An emission reduction system based on carbon footprint monitoring

Abstract

The present application relates to the technical field of carbon emission monitoring and tracing, in particular to a kind of emission reduction system based on carbon footprint monitoring, comprising: multi-modal perception module, through multi-source heterogeneous perception node parallel acquisition energy flow and environmental gas information;Data fusion processing module, the perception data is executed space-time calibration and information fusion, generates the comprehensive perception atlas of unified space-time reference;Dynamic task scheduling module, according to atlas derivation and generation instruction, guide mobile inspection equipment to accurately approach suspected high emission target;Source identification evaluation module, through mobile equipment obtains target multidimensional high-resolution feature, completes the fine identification and quantitative evaluation of carbon emission source;Cloud collaborative processing module, upload the result to cloud center.The present application realizes the active discovery, accurate positioning and independent quantization of carbon emission source, improves the accuracy and tracing ability of carbon footprint monitoring.

Description

Technical Field

[0001] This invention relates to the field of carbon emission monitoring and source tracing technology, and in particular to an emission reduction system based on carbon footprint monitoring. Background Technology

[0002] Currently, carbon emission monitoring mainly relies on gas concentration sensors deployed at fixed locations or macroscopic remote sensing observations. These technologies can obtain overall concentration changes within a region or estimate total emissions over a large area. However, fixed-point monitoring has limited spatial coverage, making it difficult to capture the specific spatial location and dynamic changes of emission sources; while remote sensing technology has a wide coverage, its spatial resolution is insufficient to identify specific facilities, and it is easily affected by meteorological conditions such as clouds, failing to provide continuous and detailed monitoring data.

[0003] The aforementioned methods suffer from a disconnect between monitoring data and specific emission sources. The system can only reflect the overall level or anomalies of gas concentration within a region, and cannot actively identify and locate specific, especially mobile or intermittent, high-intensity emission points within the region. This results in monitoring results remaining at a macroscopic level, failing to achieve independent identification and accurate quantification of individual emission sources, making it difficult to accurately trace carbon emission responsibility and formulate targeted management measures.

[0004] This invention aims to address the problems of existing technologies being unable to proactively detect suspected emission sources and unable to perform close-range, refined identification and independent quantitative assessment of detected targets. A technical solution is needed that can automatically identify potential high-emission targets from wide-area monitoring, guide mobile devices for precise verification, and ultimately achieve accurate source tracing and quantitative analysis of specific emission sources. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an emission reduction system based on carbon footprint monitoring.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an emission reduction system based on carbon footprint monitoring, comprising: The multimodal sensing module, through multi-source heterogeneous sensing nodes deployed in the monitoring area, collects dynamic information on energy flow and environmental gas concentration information in parallel to form a multimodal sensing dataset; The data fusion processing module performs spatiotemporal calibration and information fusion processing on the multimodal sensing dataset to generate a comprehensive sensing map with a unified spatiotemporal reference. The dynamic task scheduling module derives and generates dynamic task scheduling instructions for the mobile inspection equipment based on the comprehensive perception map, guiding the mobile inspection equipment to accurately approach suspected high-emission targets. Source identification and assessment module: The mobile inspection device executes the dynamic task scheduling command to obtain multi-dimensional high-resolution feature information of the target, and completes the refined identification and quantitative assessment of carbon emission sources based on this. The cloud-based collaborative processing module uploads the refined identification and quantitative evaluation results to the cloud-based collaborative processing center via an encrypted data link.

[0007] Preferably, the multi-source heterogeneous sensing nodes deployed in the monitoring area collect dynamic information on energy flow and ambient gas concentration information in parallel to form a multimodal sensing dataset, specifically performing the following operations: The fixed gas concentration sensor, the embedded power metering unit, and the heat flow monitoring probe are activated simultaneously to form the multi-source heterogeneous sensing node network. The fixed gas concentration sensor periodically captures the greenhouse gas concentration readings at the monitoring points. The embedded power metering unit records the instantaneous power and cumulative power consumption of the power supply circuit. The heat flow monitoring probe measures the heat radiation flux on the surface of the pipe or equipment. The greenhouse gas concentration readings, instantaneous power and cumulative power consumption, and thermal radiation flux are locally timestamped and bound to spatial coordinates to form the original time-series sensing data stream; A multi-channel data buffer queue is established to cache and pre-sort the time-series sensing data stream, while compensating and aligning the clock deviations between different sensing nodes. Based on the preset data quality verification rules, outlier detection and missing value interpolation are performed on the time-series sensing data stream after clock compensation and alignment, invalid sampling points are removed, and standardized sensing data blocks that have undergone quality control processing are generated. All standardized sensing data blocks belonging to the same geographic grid and the same time window are aggregated and encapsulated to output a structured multimodal sensing dataset.

[0008] Preferably, the step of performing spatiotemporal calibration and information fusion processing on the multimodal sensing dataset to generate a comprehensive sensing map with a unified spatiotemporal reference specifically involves the following operations: Receive the multimodal sensing dataset and parse the greenhouse gas concentration data layer, energy consumption data layer, and thermal radiation data layer contained therein; A spatial interpolation surface model is established for the greenhouse gas concentration data layer to transform the concentration readings at discrete points into a continuous concentration distribution field covering the monitoring area; The energy consumption data layer and the thermal radiation data layer are associated and mapped to identify the spatial coupling relationship between energy consumption and thermal radiation, and to construct an energy thermal effect coupling map. The continuous concentration distribution field and the energy thermal effect coupled map are superimposed on the same geographic information base map, and pixel-level information fusion operation is performed to calculate the correlation weight between the concentration field intensity and the energy thermal effect in each geographic unit. Based on the aforementioned correlation weights, the fused data is normalized and standardized to generate a rasterized comprehensive perception map that comprehensively reflects the potential intensity and spatial distribution characteristics of carbon emissions.

