Plateau region carbon emission monitoring management method and system, electronic equipment and medium

Abstract

The invention relates to the field of environmental monitoring, and discloses a plateau region carbon emission monitoring management method and system, electronic equipment and a medium, and the method comprises the following steps: constructing an atmospheric boundary layer dynamic model for a plateau low-pressure and strong-turbulence environment, and achieving environmental adaptability modeling by correcting a turbulence diffusion coefficient and localized combustion efficiency; based on unmanned aerial vehicle group dynamic path planning and a ground station collaborative sensing network, flight parameters are adjusted in real time in combination with the concentration gradient, and time-space continuous multi-source fusion monitoring data are generated; a high-resolution carbon emission field is output through embedding a combustion efficiency correction factor and a space correlation graph structure by adopting a data assimilation algorithm with mixed ensemble Kalman filtering and a graph convolutional network; and constructing frozen soil degradation index parameters, and driving a transfer learning early warning model to realize methane flux dynamic prediction and multi-stage early warning triggering. According to the invention, the problems of insufficient carbon emission monitoring precision and lack of frozen soil degradation correlation evaluation in a plateau complex environment are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and specifically to a method, system, electronic device and medium for monitoring and managing carbon emissions in plateau areas. Background Art

[0002] Due to its special geographical and climatic conditions, the plateau area has long faced the challenges of insufficient accuracy in carbon emission monitoring and difficulty in risk assessment of permafrost degradation. Traditional carbon emission monitoring methods are mostly based on atmospheric models assuming hydrostatic equilibrium. In the plateau environment with low air pressure and strong turbulence, due to the neglect of the non-hydrostatic effect in the vertical direction and the enhancement of turbulence diffusion by terrain gravity waves, significant deviations exist in the simulation of pollutant transport.

[0003] Although existing data assimilation techniques can integrate some observed data, they fail to effectively combine the collaborative sensing advantages of unmanned aerial vehicle (UAV) dynamic cruising and ground stations, and lack a dynamic correction mechanism for the combustion efficiency in the plateau changing with air pressure and temperature, resulting in insufficient reliability of the inversion results of carbon emission source strength.

[0004] In addition, the correlation assessment of permafrost degradation and carbon emissions is mostly limited to single remote sensing or ground monitoring means, lacking a multi-factor coupling analysis model integrating surface deformation, soil temperature and the spatio-temporal distribution of carbon emissions, and it is difficult to quantify the driving effect of permafrost thawing on methane release.

[0005] At the same time, the sparsity of historical monitoring data in the plateau area limits the generalization ability of traditional machine learning models, and the timeliness and accuracy of early warning are difficult to meet the actual needs.

[0006] How to construct a monitoring-assimilation-early warning full-chain technology system adapted to the characteristics of the plateau environment has become a technical bottleneck that urgently needs to be broken through in this field. Summary of the Invention

[0007] Aiming at the deficiencies of the existing technology, the present invention provides a method, system, electronic device and medium for monitoring and managing carbon emissions in plateau areas, which solves the problems of the failure of the hydrostatic equilibrium assumption of the atmospheric model in the plateau environment, insufficient spatio-temporal coverage of monitoring data, difficulty in quantifying the coupling mechanism between permafrost degradation and carbon emissions, and weak generalization ability of the early warning model in the existing technology.

[0008] To achieve the above objectives, the present invention is realized through the following technical solutions: The first aspect of the present invention provides a method for monitoring and managing carbon emissions in plateau areas, including the following steps: Based on the characteristics of the plateau environment with low air pressure and strong turbulence, construct a dynamic model of the atmospheric boundary layer that couples the non-hydrostatic equilibrium equation and terrain gravity wave parameterization, correct the turbulence diffusion coefficient and calculate the local combustion efficiency; The dynamic cruise of the UAV swarm and the cooperation with ground stations through the multimodal perception network are used to collect multi-source environmental data in real time, where the UAV path is dynamically adjusted based on the uncertainty distribution of the concentration field output by the atmospheric boundary layer dynamic model; Input the multi-source environmental data into the data assimilation algorithm that combines the ensemble Kalman filter and the graph convolutional network to fuse the physical model prediction and the sensor observation data, and generate the spatio-temporal distribution of the carbon emission field; Based on the spatio-temporal distribution of the carbon emission field and the satellite remote sensing data, construct a permafrost degradation index and input it into the transfer learning early warning model to calculate the methane flux change rate in real time, and trigger a multi-level early warning signal when it exceeds the preset threshold.

