A method, medium, apparatus, and product for tracing a source of carbon emissions

Abstract

Embodiments of the present application disclose a carbon emission source tracing method, medium, device and product. The carbon emission source tracing method comprises: acquiring multi-source prior data of a monitoring area and carbon concentration data of a plurality of concentration sampling points; determining a feature vector of each grid unit according to the multi-source prior data and the carbon concentration data; determining a carbon emission flux value of each grid unit based on the feature vector of each grid unit, and obtaining a total regional carbon emission amount by summarizing; performing spatial superposition analysis on the monitoring area according to the multi-source prior data, the feature vector of each grid unit and the carbon emission flux value of each grid unit, to determine a carbon emission source type and a spatial position of each carbon emission source; and determining a contribution degree of each carbon emission source to the total regional carbon emission amount according to the carbon emission flux value of each grid unit, the spatial position of each carbon emission source and the total regional carbon emission amount. Embodiments of the present application realize full-chain tracing of emission source positioning, type identification and contribution degree analysis.

Description

Technical Field

[0001] This application belongs to the field of atmospheric environment monitoring and carbon emission control, and specifically relates to a method for tracing carbon emission sources, a storage medium, an electronic device, and a computer program product. Background Technology

[0002] Precise quantification, flux accounting, and source apportionment of regional carbon emissions have become core requirements for greenhouse gas management. Existing regional carbon emission monitoring technologies are mainly divided into three categories: ground-based fixed stations, satellite remote sensing inversion, and UAV mobile aerial surveys. Among them, ground-based fixed stations have limited spatial coverage and cannot achieve refined accounting of regional area source emissions. Satellite remote sensing monitoring is limited by long revisit periods, low spatial resolution, and susceptibility to atmospheric and cloud interference, making it difficult to meet the needs of high-frequency, refined monitoring in small and medium-scale regions. UAV mobile aerial survey technology, with its advantages of mobility, high spatial resolution, and no limitation by complex terrain, has become the core development direction for carbon emission monitoring in small and medium-scale regions.

[0003] However, existing drone carbon emission monitoring technologies mainly focus on improving the accuracy of regional carbon emission accounting, and cannot form a full-chain traceability capability that covers the spatial location of emission sources and the analysis of their contribution.

[0004] Therefore, there is a lack of a full-process traceability method in the existing technology that can connect regional total emission accounting with accurate source tracing. Summary of the Invention

[0005] The purpose of this application is to provide a method, medium, device, and product for tracing carbon emission sources, which can solve the problem that existing UAV carbon emission monitoring technologies cannot form a full-chain tracing capability that covers the spatial location and contribution analysis of emission sources.

[0006] In a first aspect, embodiments of this application provide a method for tracing carbon emission sources, the method comprising: Acquire multi-source prior data of the monitoring area, as well as carbon concentration data of multiple concentration sampling points within the monitoring area; the monitoring area includes multiple grid cells, and there is a correspondence between the concentration sampling points and the grid cells; Based on the multi-source prior data and the carbon concentration data, a feature vector for each grid cell is determined; the feature vector for each grid cell is used to characterize the carbon emission attributes of each grid cell. The carbon emission flux value of each grid cell is determined based on the feature vector of each grid cell, and the carbon emission flux values ​​of all grid cells are summed to obtain the total regional carbon emissions of the monitoring area. Based on the multi-source prior data, the feature vector of each grid cell, and the carbon emission flux value of each grid cell, a spatial overlay analysis is performed on the monitoring area to determine the carbon emission source type and spatial location of each carbon emission source within the monitoring area. The contribution of each carbon emission source to the total carbon emissions of the region is determined based on the carbon emission flux value of each grid cell, the spatial location of each carbon emission source, and the total carbon emissions of the region.

[0007] Optionally, acquiring multi-source prior data of the monitoring area and carbon concentration data from multiple concentration sampling points within the monitoring area includes: Acquire geographic information data, satellite remote sensing data, meteorological data, and baseline concentration data for the monitored area; The satellite remote sensing data is segmented and classified to obtain the type and spatial geographic boundary of each potential emission source within the monitoring area; An initial carbon emission source inventory of the monitoring area is determined based on the type and spatial geographical boundaries of each potential emission source within the monitoring area; Based on the baseline concentration data, hotspot areas were identified in the monitoring area; The hotspot area is scanned to determine the hotspot concentration data within the hotspot area; The geographic information data, the initial carbon emission source inventory, and the meteorological data are identified as the multi-source prior data. Based on the baseline concentration data and the hotspot concentration data, the carbon concentration data of multiple concentration sampling points within the monitoring area are determined.

[0008] Optionally, the concentration sampling points have corresponding spatial location data within the monitoring area, and determining the feature vector of each grid cell based on the multi-source prior data and the carbon concentration data includes: The carbon concentration data of each concentration sampling point is bound to the spatial location data to generate a concentration sampling point dataset with spatial coordinates and timestamps; The multi-source prior data is spatially registered and temporally synchronized with the concentration sampling point dataset to obtain multi-source feature data. From the multi-source feature data, extract the spatial geographic features, meteorological features, source strength prior features and concentration features corresponding to each concentration sampling point to obtain the feature set of each concentration sampling point; Based on the correspondence between the concentration sampling points and the grid cells, the feature set of each concentration sampling point is mapped to the corresponding grid cell to obtain the feature vector corresponding to each grid cell.

[0009] Optionally, determining the carbon emission flux value of each grid cell based on the feature vector of each grid cell, and summing the carbon emission flux values ​​of all grid cells to obtain the total regional carbon emissions of the monitoring area, includes: The measured flux value of each concentration sampling point is calculated based on the gridded meteorological data and the carbon concentration data in the multi-source prior data. The pre-built carbon emission flux inversion model is trained and optimized based on the spatial location of each grid cell in the monitoring area, the feature set of each concentration sampling point, and the measured flux value of each concentration sampling point to obtain the target carbon emission flux inversion model. The carbon emission flux value of each grid cell is obtained based on the feature vector of each grid cell and the target carbon emission flux inversion model. The total carbon emission flux values ​​of all grid cells are summed to obtain the total regional carbon emissions of the monitored area.

[0010] Optionally, the step of training and optimizing the pre-constructed carbon emission flux inversion model based on the spatial location of each grid cell in the monitoring area, the feature set of each concentration sampling point, and the measured flux value of each concentration sampling point to obtain the target carbon emission flux inversion model includes: The feature weight of each grid cell in the target carbon emission flux inversion model is determined based on the spatial location of each grid cell in the monitoring area. The pre-constructed carbon emission flux inversion model is trained and optimized based on the feature weights of each grid cell, the feature set of each concentration sampling point, and the measured flux value of each concentration sampling point to obtain the target carbon emission flux inversion model.

[0011] Optionally, the step of performing spatial overlay analysis on the monitoring area based on the multi-source prior data, the feature vector of each grid cell, and the carbon emission flux value of each grid cell to determine the carbon emission source type and spatial location of each carbon emission source within the monitoring area includes: The concentration ratio feature of each grid cell is calculated based on the concentration features in the feature vector of each grid cell; The concentration ratio feature of each grid cell is matched with a preset emission source feature library to determine the emission source type identification result of each grid cell. Based on the carbon emission flux value of each grid cell, target grid areas with carbon emission flux values ​​higher than a preset threshold are identified. The spatial overlay analysis of the emission source type identification results of each grid cell, the target grid area, and the initial carbon emission source list in the multi-source prior data is performed to determine the spatial location of each carbon emission source. The set of grid cells covered by each carbon emission source is determined based on the spatial location of each carbon emission source, and the carbon emission source type of each carbon emission source is determined based on the emission source type identification results of all grid cells in the set of grid cells.

[0012] Optionally, determining the contribution of each carbon emission source to the total carbon emissions of the region based on the carbon emission flux value of each grid cell, the spatial location of each carbon emission source, and the total carbon emissions of the region includes: For each carbon emission source, the carbon emission flux values ​​of all grid cells in the grid cell set are summed to obtain the carbon emission amount of each carbon emission source. The contribution of each carbon emission source to the total carbon emissions of the region is determined based on the carbon emissions of each carbon emission source and the total carbon emissions of the region.

[0013] Secondly, embodiments of this application provide a storage medium that stores computer instructions, which, when executed by a computer, are used to perform the steps of the carbon emission source tracing method as described in the first aspect.

[0014] Thirdly, embodiments of this application provide an electronic device, including at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the carbon emission source tracing method as described in the first aspect.

[0015] Fourthly, embodiments of this application provide a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the carbon emission source tracing method as described in the first aspect.

[0016] In this embodiment, the monitoring area is divided into grid cells, and the feature vector corresponding to each grid cell is determined. This achieves a scale transformation from discrete sampling points to continuous spatial cells, enabling monitoring data to be represented in a unified spatial grid. Based on the feature vector of each grid cell, its carbon emission flux value is determined, and the total regional carbon emissions are obtained by summarizing these values. This completes the mapping from features to carbon emission intensity and realizes the calculation of the total regional carbon emissions. Furthermore, by performing spatial overlay analysis on the monitoring area, the carbon emission source type and spatial location of each carbon emission source are determined. The flux inversion results are fused with multi-source prior data, realizing the source tracing and location from the total regional emissions to individual carbon emission sources. Based on the carbon emission flux value of each grid cell, the spatial location of each carbon emission source, and the total regional carbon emissions, the contribution of each carbon emission source to the total regional emissions can be determined. This solves the problem that existing UAV carbon emission monitoring technologies cannot form a full-chain source tracing capability that covers the spatial location and contribution analysis of carbon emission sources, realizing the entire process of source tracing from the calculation of the total regional carbon emissions to the accurate source tracing of carbon emission sources. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the steps of a carbon emission source tracing method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of an unmanned aerial vehicle (UAV) swarm system provided in an embodiment of this application; Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0020] The carbon emission source tracing method provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0021] Reference Figure 1 This is a flowchart illustrating the steps of a carbon emission source tracing method provided in this application embodiment, specifically including the following steps: Step 101: Obtain multi-source prior data of the monitoring area, and carbon concentration data of multiple concentration sampling points within the monitoring area; the monitoring area includes multiple grid units, and the concentration sampling points correspond to the grid units; In this embodiment, multi-source prior data refers to background information data collected within the monitoring area through methods such as geographic information acquisition, satellite remote sensing inversion, and meteorological observation. This includes, but is not limited to, standardized geographic data, initial carbon emission source inventories, and gridded meteorological data, providing spatial, meteorological, and source strength prior information for subsequent feature extraction and flux inversion. Concentration sampling points refer to the specific spatial locations where the UAV swarm system collects volumetric concentration data of key carbon components along its flight path during baseline scanning and encrypted tracking missions. The key carbon components include CO2, CH4, and CO. Carbon concentration data refers to the measured volumetric concentration values ​​of CO2, CH4, and CO at each concentration sampling point.