[0009] Preferably, a spatial interpolation surface model is established for the greenhouse gas concentration data layer, and the concentration readings at discrete points are converted into a continuous concentration distribution field covering the monitoring area. Specifically, the following operations are performed: The greenhouse gas concentration data layer in the multimodal sensing dataset is analyzed to extract the spatial coordinates of all monitoring points and their corresponding concentration readings. Using the Kriging spatial interpolation algorithm, based on the spatial coordinates and concentration readings of the monitoring points, the predicted concentration value at any unsampled location within the monitoring area is calculated; Based on the digital elevation model data of the monitoring area, the concentration prediction value is corrected for topographic elevation to eliminate the influence of topographic undulation on gas diffusion; The concentration prediction values ​​after terrain elevation correction are gridded to generate a raster data layer with fixed spatial resolution. The raster data layer is smoothed and filtered to eliminate local abnormal fluctuations that may be caused by interpolation calculations, and finally outputs a continuous concentration distribution field covering the entire monitoring area.

[0010] Preferably, the step of deriving and generating dynamic task scheduling instructions for the mobile inspection equipment based on the comprehensive sensing map specifically involves the following operations: Load the comprehensive perception map, cluster and partition the standardized scores in the map to divide it into high-interest areas, medium-interest areas and low-interest areas; Based on the spatial boundary and shape characteristics of the high-interest area, an initial set of candidate waypoints is automatically generated, and an initial priority weight is assigned to each candidate waypoint. Introduce the current status information of the mobile inspection device, including remaining battery power, current location, and sensor operating status, and construct a device state constraint model; Based on the equipment state constraint model, the initial priority weights of the candidate inspection waypoints are dynamically adjusted, and multi-objective path planning is performed while considering the transfer distance and time cost between waypoints. Output the waypoint access sequence and path planning results that satisfy the equipment state constraints and have the best overall cost, and compile them into dynamic task scheduling instructions that can be parsed by mobile inspection equipment.

[0011] Preferably, the mobile inspection device executes the dynamic task scheduling instruction to obtain multi-dimensional high-resolution feature information of the target, specifically performing the following operations: The mobile inspection equipment parses and executes the received dynamic task scheduling instructions, and autonomously navigates to the waypoint location specified in the instructions; Upon arrival at the target waypoint, the hyperspectral imager, lidar scanner, and miniature gas chromatograph are activated simultaneously to conduct a coordinated scan and sampling of the target emission source or its surrounding environment. The hyperspectral imager collects reflectance data of the target in hundreds of narrow spectral bands, forming a hyperspectral data cube; The lidar scanner emits laser pulses and receives echoes to generate point cloud data describing the three-dimensional structure of the target surface. The miniature gas chromatograph extracts on-site air samples and performs rapid separation and detection, outputting precise concentration spectra of gas components; The hyperspectral data cube, the three-dimensional structural point cloud data, and the precise concentration spectrum are synchronized in time and registered spatially, and packaged into multi-dimensional high-resolution feature information of the target.

[0012] Preferably, the detailed identification and quantitative assessment of carbon emission sources based on this involves the following operations: A multimodal feature fusion identification model is established, wherein the input of the model is the multi-dimensional high-resolution feature information of the target; The spectral fingerprints of substances are extracted from the hyperspectral data cube and matched with a preset typical emission substance spectral library to preliminarily identify the types of emission substances. By combining the three-dimensional structural point cloud data, the external geometric shape, physical size and spatial orientation of the emission source and the emission outlet are analyzed to help determine the equipment type and operating status of the emission source. Using the precise concentration spectrum, the types of emitted substances initially identified are verified, and the precise volume or mass concentration of various greenhouse gases is calculated. By taking into account the types of emitted substances, the types and operating status of the equipment, and the precise volume concentration or mass concentration, a refined quantitative assessment of the carbon emission rate and total amount of a specific emission source is completed through preset emission factor mapping rules and mass balance calculations.

[0013] Preferably, after deriving and generating dynamic task scheduling instructions for the mobile inspection equipment based on the comprehensive sensing map, the following adaptive optimization steps are further included: The system receives real-time on-site confirmation data transmitted back by the mobile inspection equipment during the execution of dynamic task scheduling instructions. The on-site confirmation data includes the verified actual location and emission intensity of the emission source. The on-site confirmation data is compared and analyzed with the comprehensive perception map on which the dynamic task scheduling instruction is generated, and the map prediction accuracy index is calculated. Based on the map prediction accuracy index, the scoring threshold used to divide high-interest areas is dynamically adjusted, and the clustering partitioning results are updated. Based on the updated clustering and partitioning results, a replanning process for the dynamic task scheduling instructions is triggered to generate optimized task scheduling instructions that adapt to the latest monitoring and understanding.

[0014] Preferably, after establishing a multimodal feature fusion identification model, the following model evolution steps are also included: Continuously collect all detailed identification and quantitative evaluation cases completed by mobile inspection equipment to form a historical case database; Periodically extract newly emerging emission source characteristic patterns or variation patterns of existing emission sources from the historical database of the aforementioned cases; The extracted new feature patterns or mutation patterns are used to incrementally train the multimodal feature fusion identification model, and update the identification rules and parameters inside the model. The updated multimodal feature fusion identification model was redeployed to the mobile inspection equipment, enabling it to identify newly emerging or mutated emission sources.

[0015] Preferably, after the refined identification and quantitative evaluation results are uploaded to the cloud collaborative processing center via an encrypted data link, the following data aggregation and distribution steps are also included: The cloud-based collaborative processing center receives refined identification and quantitative assessment results uploaded from multiple mobile inspection devices, and performs deduplication, correlation, and integration on the refined identification and quantitative assessment results to construct a regional-level panoramic view of carbon footprint. The regional carbon footprint panoramic view is subjected to trend analysis and pattern mining in multiple time dimensions to identify the spatiotemporal evolution of carbon emissions. Based on preset permissions and subscription rules, the regional carbon footprint panoramic view and its analysis results are distributed to different regulatory terminals or data application systems.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: Based on a comprehensive sensing map formed by fusing heterogeneous data from multiple sources, the system automatically derives and generates dynamic scheduling commands. Mobile inspection equipment proactively adjusts its inspection path and task priority according to these commands, enabling directional movement and approach to potential high-emission targets identified in the map. This operational mechanism transforms the monitoring mode from a fixed, passive data reception model to one driven by sensing data, actively tracking and verifying abnormal targets. The system can respond in real time to spatiotemporal changes in emission intensity, automatically covering blind spots or mobile sources that traditional fixed networks cannot detect, improving the efficiency and accuracy of detecting hidden or sudden emission events.