[0009] Preferably, the construction of the atmospheric boundary layer dynamic model includes: Construct a non-hydrostatic balance equation by introducing a vertical acceleration term to describe the non-hydrostatic effect of the plateau atmosphere, and its control equation is: ; Among them, is the three-dimensional wind speed vector, is the air pressure, is the air density, is the kinematic viscosity coefficient, is the gravity term, is the Coriolis force term; The modification of the turbulent diffusion coefficient and the calculation of the local combustion efficiency include: Modify the turbulent diffusion coefficient based on the terrain gravity wave parameterization method , and the expression is: ; Among them, is the turbulent diffusion coefficient in the plain area, is the terrain slope correction factor, is the terrain elevation, is the spatial change rate of the terrain in the horizontal direction; Combine the plateau environmental temperature and the air pressure , calculate the local combustion efficiency correction factor , and its expression is: ; Among them, is the standard atmospheric pressure, is the standard temperature.

[0010] Preferably, the steps of the dynamic cruise of the UAV swarm and the cooperation with ground stations through the multimodal perception network to collect multi-source environmental data in real time include: The monitoring area is spatially partitioned based on the Venn diagram, and the initial cruise path of the UAV swarm is determined with the maximization of information entropy as the objective function. The information entropy maximization objective function is expressed as: ; where, is the carbon emission uncertainty probability of area , which is jointly calculated from the concentration field prediction results output by the atmospheric boundary layer dynamic model and historical monitoring data; The flight altitude of the UAV is dynamically adjusted according to the vertically acquired concentration gradient in real time. The adjustment formula is: ; where, is the flight altitude of the UAV, is the preset reference flight altitude, is the altitude dynamic adjustment amplitude, is the pollutant concentration gradient in the vertical direction; Meanwhile, the cruise speed of the UAV is adjusted according to the horizontal concentration gradient. The expression is: ; where, is the cruise speed of the UAV, is the maximum designed cruise speed of the UAV, is the horizontal concentration gradient, is the maximum value of the horizontal concentration gradient modulus length within the current monitoring period; The lidar data carried by the UAV and the temperature and humidity sensor data of the ground station are fused to generate a spatio-temporally continuous fused concentration field. The calculation formula is: ; where, is the spatio-temporally continuous fused concentration field, is the lidar data carried by the UAV, is the temperature and humidity sensor data of the ground station, and are the dynamic weight coefficients of the UAV and the ground station data respectively.

[0011] Preferably, the steps of inputting the multi-source environmental data into the data assimilation algorithm combining the ensemble Kalman filter and the graph convolutional network to fuse the physical model prediction and the sensor observation data to generate the spatio-temporal distribution of the carbon emission field include: Embedding a plateau environment combustion efficiency correction factor in the observation operator of the ensemble Kalman filter; Constructing the graph structure of the graph convolutional network , where, is the set of nodes, and node Represents the monitoring point, the node feature vector includes real-time concentration gradient, temperature and pressure, is a set of edges, and the edge weight is calculated jointly by the spatial distance and the concentration gradient difference, and the expression is: ; in, For nodes and The edge weights between For monitoring points The spatial coordinates of For nodes and The Euclidean distance of is the distance attenuation coefficient, For monitoring points The concentration gradient vector, For nodes and The horizontal concentration gradient mode length difference, is the maximum value of the horizontal concentration gradient modulus during the current monitoring period; The spatiotemporal correlation features output by the graph convolutional network are input into the collective Kalman filter to update the state variables. The formula is: ; in, is the state variable after assimilation, is the predicted state variable, is the observation data, is the observation operator matrix, is the Kalman gain matrix; The final output is the spatiotemporal distribution of carbon emissions.

[0012] Preferably, the calculation formula of the Kalman gain matrix is: ; in, is the forecast error covariance matrix, is the observation noise covariance matrix; The observation operator matrix The construction method is: ; in, is the observation matrix based on the physical model, It is a diagonal matrix, and its diagonal elements are the combustion efficiency correction factors of each monitoring point ; is the total number of monitoring points.