[0022] To map discrete concentration sampling data to a continuous spatial scale, the monitoring area needs to be spatially discretized using a regular grid, resulting in multiple raster cells. Specifically, using the CGCS2000 national geodetic coordinate system as a reference, the monitoring area is divided into regular rectangular raster grids according to a preset spatial resolution, with each raster cell having a unique spatial index and geographic boundary.

[0023] It should be noted that each concentration sampling point falls into a unique corresponding grid cell based on its spatial location coordinates through spatial registration; a grid cell may contain multiple sampling points or may not contain any sampling points.

[0024] In existing technologies, UAV carbon emission monitoring mostly focuses on monitoring single greenhouse gases such as methane or carbon dioxide, lacking simultaneous collection and correlation analysis of carbon monoxide, an indicator gas of incomplete combustion. This makes it impossible to distinguish different types of emission sources based on multi-gas concentration ratio characteristics, and the monitoring data can only support preliminary calculations of total carbon emissions, making it difficult to achieve accurate source analysis and tracing. This application's embodiment uses a UAV swarm system to simultaneously collect volumetric concentration data of three key carbon components: CO2, CH4, and CO, providing a data foundation for subsequent differentiation of emission source types based on multi-gas concentration ratio characteristics.

[0025] Step 102: Based on the multi-source prior data and the carbon concentration data, determine the feature vector of each grid cell; the feature vector of each grid cell is used to characterize the carbon emission attribute of each grid cell; In this embodiment, the feature vector of each grid cell is used to characterize the carbon emission attribute of each grid cell. That is, the feature vector of each grid cell refers to a multi-dimensional data set of carbon emission-related attributes of each grid cell. Specifically, by spatially registering and fusing multi-source prior data with carbon concentration data from concentration sampling points, each grid cell is assigned a feature vector that includes spatial geographic features, meteorological features, source strength prior features, and concentration features. Spatial geographic features, meteorological features, source strength prior features, and concentration features are all related to carbon emission attributes.

[0026] Step 103: Determine the carbon emission flux value of each grid cell based on the feature vector of each grid cell, and sum up the carbon emission flux values ​​of all grid cells to obtain the total regional carbon emissions of the monitoring area. In this embodiment, the carbon emission flux value of each grid cell refers to the amount of carbon emissions passing through that grid cell per unit time and per unit area, and is an indicator for measuring the carbon emission intensity of that grid cell. The total regional carbon emissions refer to the overall regional carbon emissions obtained by spatially summarizing the carbon emission flux values ​​of all grid cells within the monitoring area, reflecting the total carbon emission level of the monitoring area over a certain period of time.

[0027] Step 104: Based on the multi-source prior data, the feature vector of each grid cell, and the carbon emission flux value of each grid cell, perform spatial overlay analysis on the monitoring area to determine the carbon emission source type and spatial location of each carbon emission source within the monitoring area; In this embodiment, multi-source data needs to be overlaid in layers within the same geographic coordinate system. Spatial overlay analysis is achieved through comprehensive analysis of the spatial relationships between the layers. Specifically, spatial overlay comparison is performed based on multi-source prior data, the feature vector of each raster cell, and the carbon emission flux value of each raster cell to identify the carbon emission source type and spatial location of each carbon emission source.

[0028] Among them, carbon emission source type refers to the category to which the carbon emission source belongs, such as coal combustion source, gas leak source, transportation mobile source, agricultural / biological emission source, etc. Spatial location refers to the geographical coordinates of the carbon emission source within the monitoring area and the geographical boundary range it covers.

[0029] Step 105: Determine the contribution of each carbon emission source to the total carbon emissions of the region based on the carbon emission flux value of each grid cell, the spatial location of each carbon emission source, and the total carbon emissions of the region.

[0030] In this embodiment of the application, in order to quantitatively assess the responsibility of each carbon emission source in regional carbon emissions, after determining the spatial location of each carbon emission source, it is also necessary to obtain the carbon emission amount of the emission source based on the spatial location of each emission source and the carbon emission flux value of each grid cell, and then calculate its proportion of the total regional carbon emissions to obtain the contribution of each carbon emission source to the total regional carbon emissions.

[0031] In this embodiment, the feature vector corresponding to each grid cell is determined, realizing the scale transformation from discrete sampling points to continuous spatial cells. This allows monitoring data to be represented in a unified spatial grid. Based on the feature vector of each grid cell, its carbon emission flux value is determined and aggregated to obtain the total regional carbon emissions, completing the mapping from features to carbon emission intensity and realizing the calculation of the total regional carbon emissions. Furthermore, by performing spatial overlay analysis on the monitoring area, the carbon emission source type and spatial location of each carbon emission source are determined. The flux inversion results are fused with multi-source prior data, realizing the source tracing and location from the total regional emissions to individual carbon emission sources. Based on the carbon emission flux value of each grid cell, the spatial location of each carbon emission source, and the total regional carbon emissions, the contribution of each carbon emission source to the total regional emissions can be determined. This solves the problem that existing UAV carbon emission monitoring technologies cannot form a full-chain source tracing capability covering the spatial location and contribution analysis of carbon emission sources, realizing the entire process of source tracing from the calculation of the total regional carbon emissions to the accurate source tracing of carbon emission sources.

[0032] In one embodiment of this application, acquiring multi-source prior data of the monitoring area and carbon concentration data from multiple concentration sampling points within the monitoring area includes: Acquire geographic information data, satellite remote sensing data, meteorological data, and baseline concentration data for the monitored area; The satellite remote sensing data is segmented and classified to obtain the type and spatial geographic boundary of each potential emission source within the monitoring area; An initial carbon emission source inventory of the monitoring area is determined based on the type and spatial geographical boundaries of each potential emission source within the monitoring area; Based on the baseline concentration data, hotspot areas were identified in the monitoring area; The hotspot area is scanned to determine the hotspot concentration data within the hotspot area; The geographic information data, the initial carbon emission source inventory, and the meteorological data are identified as the multi-source prior data. Based on the baseline concentration data and the hotspot concentration data, the carbon concentration data of multiple concentration sampling points within the monitoring area are determined.

[0033] In this embodiment of the application, high-precision geographic information data of the monitoring area is first collected. The high-precision geographic information data includes a digital elevation model (DEM) with a spatial resolution of ≥5m and a digital orthophoto map (DOM). The plane coordinate system of the high-precision geographic information data is configured as the CGCS2000 national geodetic coordinate system, and the elevation datum is configured as the 1985 national elevation datum, so as to achieve the unification of the coordinate datum of all spatial data.

[0034] To eliminate noise and distortion in the raw geographic information data and ensure it meets the accuracy requirements for subsequent spatial analysis and feature extraction, high-precision geographic information data needs to be preprocessed to obtain standardized geographic data. Specifically, depression filling is performed on the DEM data to eliminate elevation anomalies. Based on the processed DEM data, slope and aspect raster data for the monitoring area are calculated and generated, ensuring that the spatial resolution of the slope and aspect data is consistent with that of the DEM data. Radiometric calibration, atmospheric correction, and geometric fine correction are performed on the DOM data, ensuring that the planar error of the corrected DOM data does not exceed one pixel. For ease of description, the processed DEM data and the corrected DOM data are defined as standardized geographic data.

[0035] In this embodiment, it is also necessary to acquire multi-temporal satellite remote sensing data of the monitoring area, and to complete image segmentation and land cover classification using object-based image analysis (OBIA) to obtain the types and spatial geographic boundaries of each potential emission source within the monitoring area. Simultaneously, an initial carbon emission source inventory of the monitoring area is constructed. The satellite remote sensing data includes, but is not limited to, Sentinel-2 visible light imagery, Landsat thermal infrared imagery, and Suomi-NPP nighttime light remote sensing imagery.

[0036] It should be noted that potential emission sources are candidate objects initially identified based on satellite remote sensing data in terms of spatial morphology and land cover categories, while the aforementioned carbon emission sources are emission responsibility entities finally determined through spatial overlay analysis by integrating multi-source prior data, UAV measured concentration data, and carbon emission flux inversion results. Among them, potential emission sources can be understood as the initial candidate set and spatial reference benchmark for carbon emission sources, while carbon emission sources are the confirmed results of potential emission sources after multi-source data verification and flux quantification.

[0037] In one embodiment, the object-oriented image analysis method can be an image segmentation algorithm and an object classification algorithm. The image segmentation algorithm can be any one of the following: multi-scale segmentation algorithm, mean-shift segmentation algorithm, quadtree segmentation algorithm, and watershed segmentation algorithm. The object classification algorithm can be any one of the following: random forest method, support vector machine (SVM), K-Nearest Neighbor (KNN), CART (Classification and Regression Tree) decision tree classification method, and U-Net (a fully convolutional neural network-based image segmentation network) / MaskR-CNN (Mask Region-based Convolutional Neural Network) instance segmentation algorithm. It should be noted that this embodiment does not limit the scope of the method; those skilled in the art can choose according to actual needs. Furthermore, object-oriented image analysis methods are mature technologies in the field, and therefore will not be elaborated further here.