[0017] Upon approaching the target, the mobile inspection equipment simultaneously collects multi-dimensional information, including close-range high-precision gas concentration gradients, optical images, and thermal radiation characteristics. Through the fusion and analysis of this high-resolution on-site characteristic data, the system achieves refined identification of the physical properties and emission types of the emission source. Simultaneously, utilizing the high signal-to-noise ratio concentration data and characteristic parameters obtained from close-range measurements, the system can more accurately invert and quantitatively calculate the emission flux of this independent emission source. This shifts the assessment of carbon emissions from a macro-regional perspective to specific, identifiable individual facilities or emission outlets, enabling precise traceability of emission responsibility and providing a technological foundation for carbon accounting and management decisions based on accurate source-level data. Attached Figure Description

[0018] Figure 1 This is a flowchart of the emission reduction system based on carbon footprint monitoring described in this invention; Figure 2 A flowchart for generating a comprehensive sensing map; Figure 3 A flowchart generated for dynamic task scheduling instructions; Figure 4 A bar chart comparing methane and ethylene concentrations at different detection sites; Figure 5 A graph showing the deduplication effect and quality analysis of emission source data. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 Within the monitoring area, the multimodal sensing module collects dynamic energy flow information and ambient gas concentration information in parallel through deployed multi-source heterogeneous sensing nodes, forming a multimodal sensing dataset. The data fusion processing module receives this dataset and performs spatiotemporal calibration and information fusion processing to generate a comprehensive sensing map with a unified spatiotemporal benchmark. The dynamic task scheduling module derives dynamic task scheduling instructions for the mobile inspection equipment based on this comprehensive sensing map, guiding the mobile inspection equipment to accurately approach suspected high-emission targets. The source identification and assessment module controls the mobile inspection equipment to execute the dynamic task scheduling instructions, acquire multi-dimensional, high-resolution feature information of the targets, and based on this, complete the refined identification and quantitative assessment of carbon emission sources. The cloud-based collaborative processing module uploads the refined identification and quantitative assessment results to the cloud-based collaborative processing center via an encrypted data link.

[0022] In one embodiment of the present invention, see [reference] Figure 2 By deploying multi-source heterogeneous sensing nodes in the monitoring area to collect information in parallel, a multimodal sensing dataset is formed. In an exemplary industrial park monitoring scenario, the monitoring area is divided into multiple regular geographical grids, with a cluster of sensing nodes deployed in each grid. In some embodiments, fixed gas concentration sensors are installed on factory roofs and streetlight poles to periodically capture readings of greenhouse gas concentrations such as carbon dioxide and methane at the monitoring points; embedded power metering units are integrated into the power supply circuits of various major production equipment and workshops to continuously record the instantaneous power and cumulative power consumption of the power supply circuits; and heat flow monitoring probes are attached to the surfaces of steam pipes, reactor walls, and heating pipelines to measure the heat radiation flux on the surface of the pipes or equipment. It can be understood that all sensing nodes are synchronously activated and work in parallel after being powered on, forming a multi-source heterogeneous sensing node network covering the monitoring area.

[0023] In practice, greenhouse gas concentration readings, instantaneous power and cumulative power consumption, and thermal radiation flux are locally timestamped and bound to spatial coordinates. The microprocessor inside each sensing node adds a local timestamp accurate to milliseconds to each collected data record and binds the node's preset unique geographic coordinates to the data record, forming the original time-series sensing data stream. It is understandable that the time-series sensing data streams generated by different nodes may have slight time deviations and may also have delay differences along the transmission path.

[0024] In practical implementation, a multi-channel data buffer queue is established to cache and pre-sort the time-series sensing data stream. The data fusion processing module sets up an independent buffer for each data type, receiving the raw data stream from the sensing nodes. The multi-channel data buffer queue caches the data streams from fixed gas concentration sensors, embedded power metering units, and heat flow monitoring probes separately. The pre-sorting operation arranges data of the same type according to their chronological order based on the timestamp information within the data packets. Simultaneously, clock deviations between different sensing nodes are compensated and aligned. This compensation and alignment operation receives network time protocol timing signals, calculates the offset between each node's local clock and the standard time, and corrects the data timestamps.

[0025] In practice, outlier detection and missing value imputation are performed on the clock-compensated and aligned time-series sensing data stream according to preset data quality verification rules. These rules include range verification, mutation rate verification, and consistency verification. Outlier detection is based on statistical characteristics within a sliding time window. For example, for a concentration reading sequence generated by a fixed gas concentration sensor, the median and absolute median difference of the sequence are calculated within a specific time window. If a reading meets the criteria, it is considered an outlier and marked. Here, represents the i-th greenhouse gas concentration reading, represents the median of greenhouse gas concentration readings within the time window, represents the absolute median difference of greenhouse gas concentration readings within the time window, and is a preset proportionality coefficient. Marked outliers and missing values ​​due to communication interruptions are filled using linear interpolation based on adjacent valid sampling points. After outlier detection and missing value imputation, invalid sampling points are removed, generating a standardized sensing data block with regular length that has undergone quality control processing.

[0026] In practice, all standardized sensing data blocks belonging to the same geographic grid and within the same time window are aggregated and encapsulated. The data fusion processing module sets a unified time window length. Within an aggregation cycle, the system collects all standardized sensing data blocks whose geographic coordinates fall within the same geographic grid boundary and whose timestamps belong to the same time window. The aggregated data is encapsulated according to a preset structured format, which includes the grid number, the time window start timestamp, the set of greenhouse gas concentration readings provided by all fixed gas concentration sensors within the grid, the set of instantaneous power and cumulative power consumption recorded by all embedded power metering units, and the set of thermal radiation flux measured by all thermal energy flow monitoring probes. Finally, a structured multimodal sensing dataset is output, which serves as the input for subsequent processing.

[0027] In one embodiment of the present invention, see [reference] Figure 3 The system performs spatiotemporal calibration and information fusion processing on the multimodal sensing dataset to generate a comprehensive sensing map. In a monitoring example for a chemical plant area, the data fusion processing module receives a structured multimodal sensing dataset from the multimodal sensing module. The data fusion processing module parses the multimodal sensing dataset, separating it into a greenhouse gas concentration data layer, an energy consumption data layer, and a thermal radiation data layer. The greenhouse gas concentration data layer contains carbon dioxide concentration readings reported by discretely distributed fixed gas concentration sensors within the plant area; the energy consumption data layer contains instantaneous power data of equipment recorded by embedded power metering units in each workshop; and the thermal radiation data layer contains radiative flux values ​​measured by heat flow monitoring probes installed on the surfaces of reaction devices and pipelines.

