud and sand / dust information; based on... The cloud information is used to determine the specific location and shape of the cloud layer in space using image processing and inversion algorithms. A typical aerosol optical thickness model of the Shagohuang region is introduced and used together with the optical characteristics of the cloud layer as input parameters for Monte Carlo ray tracing. The solar trajectory tracking method is used to calculate the real-time position of the sun in the sky based on geographical location, date, and time. The Buie distribution model is used as the shape model of the sun to describe the characteristics of sunlight with non-uniform light intensity distribution. The Monte Carlo method is used to simulate ray tracing, simulating the obstruction and scattering of cloud layers, and simultaneously simulating the absorption and scattering effects of dust aerosols on solar radiation, to calculate the distribution of solar irradiance intensity at different locations on the ground of the Shagohuang new energy base in the presence of clouds. The third setting module is used to acquire the input boundary parameters and input source term parameters of the atmosphere within the monitoring area, so as to set the atmospheric boundary conditions and atmospheric material source terms of the LBM model. The solution module is used to solve the LBM model, obtain detailed simulation results of carbon dioxide concentration distribution, flow velocity distribution, temperature distribution and pollutant concentration distribution in the Shagohuang New Energy Base, and calculate the carbon flux of the Shagohuang New Energy Base based on the detailed simulation results.

Images