[0038] Preferably, based on the type and spatial geographic boundaries of each potential emission source within the monitoring area, the activity level coefficient of each potential emission source is calculated simultaneously; the source strength prior characteristic data of each grid cell within the monitoring area is calculated; and an initial carbon emission source inventory of the monitoring area is constructed based on the type, spatial geographic boundaries, and activity level coefficients of the carbon emission sources. The source strength prior characteristic data includes the land cover type, building density, normalized difference vegetation index (NDVI), land surface temperature, distance to the nearest potential emission source, and the activity level coefficient of the corresponding potential emission source for the grid cell where the concentration sampling point is located.

[0039] Specifically, the activity level coefficient for each potential emission source is calculated using the following formula:

[0040] Wherein, α is the activity level coefficient for each potential emission source, ranging from 0 to 1; L i L represents the average nighttime solar radiation intensity in the region corresponding to the potential emission source. max The maximum value of nighttime solar radiation intensity within the monitoring area; T i The average surface temperature (T) of the area corresponding to the potential emission source; b To monitor the average background surface temperature of the area; T max The maximum surface temperature within the monitoring area is represented by w1 and w2, which are weighting coefficients; for example, w1 = 0.4 and w2 = 0.6. It should be noted that L... i and Ti The value of L is obtained by averaging the pixel values ​​of all raster units within the boundary after spatially overlaying the spatial vector boundary of a potential emission source with nighttime light remote sensing imagery and thermal infrared remote sensing imagery; max and T max The value of T is obtained by taking the maximum value of the pixel values ​​of nighttime solar radiation intensity and surface temperature across the entire monitoring area; b This is obtained by averaging the surface temperature pixel values ​​across the entire monitoring area.

[0041] Land cover type was obtained by object-oriented image classification of DOM data and satellite remote sensing data; building density was obtained by extracting building outlines and statistically analyzing the proportion of building area in the region from DOM data; normalized vegetation index (NDVI) was obtained by normalizing the near-infrared and red bands of Sentinel-2 and other multispectral remote sensing images; land surface temperature was obtained by inverting the radiative transfer equation from Landsat thermal infrared remote sensing images; and the distance to the nearest potential emission source was determined by calculating the shortest spatial distance between the center point of the raster cell and the spatial vector boundary of all potential emission sources.

[0042] At this point, an initial carbon emission source inventory is constructed based on the type of potential emission sources, spatial geographic boundaries, and activity level coefficients. Each record in the initial carbon emission source inventory includes six core fields: a unique emission source ID, emission source type, spatial vector boundary, activity level coefficient, prior emission intensity level, and corresponding remote sensing image time phase. The prior emission intensity level is determined based on the activity level coefficient of the emission source. Specifically, an activity level coefficient ≥ 0.7 indicates a high emission intensity level, 0.3 ≤ activity level coefficient < 0.7 indicates a medium emission intensity level, and an activity level coefficient < 0.3 indicates a low emission intensity level.

[0043] In this embodiment of the application, it is also necessary to acquire meteorological data monitored by ground meteorological stations within the monitoring area, and to assimilate the meteorological data based on a mesoscale meteorological forecasting model to generate gridded meteorological data. The meteorological data includes wind speed, wind direction, atmospheric temperature, relative humidity, and air pressure data within the monitoring period. The temporal resolution of the gridded meteorological data needs to be less than or equal to 1 minute, and the spatial resolution needs to be less than or equal to 30 meters.

[0044] In one embodiment, a three-dimensional variational assimilation (3D-Var) method can be used. Ground-based meteorological station data is used as the observation field, and the simulation results of a mesoscale meteorological forecasting model are used as the background field. The analysis field is obtained by minimizing the cost function, thus achieving the assimilation of meteorological data and generating gridded meteorological data. The simulation results of the mesoscale meteorological forecasting model are forecast field data obtained by running the model in advance before the monitoring operation begins, based on historical meteorological data of the monitoring area and global reanalysis data. The assimilation process involves statistically optimally fusing the observed data and the model forecast data, ensuring that the final output gridded meteorological data retains the physical consistency of the model forecast while closely approximating the true state of the observed data.

[0045] To support the throughput calculation of concentration data at different flight altitudes, it is necessary to set up vertical layers in the gridded meteorological field that cover at least the drone's cruising altitude, for example, corresponding to altitudes of 20m, 50m, 100m, and 200m above the ground, with each altitude layer outputting corresponding meteorological data. The gridded meteorological field refers to a set of regularized data where meteorological data is discretized according to a unified spatial grid, and each grid point stores the values ​​of various meteorological parameters at that location. These meteorological parameters include wind speed, wind direction, temperature, air pressure, and humidity. The gridded meteorological data are the specific values ​​stored in the gridded meteorological field, that is, the quantified values ​​of each meteorological parameter at each grid point in each vertical layer.

[0046] Standardized geographic data, initial carbon emission source inventories, and gridded meteorological data were identified as multi-source prior data and used as background inputs for subsequent feature extraction and flux inversion.

[0047] In existing technologies, the stand-alone operation mode has inherent limitations, making it impossible to balance monitoring efficiency and accuracy. Current UAV aerial surveys generally adopt a stand-alone operation mode, which, due to limitations in payload and endurance, cannot simultaneously achieve wide-area full-coverage surveys and high-precision detailed surveys of key areas. Furthermore, most of them use preset fixed flight routes, which cannot automatically identify hotspots of concentration anomalies based on real-time monitoring data and conduct adaptive intensive monitoring. As a result, it is difficult to improve monitoring efficiency and spatial resolution simultaneously, leading to insufficient reliability and spatial resolution of flux inversion results in hotspot areas.

[0048] Reference Figure 2 This is a schematic diagram of the structure of an unmanned aerial vehicle (UAV) swarm system provided in this application. The UAV swarm system consists of a ground control station 100, a long-endurance fixed-wing mother aircraft 200, and several multi-rotor sub-aircraft 300s, wherein there are at least two multi-rotor sub-aircraft 300s.

[0049] This application embodiment utilizes a drone swarm system consisting of a fixed-wing mother drone 200 and a multi-rotor drone 300 to fully leverage the advantages of fixed-wing drones' long endurance and wide-range cruising capabilities, as well as the advantages of multi-rotor drones' low-altitude, low-speed, and flexible maneuverability, to achieve collaborative operations of wide-area full-coverage general surveys and high-precision detailed surveys of hotspot areas.

[0050] Specifically, the fixed-wing aircraft carrier 200 is equipped with a tunable diode laser absorption spectroscopy (TDLAS) gas sensing module 201. The TDLAS gas sensing module 201 is used to simultaneously collect volume concentration data of three key carbon components: CO2, CH4, and CO. The detection accuracy of the fixed-wing aircraft carrier 200 needs to meet the following requirements: CO2 measurement accuracy ±1ppm, CH4 measurement accuracy ±5ppb, and CO measurement accuracy ±2ppb. The endurance of the fixed-wing aircraft carrier 200 is ≥2h, and the cruising flight altitude is 100m-200m above the ground.

[0051] The multi-rotor sub-aircraft 300 is equipped with a lightweight non-dispersive infrared (NDIR) gas sensing module 301. The NDIR gas sensing module 301 is used to simultaneously collect volume concentration data of three key carbon components: CO2, CH4, and CO. The detection accuracy of the multi-rotor sub-aircraft 300 must meet the following requirements: CO2 measurement accuracy ±2ppm, CH4 measurement accuracy ±10ppb, and CO measurement accuracy ±5ppb. The multi-rotor sub-aircraft 300 has an endurance of ≥30min and a cruising flight altitude of 20-50m above the ground.

[0052] All UAVs in the UAV swarm system are equipped with a real-time dynamic kinematic (RTK) positioning module, with a horizontal positioning accuracy of ≤2cm and an elevation positioning accuracy of ≤5cm. They are also equipped with a wireless data transmission module to realize real-time data interaction and command transmission between the fixed-wing mother aircraft 200, the multi-rotor aircraft 300 and the ground control station 100.

[0053] In one embodiment, before acquiring carbon concentration data based on the drone swarm system, it is also necessary to perform on-site sensor calibration on the drones in the drone swarm system to ensure that the sensor detection accuracy meets the preset requirements.

[0054] Specifically, zero gas (high-purity nitrogen) and standard mixed gases of corresponding concentrations of CO2, CH4, and CO are used to calibrate the TDLAS gas sensing module 201 and the NDIR gas sensing module 301 on-site. The calibration coefficient f obtained from the standard gas calibration is recorded to ensure that the sensor detection accuracy meets the preset requirements.

[0055] In the process of collecting carbon concentration data through a drone swarm system, a fixed-wing mother drone 200 first performs a full-area baseline scan and identifies hotspot areas within the monitoring area. Specifically, the ground control station 100 plans a bow-shaped full-coverage baseline scan route for the fixed-wing mother drone 200, covering the entire monitoring area. The lateral spacing of the route is set to 100-200m, and the flight direction is perpendicular to the prevailing wind direction within the monitoring area. The fixed-wing mother drone 200 executes its flight mission according to the preset route, collecting real-time data on the concentration of key carbon components, spatial location data, and flight attitude data, and simultaneously transmitting this data back to the ground control station 100. The real-time data processing unit of the ground control station 100 analyzes the transmitted data in real time, marking areas where the concentration of key carbon components exceeds a preset threshold as hotspot areas with abnormal concentrations. The concentration threshold can be set to twice the average baseline scan concentration within the monitoring area.