[0028] In practical implementation, a spatial interpolation surface model is established for the greenhouse gas concentration data layer, transforming the concentration readings of discrete points into a continuous concentration distribution field covering the monitoring area. The data fusion processing module parses the greenhouse gas concentration data layer in the multimodal sensing dataset, extracting the spatial coordinates of all fixed gas concentration sensor nodes and their corresponding concentration readings. In some embodiments, the Kriging spatial interpolation algorithm is used to calculate the predicted concentration value at any unsampled location within the monitoring area based on the spatial coordinates and concentration readings of the monitoring points. The Kriging algorithm constructs a variogram model by analyzing the spatial autocorrelation between monitoring point readings and uses this model to assign weights to the known monitoring point readings around each prediction point. For a location to be predicted within the monitoring area... Its greenhouse gas concentration estimates By weighting the readings of nearby known monitoring points The expression is:

[0029] in: It is used to estimate the number of nearby monitoring points. It is a weighting factor assigned to the reading of each known monitoring point. By solving the Kriging equations, it was determined that the equations guarantee the unbiasedness and minimum variance of the estimate.

[0030] In practice, the concentration prediction values ​​are corrected for terrain elevation based on the digital elevation model data of the monitoring area to eliminate the impact of terrain undulations on gas diffusion. It is understandable that in a chemical plant setting, the height differences between buildings and storage tanks can affect gas accumulation and diffusion. The data fusion processing module overlays the calculated initial concentration prediction grid with the plant's digital elevation model data, introducing a correction factor based on its elevation for each grid cell. The concentration prediction values ​​for grid cells located in low-lying areas are appropriately adjusted upwards according to the diffusion model, while those for grid cells located at higher elevations are adjusted downwards accordingly. The terrain-corrected concentration prediction values ​​are then gridded to generate a regular raster data layer with a fixed spatial resolution. The value of each pixel in the raster data layer represents the corrected concentration prediction value for that geographic unit. The raster data layer is then smoothed using a mean filter or Gaussian filter convolution kernel to eliminate local anomalies or "bull's-eye" effects that may arise from interpolation calculations. The final output is a spatially continuous, numerically smooth, continuous concentration distribution field covering the entire plant monitoring area.

[0031] In practical implementation, the energy consumption data layer and the thermal radiation data layer are correlated and mapped to identify the spatial coupling relationship between energy consumption and thermal radiation, constructing an energy-thermal effect coupling map. The data fusion processing module correlates the geographical location of each embedded power metering unit with its recorded instantaneous power value, and the geographical location of each thermal energy flow monitoring probe with its measured thermal radiation flux value. Optionally, through spatial connection operations, energy consumption data and thermal radiation data in similar locations are paired to calculate the "energy consumption-thermal radiation" ratio or correlation coefficient for each local area. The constructed energy-thermal effect coupling map is a raster layer, where each raster cell contains a coupling strength index, which reflects the efficiency or degree of correlation between electrical energy consumption and thermal radiation within the cell. Areas with high coupling strength indices may correspond to production equipment with high energy consumption and significant heat dissipation.

[0032] In practice, the continuous concentration distribution field and the coupled energy thermal effect map are overlaid on the same geographic information base map, and pixel-level information fusion operations are performed to calculate the correlation weight between the concentration field intensity and the energy thermal effect within each geographic unit. The data fusion processing module ensures that the continuous concentration distribution field and the coupled energy thermal effect map have the same spatial reference and raster resolution. In some embodiments, for each aligned raster unit, the information fusion operation uses a weighted overlay method, and the correlation weights are calculated accordingly. Determined by a function based on the numerical values ​​of two data layers, for example ,in: This represents the normalized concentration value of the grid cell in a continuous concentration distribution field. This represents the normalized coupling strength exponent of the grid cell in the energy-thermal effect coupling spectrum. (Function) The design aims to highlight areas with both high carbon emission concentrations and high energy-thermal effect coupling indices. Based on correlation weights, the fused data is normalized and standardized, mapping the fused result value of each grid cell to a preset standardized score range. This generates a gridded integrated sensing map that comprehensively reflects the potential intensity and spatial distribution characteristics of carbon emissions. The standardized score of each grid cell in the integrated sensing map indicates the priority level that the mobile inspection equipment needs to focus on in that area.

[0033] In one embodiment of the present invention, dynamic task scheduling instructions for mobile inspection equipment are derived and generated based on the comprehensive sensing map, including subsequent adaptive optimization. In a monitoring example of a large industrial park, the dynamic task scheduling module loads the rasterized comprehensive sensing map generated by the data fusion processing module. The dynamic task scheduling module clusters and partitions the standardized scores in the comprehensive sensing map, dividing it into high-interest areas, medium-interest areas, and low-interest areas, and aggregating raster cells with similar standardized scores into the same category. In some embodiments, areas with standardized scores between 85 and 100 are classified as high-interest areas, areas between 60 and 84 are classified as medium-interest areas, and areas below 60 are classified as low-interest areas. Based on the spatial boundary and shape characteristics of the high-interest areas, a set of initial candidate inspection waypoints is automatically generated. The generation rules include generating a waypoint at the geometric centroid of the polygon in the high-interest area and generating waypoints at locations with large boundary curvature or corners. An initial priority weight is assigned to each candidate inspection waypoint, and the initial priority weight is calculated based on the standardized score of the raster cell where the waypoint is located and the distance from the waypoint to the geometric center of the high-interest area.