[0056] In one embodiment, the prevailing wind direction is determined based on the arithmetic mean of wind directions measured at a height of 10m from ground meteorological stations within one hour prior to the monitoring operation. If the terrain of the monitoring area is complex, the prevailing wind direction is determined based on the regional prevailing wind direction simulated by a mesoscale meteorological forecasting model during the monitoring operation period. The fixed-wing mother aircraft 200 flies in a terrain-following mode based on DEM data, adjusting its flight altitude in real time according to the ground elevation to ensure that the actual flight altitude remains within a preset range of 100-200m. The baseline concentration average is the arithmetic mean of the corresponding gas component concentration data obtained from the mother aircraft's baseline scan. When the concentration value of any key carbon component (CO2, CH4, CO) exceeds the concentration threshold of the corresponding component, the area is determined to be a hotspot area with an abnormal concentration.

[0057] To ensure that the spatial resolution of the concentration sampling points meets the accuracy requirements of subsequent flux inversion, the sampling frequency of the TDLAS gas sensor module 201 can be set to 1-10Hz, and the sampling frequency of the NDIR gas sensor module 301 can be set to 5-20Hz. This ensures that the sampling frequency matches the flight speed of the UAV, thereby ensuring that the horizontal distance between adjacent sampling points does not exceed 10m. Simultaneously, the overlap rate of adjacent flight paths of the fixed-wing mother aircraft 200's zigzag flight path needs to be set to 10%-15% to ensure that there are no blind spots in the monitoring area.

[0058] The matching rule between the sampling frequency and the UAV flight speed is V=f sensor ×d, V is the flight speed of the fixed-wing aircraft 200, in m / s, f sensordenoted by Hz, where d is the sensor sampling frequency, and d is the preset horizontal spacing between adjacent sampling points, in meters. It should be noted that the matching rule here is based solely on the flight speed of the fixed-wing mother aircraft 200. This is because the flight speed of the fixed-wing mother aircraft 200 is relatively constant during bow-shaped baseline scanning and is the main variable determining the spatial distribution of the flight path sampling points. In contrast, the multi-rotor aircraft 300 experiences significant speed variations and complex flight patterns during encrypted scanning and aerodynamic tracking; its sampling point spatial distribution is determined by the actual flight trajectory, and therefore does not require matching calculations based on a preset speed.

[0059] After the fixed-wing mothership 200 completes the baseline scan of the entire area and identifies hotspot areas, the multi-rotor sub-aircraft 300 needs to perform intensive scanning and aerodynamic tracking of these hotspot areas to obtain hotspot concentration data. Specifically, the ground control station 100 assigns intensive detailed survey tasks to the nearest multi-rotor sub-aircraft 300 based on the number, spatial distribution, and area of ​​the identified hotspot areas, and plans a gridded intensive scanning route for the hotspot areas, with the lateral spacing of the route reduced to 10-30m. After receiving the mission command, the multi-rotor sub-aircraft 300 flies to the corresponding hotspot area to perform low-altitude intensive scanning, collecting high-resolution carbon key component concentration data and spatial location data in real time. During the intensive scanning process, the onboard processing unit of the multi-rotor sub-aircraft 300 identifies changes in concentration gradient in real time, dynamically adjusts its flight path, and performs aerodynamic tracking flight along the direction of increasing concentration to pinpoint the precise spatial location of the emission source.

[0060] The task allocation rule for the aforementioned multi-rotor sub-aircraft 300 is as follows: the ground control station 100 sets task priorities from high to low based on the concentration exceeding the threshold multiple of the hotspot area; the higher the multiple exceeding the threshold, the higher the task priority. For a single hotspot area, the number of sub-aircraft is allocated according to the area of ​​the hotspot, with priority given to areas ≤0.1km². 2 One multirotor sub-aircraft of 300 is allocated per hour, covering an area of ​​0.1-0.5 km². 2 Two multi-rotor sub-aircraft of 300 are allocated per hour, in an area ≥0.5 km². 2 At least three multi-rotor aircraft 300 shall be allocated per hour; the principle of allocating multi-rotor aircraft 300 based on proximity is to minimize the straight-line distance between the current position of the multi-rotor aircraft 300 and the center point of the hot spot area.

[0061] In one embodiment, the above-mentioned gas-taxis tracking method of the multi-rotor sub-aircraft 300 refers to the following: during the flight of the multi-rotor sub-aircraft 300 along a preset encrypted flight path, the concentration data of three consecutive sampling points are grouped together to calculate the direction and amplitude of the concentration gradient; when the concentration gradient amplitude exceeds the preset gradient threshold, the multi-rotor sub-aircraft 300 will pause flight and perform step-by-step flight in the direction of increasing concentration gradient at a step size of 5-10m, collecting concentration data once for each flight step size, until the concentration gradient changes from increasing to decreasing. At this time, the position corresponding to the concentration peak is the precise spatial position of the emission source; the preset gradient threshold can be set to 5% of the average value of the baseline concentration of the corresponding component.

[0062] It should be noted that during the sub-unit's aerodynamic tracking flight, the preset no-fly zone avoidance rules, minimum flight altitude limits, and maximum flight range limits must be strictly followed. If any of these limits are triggered, aerodynamic tracking will immediately terminate, and the sub-unit will return to the preset standby point. Specifically, the minimum flight altitude limit means that the multi-rotor sub-unit 300 must not fly below 10 meters above ground level or the minimum safe flight altitude for UAVs as stipulated by local regulations, to avoid collisions with ground obstacles and ensure flight safety. The maximum flight range limit means that the multi-rotor sub-unit 300's horizontal flight range must not exceed 100 meters outward from the boundary of its assigned hotspot area, to prevent excessive deviation leading to insufficient endurance or entry into the operating areas of other sub-units. The preset standby point refers to a pre-set spatial location at the edge of the monitoring area before the monitoring operation begins, for the multi-rotor sub-unit 300 to hover and await mission instructions after takeoff.

[0063] To ensure the spatial uniformity and representativeness of the encrypted scan data, the flight direction of the encrypted scan route of the multi-rotor sub-aircraft 300 needs to be perpendicular to the real-time prevailing wind direction within the hotspot area. After completing the encrypted scan task of the currently assigned hotspot area, the ground control station 100 will reassign the remaining unfinished hotspot area tasks. If there are no tasks to be performed, the multi-rotor sub-aircraft 300 will return to the preset standby point. At the same time, to ensure the coordination and safety of the cluster operation, after the fixed-wing mother aircraft 200 completes the baseline scan of the entire area, the fixed-wing mother aircraft 200 also needs to hover at a preset safe altitude over the monitoring area until all multi-rotor sub-aircraft 300s have completed all encrypted scan tasks and then return to base synchronously.

[0064] To eliminate errors introduced by instrument noise, environmental fluctuations, and changes in atmospheric conditions in the raw sensor measurement data, it is necessary to perform outlier removal, smoothing and denoising, and concentration calibration on the raw concentration values ​​collected by the UAV swarm system to obtain the calibrated actual concentration values. Specifically, the 3σ criterion can be used for outlier removal, eliminating outlier data points exceeding three times the standard deviation; a moving average filter can be used for smoothing and denoising, with the sliding window width set to five sampling points; and a calibration curve calibrated using a standard gas can be used for concentration calibration, performing a secondary calibration on the carbon concentration data converted from the collected raw voltage signal to eliminate the influence of ambient temperature and pressure on detection accuracy.

[0065] The original measured concentration value is then calibrated a second time using the following formula:

[0066] Among them, C corr C represents the actual concentration value after calibration. raw The original concentration value measured by the sensor is P0, which is the standard atmospheric pressure, i.e., 101.325 kPa. meas T represents the measured atmospheric pressure at the concentration sampling point, T0 represents the standard temperature (273.15 K), and T... meas The measured ambient thermodynamic temperature values ​​at the concentration sampling points are denoted as , and f is the calibration coefficient obtained from standard gas calibration. It can be understood that the baseline concentration data and hotspot concentration data refer to the actual concentration values ​​after calibration, i.e., the volume concentration values ​​of CO2, CH4, and CO at each concentration sampling point obtained after the above outlier removal, smoothing, noise reduction, and concentration calibration processes.

[0067] After the multi-rotor sub-aircraft 300 completes the hotspot area encryption scan and aerodynamic tracking flight and returns to base, it immediately uses the same zero gas as the pre-flight calibration and a standard mixture of CO2, CH4, and CO at the corresponding concentrations to perform post-flight zero-point and span calibration on the TDLAS gas sensor module 201 and the NDIR gas sensor module 301. If the zero-point drift of the sensors after flight exceeds 2% of the full scale, drift correction is applied to the full concentration data collected during this flight. The specific correction formula is as follows: C corr =C corr +ΔC Among them, C corr 'C' represents the final concentration value after drift correction. corr The value is the concentration value after correction by the calibration factor f, and ΔC is the zero-point drift.

[0068] It should be noted that during the entire flight mission of the fixed-wing mother aircraft 200 and all multi-rotor aircraft 300, the gas sensing modules and positioning modules of all UAVs must be synchronized with the ground meteorological station network with a synchronization accuracy of ≤10ms, so as to achieve the time reference unification of concentration data, location data and meteorological data, and provide a data foundation for subsequent data fusion.

[0069] Specifically, the fixed-wing mother aircraft 200 and all multi-rotor aircraft 300 take off synchronously before the monitoring operation begins. After takeoff, the multi-rotor aircraft 300 hovers at a preset standby point at the edge of the monitoring area. The straight-line distance between the preset standby point and the boundary of the monitoring area does not exceed 500m. Throughout the baseline scanning process performed by the fixed-wing mother aircraft 200, the multi-rotor aircraft 300 remains in a hovering standby state and receives mission instructions from the ground control station 100 in real time.

[0070] In one embodiment, the spatial resolution of the grid unit can be set to 5-30m. At the same time, the spatial resolution of the grid unit is not higher than the lateral spacing of the encrypted scanning route of the slave unit, so as to ensure that the scale of the grid unit matches the spatial resolution of the monitoring data, and avoid the loss of spatial details due to the grid scale being too coarse or the introduction of too much noise due to the scale being too fine.

[0071] Based on the baseline and hotspot concentration data mentioned above, carbon concentration data at multiple concentration sampling points within the monitoring area are determined. Specifically, baseline concentration data collected at all sampling points during the mother unit's baseline scan are used as background concentration sampling point data for the entire area, while hotspot concentration data collected at all sampling points during the slave unit's intensive scan and gas tracing are used as intensive concentration sampling point data for hotspot areas. Together, these two sets of data constitute carbon concentration data at multiple concentration sampling points within the monitoring area, which are then used for feature extraction and flux inversion in subsequent steps.

[0072] It should be noted that the parameters mentioned above, such as sampling frequency, lateral spacing of flight paths, flight speed, preset threshold, step size, and distance limit, which involve specific values, are all exemplary descriptions of the embodiments of this application. Those skilled in the art can reasonably adjust and optimize the above parameters according to actual monitoring needs, UAV performance parameters, and on-site environmental conditions. The embodiments of this application do not limit this.

[0073] In one embodiment, in order to visually demonstrate the spatial distribution characteristics of key carbon components within the monitoring area, a concentration field raster map can be generated based on the spatial relationship between each concentration sampling point and the raster unit, as well as the carbon concentration data of each concentration sampling point.

[0074] This application's embodiments, through the acquisition and standardized processing of high-precision geographic information data (DEM / DOM), provide a unified, distortion-free geographic base map and terrain feature data for subsequent spatial analysis. By segmenting and classifying multi-temporal satellite remote sensing data, combined with activity level coefficient calculation and source strength prior feature extraction, an initial carbon emission source inventory containing emission source spatial boundaries and prior emission intensities is constructed, providing a reference benchmark for subsequent emission source type identification and spatial overlay analysis. Through the assimilation processing of ground meteorological station measured data and mesoscale meteorological forecast models, gridded meteorological field data covering different vertical heights is generated, providing key meteorological parameters for high-precision carbon emission flux inversion. A fixed-wing mother drone performs a full-area baseline scan to identify hotspot areas, and then a multi-rotor drone performs intensive scanning of these hotspot areas and atmospheric tracing to pinpoint the precise spatial location of emission sources. The collaborative operation of these two heterogeneous UAVs addresses both wide-area coverage and detailed local investigation needs, providing a high-quality, multi-dimensional, and high-resolution data foundation for the complete chain from regional total emission accounting to precise emission source tracing.

[0075] In one embodiment of this application, the concentration sampling points have corresponding spatial location data within the monitoring area, and the step of determining the feature vector of each grid cell based on the multi-source prior data and the carbon concentration data includes: The carbon concentration data of each concentration sampling point is bound to the spatial location data to generate a concentration sampling point dataset with spatial coordinates and timestamps; The multi-source prior data is spatially registered and temporally synchronized with the concentration sampling point dataset to obtain multi-source feature data. From the multi-source feature data, extract the spatial geographic features, meteorological features, source strength prior features and concentration features corresponding to each concentration sampling point to obtain the feature set of each concentration sampling point; Based on the correspondence between the concentration sampling points and the grid cells, the feature set of each concentration sampling point is mapped to the corresponding grid cell to obtain the feature vector corresponding to each grid cell.

[0076] In this embodiment of the application, each concentration sampling point has associated spatial location data, which is used to clarify the specific spatial location of the sampling point within the monitoring area. The spatial location data comes from the real-time measurement results of the real-time dynamic differential positioning module carried by the UAV, and includes information in four dimensions: longitude, latitude, elevation and timestamp.

[0077] To correlate UAV-measured carbon concentration data with spatial location, and to give each carbon concentration data point clear geographic coordinates and time information, it is necessary to bind the carbon concentration data of each concentration sampling point with spatial location data. Specifically, using the CGCS2000 National Geodetic Coordinate System as a unified benchmark, the carbon concentration data is bound with the spatial location data of the corresponding sampling point, generating a concentration sampling point dataset with spatial coordinates and timestamps.

[0078] Subsequently, to unify multi-source prior data from different sources and with varying spatiotemporal references under the same spatial grid and temporal reference, and to achieve spatial alignment and temporal synchronization of the multi-source data, it is necessary to spatially register and temporally synchronize the multi-source prior data with the concentration sampling point dataset. Specifically, standardized geographic data, source strength prior feature data from the initial carbon emission source inventory, and gridded meteorological data are all resampled to a spatial resolution and temporal reference that matches the concentration sampling point dataset, achieving spatial registration and temporal synchronization of all data. At this point, the resampled multi-source prior data and the concentration sampling point dataset correspond spatially grid-by-grid and temporally time-by-time, forming multi-source feature data for feature extraction.

[0079] In one embodiment, for numerical data such as DEM, meteorological data, NDVI, and surface temperature, bilinear interpolation can be used to achieve resampling; for categorical data such as land cover type and emission source classification, nearest neighbor interpolation can be used to achieve resampling.

[0080] After completing spatial registration and time synchronization, for each concentration sampling point, multi-dimensional features corresponding to that sampling point are extracted from multi-source feature data to construct the input feature set of the carbon emission flux inversion model. The input feature set includes the feature set of each concentration sampling point, which includes spatial geographic features, meteorological features, source strength prior features, and concentration features.

[0081] Spatial geographic features include the longitude, latitude, elevation, slope, and aspect of the sampling point; meteorological features include wind speed, wind direction, ambient temperature, atmospheric pressure, and relative humidity at the corresponding time of sampling point; prior source strength features include the land cover type, building density, vegetation normalization index, surface temperature, distance to the nearest potential emission source, and activity level coefficient of the corresponding emission source in the grid cell where the sampling point is located; concentration features include the measured concentration values ​​of CO2, CH4, and CO at the sampling point, as well as the concentration ratios of CO / CO2 and CH4 / CO2.

[0082] It should be noted that if the grid cell containing the sampling point does not fall within the geographical boundary of any potential emission source, then there is no corresponding potential emission source for that grid cell, and its corresponding emission source activity level coefficient is assigned a value of 0, indicating that there is no prior emission source information at that location. In this case, except for the activity level coefficient being 0, the other features in the source strength prior features of that grid cell are still assigned normal values ​​based on the actual remote sensing interpretation and spatial analysis results.

[0083] In one embodiment, when calculating the concentration ratio feature, if the CO2 concentration value of a sampling point after calibration is less than 1.1 times the average background CO2 concentration of the monitoring area, the concentration ratio feature of that sampling point is not calculated, and it is directly marked as a background sampling point to avoid ratio anomalies caused by an excessively small denominator. The average background CO2 concentration of the monitoring area is calculated as follows: after preprocessing the CO2 concentration data obtained from the baseline scan of the main unit, hotspot area data with concentrations exceeding a preset threshold are removed, and the remaining data are the arithmetic mean. The average background CH4 concentration is calculated using the same method.

[0084] After obtaining the feature set of each concentration sampling point, since the monitoring area is divided into several uniformly sized grid cells, to map the discrete concentration sampling point features to a continuous spatial grid, it is necessary to, for each grid cell, based on the correspondence between concentration sampling points and grid cells (i.e., each concentration sampling point falls into a unique corresponding grid cell according to its spatial coordinates), synthesize the feature sets of all concentration sampling points falling into that grid cell. For example, averaging numerical features and taking the mode for categorical features, the feature vector corresponding to each grid cell is obtained. For grid cells that do not contain any concentration sampling points, spatial interpolation methods are used to fill them based on the feature values ​​of neighboring sampling points, ensuring that each grid cell has a complete feature vector. Spatial interpolation methods include, but are not limited to, inverse distance weighted interpolation or Kriging interpolation.

[0085] This application's embodiments spatially register and temporally synchronize multi-source prior data with UAV-measured carbon concentration data, constructing a multi-dimensional feature set for each concentration sampling point, encompassing four categories of features: spatial geography, meteorology, source strength priors, and concentration. Then, feature vectors for each grid cell are obtained through point-to-surface mapping, enabling discrete sampling point data to be represented within a unified spatial grid. This provides structured input data for subsequent carbon emission flux inversion models, ensuring data consistency from discrete sampling points to continuous flux field inversion. Simultaneously, the construction of CO / CO2 and CH4 / CO2 concentration ratio features within the feature set provides crucial evidence for the accurate identification of subsequent emission source types.

[0086] In one embodiment of this application, determining the carbon emission flux value of each grid cell based on the feature vector of each grid cell, and summing the carbon emission flux values ​​of all grid cells to obtain the total regional carbon emissions of the monitoring area, includes: The measured flux value of each concentration sampling point is calculated based on the gridded meteorological data and the carbon concentration data in the multi-source prior data. The pre-built carbon emission flux inversion model is trained and optimized based on the spatial location of each grid cell in the monitoring area, the feature set of each concentration sampling point, and the measured flux value of each concentration sampling point to obtain the target carbon emission flux inversion model. The carbon emission flux value of each grid cell is obtained based on the feature vector of each grid cell and the target carbon emission flux inversion model. The total carbon emission flux values ​​of all grid cells are summed to obtain the total regional carbon emissions of the monitored area.

[0087] In this embodiment, to obtain the label values ​​used to train the carbon emission flux inversion model, it is necessary to calculate the corresponding measured flux value for each concentration sampling point, i.e., the measured carbon concentration flux value for each concentration sampling point. Specifically, based on atmospheric boundary layer flux theory, the gradient method is used to calculate the measured flux value for that concentration sampling point.