[0034] In practical implementation, a device state constraint model is constructed by introducing the current state information of the mobile inspection equipment. The current state information of the mobile inspection equipment is reported in real time through its onboard telemetry system, including the remaining battery percentage, current latitude and longitude coordinates, and the operating status of sensors such as the hyperspectral imager and lidar scanner. The device state constraint model defines the physical limitations on the tasks performed by the mobile inspection equipment. For example, it cannot perform new long-distance inspection tasks when the remaining battery is below 20%, and it must avoid detection tasks that rely on a specific sensor when that sensor fails. Based on the device state constraint model, the initial priority weights of candidate inspection waypoints are dynamically adjusted. For example, for a candidate inspection waypoint with a high initial priority weight, if the mobile inspection equipment's current remaining battery is insufficient to support its round trip to that waypoint, the weight of that waypoint will be significantly reduced or even set to zero. Simultaneously considering the transfer distance and time cost between waypoints, multi-objective path planning is performed. The goal of path planning is to visit as many high-weight waypoints as possible while satisfying the device state constraints, while minimizing the total travel distance and time consumption. Overall cost. The model can be performed and minimized using an objective function, which is expressed as follows:

[0035] in: This represents the sum of the transfer distances between all waypoints in the planned path. This represents the estimated total time to complete visits to all waypoints. This represents the sum of the dynamically adjusted priority weights of all waypoints in the waypoint sequence. , , These are the conversion factors for distance cost, time cost, and weighted benefit, respectively. The output is a waypoint visit sequence and path planning result that satisfies equipment state constraints and achieves optimal overall cost. This is then compiled into a dynamic task scheduling instruction that includes latitude and longitude coordinates, arrival time windows, and expected action sequences, and can be parsed by the mobile inspection equipment's onboard flight control system.

[0036] In practical implementation, the system receives real-time on-site confirmation data transmitted from mobile inspection equipment during the execution of dynamic task scheduling commands. This data is uploaded immediately after the mobile inspection equipment completes a detailed inspection of a target waypoint and includes the verified true location and emission intensity of the emission source. The true location of the emission source in the on-site confirmation data is understood to be the precise geographic coordinates determined jointly by a lidar scanner and a high-precision GNSS module, while the emission intensity is the quantified carbon emission rate calculated by the source identification and assessment module. The on-site confirmation data is compared and analyzed with the integrated sensing map upon which the dynamic task scheduling commands are generated. The map prediction accuracy index is calculated, and the comparative analysis includes location deviation comparison and intensity deviation comparison. Location deviation refers to the spatial distance between the predicted high-concern area in the integrated sensing map and the actual emission source location confirmed on-site. Intensity deviation refers to the difference between the standardized score of the corresponding grid cell in the integrated sensing map and the normalized value of the emission intensity confirmed on-site. Map prediction accuracy index. It can be quantified as:

[0037] in: This refers to the number of waypoints whose on-site confirmation data has been received. It is the first Normalized positional deviation of each point It is the first Normalized intensity deviation at each point and These are the weighting coefficients assigned to the position and intensity deviations, respectively.

[0038] In practice, based on the map prediction accuracy index, the scoring threshold used to delineate high-interest areas is dynamically adjusted, and the clustering partitioning results are updated. It can be understood that the map prediction accuracy index... The lower the value, the greater the deviation between the prediction results of the integrated sensing map and the actual situation. When the map prediction accuracy index... When the score falls below the preset lower limit, the system will lower the scoring threshold for classifying high-concern areas, for example, from 80 points to 70 points, thereby including more areas in the high-concern area and reducing the risk of missed reports. This applies to the map prediction accuracy index. When the score exceeds the preset upper limit, the system can raise the scoring threshold from 80 to 85 points, making the division of high-concern areas more rigorous and precise. Based on the updated clustering and partitioning results, a replanning process for dynamic task scheduling instructions is triggered. The dynamic task scheduling module takes the latest high-concern area division results and the latest status information of the mobile inspection equipment as input, and re-generates candidate waypoints, corrects weights, and plans multi-target paths to generate optimized task scheduling instructions adapted to the latest monitoring understanding and issues them to the mobile inspection equipment.

[0039] In one embodiment of the present invention, the mobile inspection device executes dynamic task scheduling instructions to acquire multi-dimensional high-resolution feature information of the target and completes refined identification and quantitative evaluation, as well as subsequent model evolution steps. In an inspection scenario targeting a suspected leak point in a condenser tower in a petrochemical plant, the mobile inspection device parses and executes the received dynamic task scheduling instructions. Based on the latitude and longitude coordinate sequence and flight waypoints specified in the instructions, the mobile inspection device autonomously navigates to the target waypoint location through the onboard navigation and flight control system. Upon arrival at the target waypoint location, the mobile inspection device simultaneously activates its onboard hyperspectral imager, lidar scanner, and miniature gas chromatograph to perform coordinated scanning and sampling of the condenser tower and its surrounding area. The hyperspectral imager performs push-broom imaging of the outer surface of the condenser tower and the plume above it at a preset scanning angle, collecting reflectance data of the target in hundreds of narrow spectral bands, forming a hyperspectral data cube combining spatial and spectral dimensions. A lidar scanner emits laser pulses and receives the echoes reflected from the surface of the condenser tower. By measuring the laser's time-of-flight, it calculates the distance and generates high-density 3D point cloud data describing the fine 3D structure of the condenser tower surface and the approximate morphology of the leak. A miniature gas chromatograph extracts air samples from specific locations around the condenser tower via an extended sampling probe. The samples undergo rapid separation and detection inside the miniature gas chromatograph, outputting a spectrum containing gaseous components such as carbon dioxide, methane, and non-methane total hydrocarbons, along with their precise concentrations.

[0040] In practice, the hyperspectral data cube, 3D structural point cloud data, and precise concentration spectrum are synchronized in time and registered spatially. The central processor of the mobile inspection equipment adds a unified high-precision time stamp to the three sets of data, and the 3D structural point cloud data acquired by the lidar scanner is used as the spatial reference. Through coordinate transformation, the geographical location corresponding to each pixel of the hyperspectral imager is matched with the point in the 3D structural point cloud. At the same time, the sampling point location of the miniature gas chromatograph is marked on the 3D structural point cloud. The data that has completed time synchronization and spatial registration is packaged into a multi-dimensional high-resolution feature information data package of the target. A multi-modal feature fusion identification model is established, and the input of the multi-modal feature fusion identification model is the multi-dimensional high-resolution feature information data package of the target. The spectral fingerprint of a substance is extracted from the hyperspectral data cube. The spectral fingerprint refers to the absorption or reflection characteristic curve of a substance in a specific wavelength range. The extracted spectral fingerprint is matched with a preset typical emission substance spectral library. The typical emission substance spectral library stores standard spectral curves of substances such as methane and volatile organic compounds. By calculating the spectral similarity, the types of emission substances are initially identified as methane and a small amount of ethylene.

[0041] In practical implementation, three-dimensional structural point cloud data is used to assist in determining the equipment type and operating status of the emission source. The geometric shape, physical dimensions, and spatial orientation of the emission source reflected in the three-dimensional structural point cloud data are analyzed. In some embodiments, the point cloud data shows the target structure as a cylinder, approximately 15 meters high and 5 meters in diameter, with an irregularly shaped protrusion on the side and a continuously expanding sparse point cloud region. This characteristic matches the physical morphology of a gas leak occurring on the maintenance valve flange of a condenser tower. Precise concentration spectra are used to verify the initially identified emission substances. The precise concentration spectra output by the micro gas chromatograph show that the volume concentration of methane in the sampled air is significantly higher than the background value, and ethylene characteristic peaks are detected, verifying the preliminary hyperspectral identification results. Based on the precise concentration spectra, the precise volume or mass concentrations of various greenhouse gases are calculated, for example, the volume concentration of methane. The volume concentration of ethylene is 850 ppmV. The value is 120 ppmV. This can be understood as a comprehensive assessment of the carbon emission rate and total amount from a specific emission source, considering the types of emitted substances, equipment type and operating status, as well as precise volumetric or mass concentrations. This is achieved through pre-defined emission factor mapping rules and mass balance calculations. The emission factor mapping rules retrieve the corresponding carbon dioxide equivalent conversion factor from a pre-defined emission factor database based on equipment type (condenser), operating status (leakage), and the type of leaked substance (methane, ethylene). The mass balance calculation estimates the emission rate based on concentration data and leak point characteristics, providing a simplified emission rate estimation method. The estimation formula can be expressed as:

[0042] in: Representative gas components Volume concentration at the leak point, This represents the gas velocity at the leak point estimated using three-dimensional structural point cloud data. This represents the equivalent area of ​​the leak point calculated using three-dimensional structural point cloud data. Representative gas components Density under standard conditions. Ultimately, total carbon emissions are obtained by integrating the emission rates over a specific time period. See Table 1.

[0043] Table 1: Emission Factor Mapping Table

[0044] In practice, a historical case database is continuously collected, containing all detailed identification and quantitative assessment cases completed by mobile inspection equipment. This database is stored in a cloud-based collaborative processing center. Each case record includes a multi-dimensional, high-resolution feature information data package, the identified emission substance types, the determined equipment type and operating status, the calculated precise concentration, and the final quantitative assessment result. New emission source feature patterns or variations of existing emission sources are periodically extracted from the historical case database. This extraction is performed by a pattern discovery algorithm deployed in the cloud, which compares the differences in multi-dimensional, high-resolution feature information between newly added cases and historical cases.

[0045] In some embodiments, the pattern discovery algorithm identifies a novel thermal imaging spectral feature resulting from insulation material damage. This feature differs from known leakage spectral features and is marked as a newly emerging emission source feature pattern. The extracted new feature pattern or variant pattern is used to incrementally train the multimodal feature fusion identification model. During incremental training, the new feature pattern is added as a sample to the model's training set, and the classifier parameters or neural network weights within the model are adjusted, updating the model's internal identification rules and parameters. The updated multimodal feature fusion identification model is redeployed to the mobile inspection device. The model update file is sent to the mobile inspection device via an encrypted communication link. After loading the new multimodal feature fusion identification model, the mobile inspection device gains the ability to identify newly emerging or mutated emission sources.

[0046] See Figure 4 This is a bar chart comparing methane and ethylene concentrations at different detection points, clearly reflecting the gas diffusion characteristics in a petrochemical plant condenser tower leak scenario. The methane concentration is highest directly below the leak and gradually decreases with increasing distance, consistent with the physical laws of gas diffusion. The methane concentration (approximately 750 ppmV) 2 meters above the leak is slightly higher than on the left and right sides, indicating an upward diffusion trend of the leaked gas. This characteristic can provide a basis for route planning for inspection equipment. This chart directly supports the verification of the multimodal feature fusion identification model: the methane and ethylene types initially identified by hyperspectral imaging corroborate the concentration detection results from gas chromatography. The concentration gradient data can be used for subsequent emission rate calculations; combined with parameters such as leak area and flow velocity, it can more accurately quantify carbon emission intensity.

[0047] In one embodiment of the present invention, the cloud-based collaborative processing center performs subsequent data aggregation and distribution steps on the refined identification and quantitative assessment results. In an example of a city-level monitoring network covering multiple industrial parks, the cloud-based collaborative processing center continuously receives data packets uploaded by multiple mobile inspection devices from different areas and shifts via a network interface. Each data packet encapsulates the refined identification and quantitative assessment results generated by the source identification and assessment module. Specifically, the refined identification and quantitative assessment results include the unique identifier of the emission source, precise geographical coordinates, emission source type, main greenhouse gas species, real-time carbon emission rate, and calculated cumulative emission estimate.

[0048] In practice, the cloud-based collaborative processing center deduplicates, correlates, and integrates the received refined identification and quantitative assessment results to construct a regional-level carbon footprint panorama. The deduplication operation is based on the precise geographic coordinates and characteristic information of the emission sources. The cloud-based collaborative processing center compares the coordinates of newly uploaded emission sources with the coordinates of existing records in the regional-level carbon footprint panorama. When the distance between two coordinates is less than a preset spatial tolerance threshold and the emission source type is consistent, the system determines it as a duplicate report of the same emission source and retains the record with the latest timestamp or higher data confidence. The correlation operation connects emission source records with the basic geographic information database, corporate legal entity database, and energy consumption ledger, supplementing each emission source record with management attribute information such as the company name, industry classification, and registered address. The integration operation summarizes all deduplicated and correlated emission source records according to spatial grids or administrative regional units, calculating the total number of emission sources, the total emission rate of various greenhouse gases, and the cumulative emission amount within each spatial unit, constructing a structured regional-level carbon footprint panorama database that includes spatial, temporal, and management dimensions.