[0088] The measured flux value at each concentration sampling point is calculated using the following formula: F=ρ×U×C corr ×K Where F is the measured flux value at each concentration sampling point, in mg / (m³). 2 ·s), where ρ is the density of the greenhouse gas under standard conditions, and CO2 is taken as 1.977 kg / m³. 3 CH4 was taken as 0.717 kg / m³ 3 U represents the measured average wind speed in the gridded meteorological data at the corresponding flight altitude and sampling time, in m / s. C corr The concentration value is corrected by the calibration factor f, and the unit is mg / m³. 3 K is the vertical diffusion coefficient, which can be calculated based on the Moning-Obukhov similarity theory. The specific calculation formula is as follows:

[0089] in, This is the von Kármán constant, which can take a value of 0.4; The frictional wind speed is calculated based on measured wind speed data from ground weather stations; z is the height of the sampling point above the ground. The momentum stability correction function is determined based on the Monin-Obukhov length L; the Monin-Obukhov length L is calculated from measured temperature and wind speed data.

[0090] Specifically, the momentum stability correction function The specific calculation method is as follows: exist At that time, it was a neutral stratification. ; exist At that time, it is an unstable stratification. ; exist At that time, it is a stable stratification. .

[0091] Frictional wind speed The calculation method is as follows:

[0092] Among them, U z Z represents the measured average wind speed at height z, and z0 represents the surface roughness length, determined by referring to a table based on the surface cover type. For example, 0.5~1.0m for urban built-up areas, 0.03~0.1m for farmland, and 0.5~2.0m for forest land. This is the integral correction function for momentum stability.

[0093] In this embodiment of the application, a carbon emission flux inversion model based on geographically weighted random forest is pre-constructed to invert the carbon emission flux value of each grid cell, thereby achieving high-precision flux inversion in the region.

[0094] Specifically, the measured flux value of each concentration sampling point needs to be used as the label value for training the carbon emission flux inversion model, and the feature set of each concentration sampling point needs to be used as the input feature set for training the carbon emission flux inversion model. The input feature set and label value corresponding to all concentration sampling points are then used to construct a sample dataset.

[0095] Before training the carbon emission flux inversion model, the input feature set needs to be preprocessed. Specifically, the numerical features are normalized by min-max to map the values ​​to the [0, 1] interval, and the categorical features are encoded using a unique encoding method.

[0096] To ensure the effectiveness and generalization ability of model training, the sample dataset is randomly divided into training, validation, and test sets. The training set is used for learning model parameters, the validation set for hyperparameter optimization, and the test set for independent evaluation of the final model accuracy. Hyperparameter optimization can employ a grid search method, optimizing parameters including the number of decision trees, maximum tree depth, minimum number of split samples per node, and minimum number of leaf node samples. The number of decision trees is 100-300, and the maximum tree depth is 10-20 layers. During training, a fixed random seed is used, and out-of-bag (OBC) validation is employed to suppress overfitting, with OBC data accounting for 30%. For accuracy validation of the trained model, 10-fold cross-validation is used to ensure the model's generalization ability and robustness. When the model's coefficient of determination R0 is... 2 The model is considered to be successfully trained when the mean absolute percentage error (MAPE) is ≥0.85 and ≤15%.

[0097] In one embodiment, the ratio of training set, validation set, and test set data volume can be set to 7:2:1. The sample dataset is partitioned using a spatially hierarchical random partitioning method, with grid cells of the monitoring area as the unit of division. This ensures that sampling points within the same grid cell appear in only one of the training, validation, and test sets, avoiding artificially high model accuracy due to spatial autocorrelation. Based on this, the carbon emission flux inversion model takes the feature set of concentration sampling points as input and the measured flux value of the concentration sampling points as output. The model is trained using the training set, its hyperparameters are optimized using the validation set, and the accuracy of the trained model is verified using the test set. Once the model accuracy meets the preset requirements, the target carbon emission flux inversion model is obtained.

[0098] The embodiments of this application systematically optimize the model hyperparameters using a grid search method, suppress overfitting by combining out-of-bag data validation, and ensure the model's generalization ability in different spatial regions through 10-fold spatial cross-validation, thereby further improving the robustness and reliability of the model.

[0099] The feature vector of each grid cell is used as input to the target carbon emission flux inversion model with the required accuracy. At this time, the target carbon emission flux inversion model will output the carbon emission flux value of each grid cell.

[0100] After obtaining the carbon emission flux value of each grid cell, a continuous carbon emission flux field grid map of the monitoring area can be generated based on the carbon emission flux value of each grid cell. This grid map is the spatial continuous distribution field of carbon emission flux in the monitoring area.

[0101] Based on this, the flux field grid map is divided into zones for statistical analysis, and the total emissions of CO2 and CH4 are calculated separately. The CH4 emissions are then converted into CO2 equivalents according to the global warming potential over a 100-year timescale, and the total carbon emissions of the monitored area are finally obtained.

[0102] In one embodiment, to ensure the physical consistency and statistical reliability of the flux field grid map, the grid flux values ​​need to be validated and corrected for reasonableness after the carbon emission flux field grid map is generated. Specifically, if the carbon emission flux value is negative, it is determined to be an outlier and replaced with the average flux of the eight neighboring grids. If the carbon emission flux value exceeds five times the average flux of the entire region, it is determined to be an outlier using the 3σ criterion and replaced with the upper limit of flux prediction corresponding to the prior source strength characteristics of the corresponding grid. The upper limit of flux prediction is the 95th percentile of the flux prediction values ​​of all grids within the monitoring area that have the same land cover type and emission intensity level as the grid. The 100-year global warming potential of CH4 is set at 29.8, based on the recommended value in the IPCC Sixth Assessment Report.

[0103] This application's embodiments calculate the measured flux value at each concentration sampling point using the gradient method based on atmospheric boundary layer flux theory. Combined with gridded meteorological data and calibrated carbon concentration data from multi-source prior data, it provides highly reliable training label values ​​for the carbon emission flux inversion model. Based on this, a geographically weighted random forest model is trained. Leveraging this model's strong fitting ability for nonlinear relationships and its adaptive weighting ability to spatial heterogeneity, a high-precision mapping from grid cell feature vectors to carbon emission flux values ​​is achieved, effectively solving the problem that traditional spatial interpolation methods struggle to characterize the spatial nonlinear distribution of carbon emission flux. Simultaneously, by verifying and correcting the prediction results for reasonableness, the physical consistency and statistical reliability of the flux field grid map are ensured. Finally, through zonal statistics and the conversion of CH4 to CO2 equivalents, a high-precision calculation of the total carbon emissions in the monitoring area is achieved.

[0104] In one embodiment of this application, the step of training and optimizing a pre-constructed carbon emission flux inversion model based on the spatial location of each grid cell in the monitoring area, the feature set of each concentration sampling point, and the measured flux value of each concentration sampling point to obtain a target carbon emission flux inversion model includes: The feature weight of each grid cell in the target carbon emission flux inversion model is determined based on the spatial location of each grid cell in the monitoring area. The pre-constructed carbon emission flux inversion model is trained and optimized based on the feature weights of each grid cell, the feature set of each concentration sampling point, and the measured flux value of each concentration sampling point to obtain the target carbon emission flux inversion model.

[0105] In this embodiment of the application, in order to characterize the continuity and heterogeneity of carbon emission flux as it changes with spatial location, a geographically weighted kernel function can be introduced on the basis of the traditional random forest algorithm to assign differentiated weights to samples at different spatial locations. The weight value is negatively correlated with the spatial distance between samples, that is, the closer the spatial distance between samples, the greater the influence of the target sample, and the farther the distance between samples, the smaller the influence of the target sample, thereby solving the problem of spatial heterogeneity of carbon emission flux.

[0106] In one embodiment, the geographic weighting kernel function uses a Gaussian kernel function, and the feature weight of each raster cell is specifically calculated using the following formula:

[0107] Among them, w ij Let d be the feature weight of the i-th grid cell, representing the influence of the j-th concentration sampling point on flux prediction of the i-th grid cell. ij Let be the Euclidean spatial distance between grid cell i and concentration sampling point j, calculated based on the spatial positions of grid cell i and concentration sampling point j, and let b be the kernel function bandwidth. The optimal value of the kernel function bandwidth can be determined using leave-one-out cross-validation. In this case, the bandwidth search range can be set to 100m-2000m, and the search step size can be set to 50m; the optimal bandwidth value b is determined by minimizing the mean square error of cross-validation.

[0108] After determining the feature weights corresponding to each grid cell, these feature weights are used as weight coefficients for each training sample (i.e., concentration sampling point). Combining the feature set and measured flux value of each concentration sampling point, the pre-constructed carbon emission flux inversion model is trained and optimized. Specifically, within a geographically weighted random forest framework, the feature weights are used to weight the sampling of each training sample, causing the model to focus more on samples closer to the target grid during training, thus obtaining the target carbon emission flux inversion model. The feature vector of this grid cell is then input into the trained and optimized target carbon emission flux inversion model to output the predicted carbon emission flux value for that grid cell. By performing the above training, optimization, and prediction process on each grid cell within the monitoring area, the carbon emission flux value for each grid cell can be obtained.

[0109] In existing technologies, flux inversion of UAV aerial survey data often adopts traditional spatial interpolation methods, which only perform geometric interpolation based on discrete sampling point concentration data. This does not fully integrate key influencing factors such as topography, real-time meteorological conditions, and intensity of human activities on the surface. It cannot characterize the complex nonlinear relationship between carbon emission concentration distribution and environmental factors. The inversion accuracy drops significantly under complex terrain and variable weather conditions, the model's generalization ability and robustness are insufficient, and it lacks clear physical meaning.