[0049] In practical implementation, trend analysis and pattern mining are performed on the regional-level carbon footprint panoramic view across multiple time dimensions. The time-series analysis engine of the cloud-based collaborative processing center extracts historical emission data sequences for a specified region from the regional-level carbon footprint panoramic view database. In some embodiments, trend analysis calculates daily, weekly, or monthly emission change curves and fits their trend lines to determine whether emission levels are in an upward, downward, or stable phase. Pattern mining uses clustering algorithms to identify the spatiotemporal evolution patterns of carbon emissions. For example, spatial clustering reveals the spatial distribution and clustering patterns of emission hotspots, while time-series clustering identifies enterprise groups or industrial parks with similar emission fluctuation patterns. An analytical index for quantifying the spatiotemporal evolution of emission intensity is also provided. It can be defined as:

[0050] in: Representative spatial unit At the point of time Normalized emission intensity It is a spatial unit Average emission intensity within the analysis time window, It is its standard deviation. It is a spatial unit In the most recent time interval Changes in emission intensity within the region It is its historical average change. (Indicator) The value reflects both the degree of abnormality in emission intensity and the drasticness of its changing trend.

[0051] In practice, based on preset permission and subscription rules, the regional carbon footprint panoramic view and its analysis results are distributed to different regulatory terminals or data application systems. These permission and subscription rules are configured on the management platform of the cloud-based collaborative processing center. It can be understood that the rules define the data scope, granularity, and alarm types that different user roles can access. In some embodiments, a municipal-level environmental monitoring terminal is granted access to the entire city's regional carbon footprint panoramic view and can receive alarms indicating excessive emissions across the city; while the environmental management system of a company within a specific industrial park is only subscribed to and distributed with emission source inventories and trend analysis reports related to the geographical boundaries of that company's factory area. The distribution is accomplished through secure application programming interfaces (APIs) for data push or by generating and publishing encrypted data packages, ensuring that the regional carbon footprint panoramic view and its analysis results are securely and accurately delivered to authorized regulatory terminals or data application systems.

[0052] See Figure 5 This chart illustrates the deduplication effect and quality analysis of emission source data, visually demonstrating the deduplication effect and data quality changes in different batches of data uploaded to the cloud-based collaborative processing center. In all batches, the amount of data before deduplication was significantly higher than the amount after deduplication, with the amount of data after deduplication being approximately 45%-55% of the amount before. This indicates that there were a large number of duplicate emission source records in the original uploaded data, and the cloud-based deduplication mechanism effectively reduced redundant data, improving the efficiency of subsequent analysis. This chart directly reflects the necessity and effectiveness of the "deduplication" step in the cloud-based data aggregation process, providing a basis for optimizing data cleaning rules. The fluctuation pattern of confidence level can be used to dynamically adjust the parameters of the deduplication algorithm, thereby further improving data quality. The comparison of data volume before and after deduplication also provides a quantitative reference for the planning of cloud storage and computing resources.

[0053] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An emission reduction system based on carbon footprint monitoring, characterized in that, The system includes: The multimodal sensing module, through multi-source heterogeneous sensing nodes deployed in the monitoring area, collects dynamic information on energy flow and environmental gas concentration information in parallel to form a multimodal sensing dataset; The data fusion processing module performs spatiotemporal calibration and information fusion processing on the multimodal sensing dataset to generate a comprehensive sensing map with a unified spatiotemporal reference. The dynamic task scheduling module derives and generates dynamic task scheduling instructions for the mobile inspection equipment based on the comprehensive perception map, guiding the mobile inspection equipment to accurately approach suspected high-emission targets. Source identification and assessment module: The mobile inspection device executes the dynamic task scheduling command to obtain multi-dimensional high-resolution feature information of the target, and completes the refined identification and quantitative assessment of carbon emission sources based on this. The cloud-based collaborative processing module uploads the refined identification and quantitative evaluation results to the cloud-based collaborative processing center via an encrypted data link.

2. The working method of the emission reduction system based on carbon footprint monitoring according to claim 1, characterized in that, The process involves deploying multi-source heterogeneous sensing nodes in the monitoring area to collect dynamic information on energy flow and ambient gas concentration in parallel, forming a multimodal sensing dataset. Specifically, the following operations are performed: The fixed gas concentration sensor, the embedded power metering unit, and the heat flow monitoring probe are activated simultaneously to form the multi-source heterogeneous sensing node network. The fixed gas concentration sensor periodically captures the greenhouse gas concentration readings at the monitoring points. The embedded power metering unit records the instantaneous power and cumulative power consumption of the power supply circuit. The heat flow monitoring probe measures the heat radiation flux on the surface of the pipe or equipment. The greenhouse gas concentration readings, instantaneous power and cumulative power consumption, and thermal radiation flux are locally timestamped and bound to spatial coordinates to form the original time-series sensing data stream; A multi-channel data buffer queue is established to cache and pre-sort the time-series sensing data stream, while compensating and aligning the clock deviations between different sensing nodes. Based on the preset data quality verification rules, outlier detection and missing value interpolation are performed on the time-series sensing data stream after clock compensation and alignment, invalid sampling points are removed, and standardized sensing data blocks that have undergone quality control processing are generated. All standardized sensing data blocks belonging to the same geographic grid and the same time window are aggregated and encapsulated to output a structured multimodal sensing dataset.

3. The working method of the emission reduction system based on carbon footprint monitoring according to claim 1, characterized in that, The process of performing spatiotemporal calibration and information fusion on the multimodal sensing dataset to generate a comprehensive sensing map with a unified spatiotemporal reference involves the following operations: Receive the multimodal sensing dataset and parse the greenhouse gas concentration data layer, energy consumption data layer, and thermal radiation data layer contained therein; A spatial interpolation surface model is established for the greenhouse gas concentration data layer to transform the concentration readings at discrete points into a continuous concentration distribution field covering the monitoring area; The energy consumption data layer and the thermal radiation data layer are associated and mapped to identify the spatial coupling relationship between energy consumption and thermal radiation, and to construct an energy thermal effect coupling map. The continuous concentration distribution field and the energy thermal effect coupled map are superimposed on the same geographic information base map, and pixel-level information fusion operation is performed to calculate the correlation weight between the concentration field intensity and the energy thermal effect in each geographic unit. Based on the aforementioned correlation weights, the fused data is normalized and standardized to generate a rasterized comprehensive perception map that comprehensively reflects the potential intensity and spatial distribution characteristics of carbon emissions.