[0110] This application's embodiment introduces a geographically weighted kernel function, enabling the target carbon emission flux inversion model to adaptively adjust weights based on the spatial distance between samples. This fully considers the continuity and heterogeneity of carbon emission flux as it changes with spatial location. Compared to traditional spatial interpolation methods that only perform geometric interpolation based on discrete sampling points and do not integrate key influencing factors, the model constructed in this application's embodiment uses a multi-dimensional feature vector that integrates spatial geographic features, meteorological features, and source strength prior features as input. It learns and characterizes the complex nonlinear relationship between carbon emission flux and multiple influencing factors through a random forest algorithm, maintaining high inversion accuracy even under complex terrain and variable weather conditions.

[0111] Compared to traditional random forest models that treat all samples equally and ignore spatial relationships, the embodiments in this application employ a geographical weighting mechanism to enable the model to have spatial adaptability, effectively addressing the spatial heterogeneity problem of carbon emission fluxes and significantly improving the model's generalization ability and robustness. Furthermore, since the model input features include environmental variables with clear physical meaning such as wind speed, temperature, topography, and land cover, the inversion results have clear physical interpretability, avoiding the limitations of purely geometric interpolation methods that lack physical meaning.

[0112] In one embodiment of this application, the step of performing spatial overlay analysis on the monitoring area based on the multi-source prior data, the feature vector of each grid cell, and the carbon emission flux value of each grid cell to determine the carbon emission source type and spatial location of each carbon emission source within the monitoring area includes: The concentration ratio feature of each grid cell is calculated based on the concentration features in the feature vector of each grid cell; The concentration ratio feature of each grid cell is matched with a preset emission source feature library to determine the emission source type identification result of each grid cell. Based on the carbon emission flux value of each grid cell, target grid areas with carbon emission flux values ​​higher than a preset threshold are identified. The spatial overlay analysis of the emission source type identification results of each grid cell, the target grid area, and the initial carbon emission source list in the multi-source prior data is performed to determine the spatial location of each carbon emission source. The set of grid cells covered by each carbon emission source is determined based on the spatial location of each carbon emission source, and the carbon emission source type of each carbon emission source is determined based on the emission source type identification results of all grid cells in the set of grid cells.

[0113] In this embodiment, due to differences in combustion efficiency, fuel composition, and emission processes, different types of emission sources exhibit varying relative proportions of CO2, CH4, and CO. Therefore, to distinguish the emission source types at different locations within the monitoring area, it is necessary to calculate the concentration ratio characteristics based on the concentration characteristics in the feature vector of each grid cell. These concentration ratio characteristics include two ratios: CO / CO2 and CH4 / CO2, which can serve as the basis for identifying the emission source type.

[0114] After calculating the concentration ratio characteristics of each grid cell, the concentration ratio characteristics of each grid cell are matched with a preset emission source feature library to determine the emission source type identification result for each grid cell. The emission source feature library contains the threshold ranges for CO / CO2 and CH4 / CO2 concentration ratio characteristics of different types of emission sources.

[0115] As an example, for coal combustion sources, the CO / CO2 ratio ranges from 0.02 to 0.1, and the CH4 / CO2 ratio ranges from 0.001 to 0.005; for gas leak sources, the CO / CO2 ratio is ≤0.005, and the CH4 / CO2 ratio is ≥0.1; for transportation sources, the CO / CO2 ratio ranges from 0.05 to 0.2, and the CH4 / CO2 ratio ranges from 0.002 to 0.01; for agricultural / biological emission sources, the CO / CO2 ratio is ≤0.01, and the CH4 / CO2 ratio ranges from 0.01 to 0.08.

[0116] The CO / CO2 and CH4 / CO2 concentration ratios of each grid cell are matched against a threshold database of emission source characteristics. When a grid cell's concentration ratio falls within the threshold range of a certain type of emission source, the emission source type corresponding to that grid cell is determined to be of that category. If a grid cell's concentration ratio falls within the threshold range of two or more emission source types simultaneously, the emission source type with the higher carbon emission flux value for that grid cell is used as the final determination result, based on the fact that the emission source with the higher flux value contributes more to the emissions of that grid cell. If the CO2 and CH4 concentration values ​​of a grid cell do not exceed 1.1 times the average background concentration of the monitoring area, the grid cell is determined to be a background area and is not included in the emission source type determination range to avoid invalid type determination for low-concentration background areas.

[0117] In one embodiment, the ratio characteristic threshold range of the emission source feature library can be calibrated based on local measured data of the monitoring area to adapt to the differences in emission source characteristics in different regions.

[0118] Specifically, field concentration samples of known emission sources within the monitoring area are collected, and the 5th and 95th percentiles of the CO / CO2 and CH4 / CO2 concentration ratios for the corresponding emission source types are calculated. These percentile ranges are then used to replace the default threshold ranges for the corresponding emission source types in the feature library.

[0119] The embodiments of this application, through local calibration, enable the feature library thresholds to more accurately reflect the characteristic distribution of actual emission sources in the monitoring area, thereby improving the accuracy of type identification.

[0120] In addition, it is necessary to identify target grid areas where the carbon emission flux value is higher than a preset threshold based on the carbon emission flux value of each grid cell. The preset threshold can be determined based on the multiple relationship of the average flux value of the entire monitoring area. For example, it can be set to 3 times the average flux value of the entire area to identify emission hotspot grids where the carbon emission flux is significantly higher than the background level. The target grid area represents the spatial location of concentrated carbon emissions and is a key area for further determining the spatial location of emission sources.

[0121] Subsequently, spatial overlay analysis was performed on the emission source type identification results of each raster unit, the target raster areas with carbon emission flux values ​​exceeding a preset threshold, and the initial carbon emission source inventory from the multi-source prior data. Spatial overlay analysis refers to overlaying the above three types of spatial data into layers within the same geographic coordinate system and making a comprehensive judgment based on their spatial relationships.

[0122] Specifically, the target grid area serves as a clue to the potential location of emission sources. The spatial vector boundary of potential emission sources in the initial carbon emission source inventory is used as a spatial reference benchmark. The emission source type identification result of the grid unit is used as the basis for type determination. Through comparison and cross-validation of the spatial locations of these three factors, the spatial location of each carbon emission source is finally determined. It should be noted that the spatial location of each carbon emission source refers to the specific spatial range of the carbon emission source within the monitoring area, which is defined by geographical boundaries.

[0123] After determining the spatial location of each carbon emission source, a spatial distribution map of emission sources can be generated based on the spatial locations of all carbon emission sources. This map determines the set of grid cells covered by each carbon emission source, that is, all grid cells contained within the geographical boundary of the carbon emission source. The carbon emission source type of each carbon emission source is then determined based on the emission source type identification results of all grid cells in this set. As an example, statistical analysis can be performed on the emission source type identification results of all grid cells covered by the carbon emission source, and the type with the highest frequency can be determined as the final carbon emission source type.

[0124] This application's embodiments effectively pinpoint the spatial location of each carbon emission source by spatially overlaying the emission source type identification results at the grid cell scale with high-value carbon emission flux areas, combined with an initial carbon emission source inventory as a spatial reference. Simultaneously, through grid cell type statistics based on spatial location, the type identification results at the grid scale are aggregated to the emission source scale, achieving accurate transfer from grid type to emission source type, thus solving the technical problem of existing technologies being unable to spatially locate and identify emission sources. Furthermore, by matching the CO / CO2 and CH4 / CO2 ratios with thresholds in the emission source feature library, the differences in combustion efficiency and gas emission ratios among different emission source types are utilized to accurately distinguish emission source types, providing a reliable basis for subsequent emission source contribution calculations and liability determination.

[0125] In one embodiment of this application, determining the contribution of each carbon emission source to the total carbon emissions of the region based on the carbon emission flux value of each grid cell, the spatial location of each carbon emission source, and the total carbon emissions of the region includes: For each carbon emission source, the carbon emission flux values ​​of all grid cells in the grid cell set are summed to obtain the carbon emission amount of each carbon emission source. The contribution of each carbon emission source to the total carbon emissions of the region is determined based on the carbon emissions of each carbon emission source and the total carbon emissions of the region.

[0126] In this embodiment, after determining the spatial location of each carbon emission source and the set of grid cells it covers, in order to quantitatively assess the responsibility of each emission source in regional carbon emissions, it is necessary to calculate the contribution of each carbon emission source to the total regional carbon emissions. Specifically, for each carbon emission source, based on its spatial location, the set of grid cells it covers is determined, and the carbon emission flux values ​​of all grid cells in this set are summarized to obtain the carbon emission amount of that source. Subsequently, based on the carbon emission amount of each carbon emission source and the total regional carbon emissions of the monitored area, the contribution of each carbon emission source to the total regional carbon emissions can be obtained. The contribution is expressed as a percentage; a higher contribution indicates a greater impact of the emission source on regional carbon emissions.

[0127] Preferably, an emission flux allocation model can be used to calculate the contribution of each carbon emission source to the total carbon emissions in the monitoring area. The emission flux allocation model can use a spatial superposition weight allocation method, and the contribution of each carbon emission source is calculated using the following formula:

[0128] in, S represents the carbon emission contribution of the nth carbon emission source. nS is the set of all grid cells within the geographic boundary of the nth carbon emission source. all To monitor the collection of all grid cells within the area, F s Let A be the carbon emission flux value of grid cell s. s Let be the area of ​​the grid cell 's'. Specifically, This represents the carbon emissions from each carbon emission source. Represents the total carbon emissions of the region.

[0129] This application embodiment scientifically decomposes the total regional carbon emissions to each emission source using a spatial superposition weight allocation method. Based on the flux values ​​of grid cells within the geographical boundaries of each emission source, the carbon emissions of each emission source are obtained, thereby quantifying the contribution of each emission source to the total regional emissions and realizing a quantitative assessment of the emission responsibility of each emission source.