4. The working method of an emission reduction system based on carbon footprint monitoring according to claim 3, characterized in that, To establish a spatial interpolation surface model for the greenhouse gas concentration data layer, the concentration readings at discrete points are transformed into a continuous concentration distribution field covering the monitoring area. Specifically, the following operations are performed: The greenhouse gas concentration data layer in the multimodal sensing dataset is analyzed to extract the spatial coordinates of all monitoring points and their corresponding concentration readings. Using the Kriging spatial interpolation algorithm, based on the spatial coordinates and concentration readings of the monitoring points, the predicted concentration value at any unsampled location within the monitoring area is calculated; Based on the digital elevation model data of the monitoring area, the concentration prediction value is corrected for topographic elevation to eliminate the influence of topographic undulation on gas diffusion; The concentration prediction values ​​after terrain elevation correction are gridded to generate a raster data layer with fixed spatial resolution. The raster data layer is smoothed and filtered to eliminate local abnormal fluctuations that may be caused by interpolation calculations, and finally outputs a continuous concentration distribution field covering the entire monitoring area.

5. The working method of an emission reduction system based on carbon footprint monitoring according to claim 1, characterized in that, Based on the comprehensive sensing map, dynamic task scheduling instructions for the mobile inspection equipment are derived and generated, specifically by performing the following operations: Load the comprehensive perception map, cluster and partition the standardized scores in the map to divide it into high-interest areas, medium-interest areas and low-interest areas; Based on the spatial boundary and shape characteristics of the high-interest area, an initial set of candidate waypoints is automatically generated, and an initial priority weight is assigned to each candidate waypoint. Introduce the current status information of the mobile inspection device, including remaining battery power, current location, and sensor operating status, and construct a device state constraint model; Based on the equipment state constraint model, the initial priority weights of the candidate inspection waypoints are dynamically adjusted, and multi-objective path planning is performed while considering the transfer distance and time cost between waypoints. Output the waypoint access sequence and path planning results that satisfy the equipment state constraints and have the best overall cost, and compile them into dynamic task scheduling instructions that can be parsed by mobile inspection equipment.

6. The working method of an emission reduction system based on carbon footprint monitoring according to claim 1, characterized in that, The mobile inspection device executes the dynamic task scheduling command to acquire multi-dimensional high-resolution feature information of the target, specifically performing the following operations: The mobile inspection equipment parses and executes the received dynamic task scheduling instructions, and autonomously navigates to the waypoint location specified in the instructions; Upon arrival at the target waypoint, the hyperspectral imager, lidar scanner, and miniature gas chromatograph are activated simultaneously to conduct a coordinated scan and sampling of the target emission source or its surrounding environment. The hyperspectral imager collects reflectance data of the target in hundreds of narrow spectral bands, forming a hyperspectral data cube; The lidar scanner emits laser pulses and receives echoes to generate point cloud data describing the three-dimensional structure of the target surface. The miniature gas chromatograph extracts on-site air samples and performs rapid separation and detection, outputting precise concentration spectra of gas components; The hyperspectral data cube, the three-dimensional structural point cloud data, and the precise concentration spectrum are synchronized in time and registered spatially, and packaged into multi-dimensional high-resolution feature information of the target.

7. The working method of an emission reduction system based on carbon footprint monitoring according to claim 6, characterized in that, The detailed identification and quantitative assessment of carbon emission sources based on this involves the following steps: A multimodal feature fusion identification model is established, wherein the input of the model is the multi-dimensional high-resolution feature information of the target; The spectral fingerprints of substances are extracted from the hyperspectral data cube and matched with a preset typical emission substance spectral library to preliminarily identify the types of emission substances. By combining the three-dimensional structural point cloud data, the external geometric shape, physical size and spatial orientation of the emission source and the emission outlet are analyzed to help determine the equipment type and operating status of the emission source. Using the precise concentration spectrum, the types of emitted substances initially identified are verified, and the precise volume or mass concentration of various greenhouse gases is calculated. By taking into account the types of emitted substances, the types and operating status of the equipment, and the precise volume concentration or mass concentration, a refined quantitative assessment of the carbon emission rate and total amount of a specific emission source is completed through preset emission factor mapping rules and mass balance calculations.

8. The working method of an emission reduction system based on carbon footprint monitoring according to claim 5, characterized in that, After deriving and generating dynamic task scheduling instructions for the mobile inspection equipment based on the comprehensive sensing map, the following adaptive optimization steps are also included: The system receives real-time on-site confirmation data transmitted back by the mobile inspection equipment during the execution of dynamic task scheduling instructions. The on-site confirmation data includes the verified actual location and emission intensity of the emission source. The on-site confirmation data is compared and analyzed with the comprehensive perception map on which the dynamic task scheduling instruction is generated, and the map prediction accuracy index is calculated. Based on the map prediction accuracy index, the scoring threshold used to divide high-interest areas is dynamically adjusted, and the clustering partitioning results are updated. Based on the updated clustering and partitioning results, a replanning process for the dynamic task scheduling instructions is triggered to generate optimized task scheduling instructions that adapt to the latest monitoring and understanding.

9. The working method of an emission reduction system based on carbon footprint monitoring according to claim 7, characterized in that, After establishing a multimodal feature fusion and identification model, the following model evolution steps are also included: Continuously collect all detailed identification and quantitative evaluation cases completed by mobile inspection equipment to form a historical case database; Periodically extract newly emerging emission source characteristic patterns or variation patterns of existing emission sources from the historical database of the aforementioned cases; The extracted new feature patterns or mutation patterns are used to incrementally train the multimodal feature fusion identification model, and update the identification rules and parameters inside the model. The updated multimodal feature fusion identification model was redeployed to the mobile inspection equipment, enabling it to identify newly emerging or mutated emission sources.

10. The working method of an emission reduction system based on carbon footprint monitoring according to claim 1, characterized in that, After the refined identification and quantitative evaluation results are uploaded to the cloud collaborative processing center via an encrypted data link, the following data aggregation and distribution steps are also included: The cloud-based collaborative processing center receives refined identification and quantitative assessment results uploaded from multiple mobile inspection devices, and performs deduplication, correlation, and integration on the refined identification and quantitative assessment results to construct a regional-level panoramic view of carbon footprint. The regional carbon footprint panoramic view is subjected to trend analysis and pattern mining in multiple time dimensions to identify the spatiotemporal evolution of carbon emissions. Based on preset permissions and subscription rules, the regional carbon footprint panoramic view and its analysis results are distributed to different regulatory terminals or data application systems.

Images