[0130] In one embodiment of this application, in order to achieve dynamic monitoring and early warning of regional carbon emissions, steps 101 to 105 above need to be repeated according to a preset monitoring cycle to obtain time-series carbon emission monitoring data of the monitoring area in different monitoring cycles. The monitoring cycle can be monthly or quarterly, and this embodiment of the application does not impose any limitation on this.

[0131] Based on multi-period monitoring data, a regional dynamic spatiotemporal map of carbon emissions can be constructed. This map visually displays the dynamic changes in carbon emission fluxes of each grid cell and carbon emission source within the monitoring area over time. Furthermore, the temporal variation patterns of emissions from each carbon emission source can be analyzed, identifying trends and seasonal characteristics of emission intensity fluctuations over time, thus obtaining the temporal variation trend. For a single carbon emission source, when the increase in emission flux compared to the previous monitoring period exceeds a preset warning threshold, an emission anomaly warning is triggered, generating warning information.

[0132] In one embodiment, the preset warning threshold is set to 30%, meaning that when the emission flux of a certain emission source increases by more than 30% compared to the previous period, the system automatically determines that the emission source has an emission anomaly. After triggering the warning, the system simultaneously generates an anomaly warning report, which includes the location, type, emission flux increase, corresponding monitoring period, and possible causes of the anomaly, such as preliminary inferences based on the emission source type and emission increase characteristics, that it is an equipment failure, process change, or leakage event.

[0133] After completing the above monitoring, inversion, source tracing and early warning analysis, the multiple results of the monitoring area will be integrated into the GIS (Geographic Information System) platform for dynamic visualization and display. A standardized regional carbon emission flux monitoring and source tracing report will be automatically generated. The displayed results include, but are not limited to, concentration field raster map, carbon emission flux field raster map, emission source spatial distribution map, contribution, time series change trend and early warning information.

[0134] This application's embodiments achieve a technological upgrade from single static monitoring to continuous dynamic monitoring through time-series data acquisition and dynamic spatiotemporal mapping within a preset monitoring cycle. This enables managers to grasp the temporal variation patterns of emissions from various emission sources, promptly detect emission anomalies, and trigger early warnings, supporting the normalized and institutionalized supervision of carbon emissions. Simultaneously, through dynamic visualization and automatic generation of standardized reports via a GIS platform, monitoring data is transformed into intuitive management information.

[0135] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of this application.

[0136] This application also provides a storage medium that stores computer instructions. When the computer executes the computer instructions, it is used to execute the various processes of the above-described carbon emission source tracing method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0137] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0138] This application also provides an electronic device, including a processor 3010, a memory 309, and a program or instructions stored in the memory 309 and executable on the processor 3010. When the program or instructions are executed by the processor 3010, they implement the various processes of the above-described carbon emission source tracing method embodiments and achieve the same technical effects. To avoid repetition, they will not be described again here.

[0139] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0140] Figure 3 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of this application. The electronic device 300 includes, but is not limited to, components such as: radio frequency unit 301, network module 302, audio output unit 303, input unit 304, sensor 305, display unit 306, user input unit 307, interface unit 308, memory 309, and processor 3010.

[0141] Those skilled in the art will understand that the electronic device 300 may also include a power supply (such as a battery) for supplying power to various components. The power supply may be logically connected to the processor 3010 through a power management system, thereby enabling functions such as managing charging, discharging, and power consumption through the power management system. Figure 3 The electronic device structure shown does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here. This application also provides a computer program product, including a computer program / instruction. When the computer program / instruction is executed by a processor, it implements the various processes of the above-described carbon emission source tracing method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0142] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0143] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, 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 is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0144] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for tracing carbon emission sources, characterized in that, include: Acquire multi-source prior data of the monitoring area, as well as carbon concentration data of multiple concentration sampling points within the monitoring area; the monitoring area includes multiple grid cells, and there is a correspondence between the concentration sampling points and the grid cells; Based on the multi-source prior data and the carbon concentration data, a feature vector for each grid cell is determined; the feature vector for each grid cell is used to characterize the carbon emission attributes of each grid cell. The carbon emission flux value of each grid cell is determined based on the feature vector of each grid cell, and the carbon emission flux values ​​of all grid cells are summed to obtain the total regional carbon emissions of the monitoring area. Based on the multi-source prior data, the feature vector of each grid cell, and the carbon emission flux value of each grid cell, a spatial overlay analysis is performed on the monitoring area to determine the carbon emission source type and spatial location of each carbon emission source within the monitoring area. The contribution of each carbon emission source to the total carbon emissions of the region is determined based on the carbon emission flux value of each grid cell, the spatial location of each carbon emission source, and the total carbon emissions of the region.

2. The tracing method according to claim 1, characterized in that, The acquisition of multi-source prior data of the monitoring area, and carbon concentration data of multiple concentration sampling points within the monitoring area, includes: Acquire geographic information data, satellite remote sensing data, meteorological data, and baseline concentration data for the monitored area; The satellite remote sensing data is segmented and classified to obtain the type and spatial geographic boundary of each potential emission source within the monitoring area; An initial carbon emission source inventory of the monitoring area is determined based on the type and spatial geographical boundaries of each potential emission source within the monitoring area; Based on the baseline concentration data, hotspot areas were identified in the monitoring area; The hotspot area is scanned to determine the hotspot concentration data within the hotspot area; The geographic information data, the initial carbon emission source inventory, and the meteorological data are identified as the multi-source prior data. Based on the baseline concentration data and the hotspot concentration data, the carbon concentration data of multiple concentration sampling points within the monitoring area are determined.

3. The tracing method according to claim 1, characterized in that, The concentration sampling points have corresponding spatial location data within the monitoring area. The step of determining the feature vector of each grid cell based on the multi-source prior data and the carbon concentration data includes: The carbon concentration data of each concentration sampling point is bound to the spatial location data to generate a concentration sampling point dataset with spatial coordinates and timestamps; The multi-source prior data is spatially registered and temporally synchronized with the concentration sampling point dataset to obtain multi-source feature data. From the multi-source feature data, extract the spatial geographic features, meteorological features, source strength prior features and concentration features corresponding to each concentration sampling point to obtain the feature set of each concentration sampling point; Based on the correspondence between the concentration sampling points and the grid cells, the feature set of each concentration sampling point is mapped to the corresponding grid cell to obtain the feature vector corresponding to each grid cell.

4. The tracing method according to claim 3, characterized in that, The process of determining the carbon emission flux value of each grid cell based on its feature vector, and summing the carbon emission flux values ​​of all grid cells to obtain the total regional carbon emissions of the monitoring area, includes: The measured flux value of each concentration sampling point is calculated based on the gridded meteorological data and the carbon concentration data in the multi-source prior data. The pre-built carbon emission flux inversion model is trained and optimized based on the spatial location of each grid cell in the monitoring area, the feature set of each concentration sampling point, and the measured flux value of each concentration sampling point to obtain the target carbon emission flux inversion model. The carbon emission flux value of each grid cell is obtained based on the feature vector of each grid cell and the target carbon emission flux inversion model. The total carbon emission flux values ​​of all grid cells are summed to obtain the total regional carbon emissions of the monitored area.

5. The tracing method according to claim 4, characterized in that, The process of training and optimizing a pre-built carbon emission flux inversion model based on the spatial location of each grid cell in the monitoring area, the feature set of each concentration sampling point, and the measured flux value of each concentration sampling point, to obtain a target carbon emission flux inversion model, includes: The feature weight of each grid cell in the target carbon emission flux inversion model is determined based on the spatial location of each grid cell in the monitoring area. The pre-constructed carbon emission flux inversion model is trained and optimized based on the feature weights of each grid cell, the feature set of each concentration sampling point, and the measured flux value of each concentration sampling point to obtain the target carbon emission flux inversion model.

6. The tracing method according to claim 1, characterized in that, The step of performing spatial overlay analysis on the monitoring area based on the multi-source prior data, the feature vector of each grid cell, and the carbon emission flux value of each grid cell to determine the carbon emission source type and spatial location of each carbon emission source within the monitoring area includes: The concentration ratio feature of each grid cell is calculated based on the concentration features in the feature vector of each grid cell; The concentration ratio feature of each grid cell is matched with a preset emission source feature library to determine the emission source type identification result of each grid cell. Based on the carbon emission flux value of each grid cell, target grid areas with carbon emission flux values ​​higher than a preset threshold are identified. The spatial overlay analysis of the emission source type identification results of each grid cell, the target grid area, and the initial carbon emission source list in the multi-source prior data is performed to determine the spatial location of each carbon emission source. The set of grid cells covered by each carbon emission source is determined based on the spatial location of each carbon emission source, and the carbon emission source type of each carbon emission source is determined based on the emission source type identification results of all grid cells in the set of grid cells.

7. The tracing method according to claim 6, characterized in that, The step of determining the contribution of each carbon emission source to the total carbon emissions of the region based on the carbon emission flux value of each grid cell, the spatial location of each carbon emission source, and the total carbon emissions of the region includes: For each carbon emission source, the carbon emission flux values ​​of all grid cells in the grid cell set are summed to obtain the carbon emission amount of each carbon emission source. The contribution of each carbon emission source to the total carbon emissions of the region is determined based on the carbon emissions of each carbon emission source and the total carbon emissions of the region.

8. A storage medium, characterized in that, The storage medium stores computer instructions, which, when executed by the computer, are used to perform a carbon emission source tracing method as described in any one of claims 1-7.

9. An electronic device, characterized in that, Includes at least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform a carbon emission source tracing method as described in any one of claims 1-7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements a carbon emission source tracing method as described in any one of claims 1-7.

Images