[0013] Preferably, the steps of constructing a permafrost degradation index based on the spatio-temporal distribution of the carbon emission field and satellite remote sensing data and inputting it into a transfer learning early warning model to calculate the methane flux change rate in real time and triggering multi-level early warning signals when the preset threshold is exceeded include: Obtain the surface deformation amount based on satellite remote sensing data, and combine it with the spatio-temporal distribution of the carbon emission field to calculate the permafrost degradation index, and its expression is: ; Among them, is the permafrost degradation index, is the surface deformation amount, is the spatio-temporal mean value of the spatio-temporal distribution of the carbon emission field in the permafrost degradation area, is the soil temperature monitored by the ground station, is the critical temperature of permafrost melting, which is calibrated through historical permafrost sample experiments; Input the permafrost degradation index into the transfer learning early warning model to calculate the methane flux change rate in real time. When the change rate exceeds the preset threshold, multi-level early warning signals are triggered, and the threshold is dynamically adjusted according to the carbon sink capacity of the plateau area.

[0014] Preferably, the transfer learning early warning model includes: A pre-trained convolutional neural network extracts the spatio-temporal characteristics of temperature and humidity in the Arctic permafrost area, and the input features are historical satellite infrared data and microwave radar humidity data in the Arctic region; Construct a long short-term memory network to fuse the permafrost degradation index and the spatio-temporal distribution data of the carbon emission field, and output the methane flux prediction value; When the system detects that the cumulative increase in the methane flux prediction value exceeds the dynamically calculated critical flux threshold within a continuously set early warning monitoring period, corresponding-level early warning signals will be automatically triggered according to the proportion exceeding the threshold.

[0015] The second aspect of the present invention provides a carbon emission monitoring and management system for the plateau area, including: An atmospheric boundary layer dynamic modeling module, which is used to construct an atmospheric boundary layer dynamic model that couples the non-hydrostatic balance equation and the terrain gravity wave parameterization based on the characteristics of the plateau low pressure and strong turbulence environment, and corrects the turbulent diffusion coefficient and calculates the local combustion efficiency; A multi-modal collaborative perception module, which is used to perform dynamic cruise and data acquisition tasks through a multi-modal perception network composed of an unmanned aerial vehicle group and a ground station, and the path of the unmanned aerial vehicle is adjusted in real time based on the concentration field uncertainty distribution output by the atmospheric boundary layer dynamic model; A data assimilation processing module, which is used to input multi-source environmental data into a data assimilation algorithm that combines the ensemble Kalman filter and the graph convolutional network, fuse the physical model prediction and sensor observation data, and generate a carbon emission field distribution result with high spatio-temporal resolution; A permafrost degradation warning module, which is used to fuse the spatio-temporal distribution of the carbon emission field and satellite remote sensing surface deformation data, construct a permafrost degradation index and input it into a transfer learning model, and trigger multi-level warning signals according to the methane flux change rate.

[0016] The third aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above-mentioned method is implemented.

[0017] The fourth aspect of the present invention provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned method is implemented.

[0018] The present invention provides a method, a system, an electronic device and a medium for monitoring and managing carbon emissions in the plateau area. It has the following beneficial effects: 1. By constructing a dynamic model of the atmospheric boundary layer that couples the non-hydrostatic balance equation and terrain gravity wave parameterization, the present invention breaks through the application limitations of the traditional hydrostatic balance assumption in the plateau low-pressure and strong turbulence environment, significantly improves the simulation accuracy of vertical turbulent diffusion and pollutant transport, and provides a highly reliable physical driving basis for carbon emission source strength inversion.

[0019] 2. Based on the dynamic path planning of the unmanned aerial vehicle swarm and the collaborative sensing mechanism of ground stations, combined with the flight parameter adaptive adjustment strategy driven by the concentration gradient, the present invention realizes the intelligent allocation of monitoring resources in spatio-temporally heterogeneous regions, effectively solves the data acquisition blind spots and redundancy problems under the complex terrain of the plateau, and improves the monitoring coverage efficiency and data quality.

[0020] 3. The present invention adopts a hybrid data assimilation algorithm that combines the ensemble Kalman filter and the graph convolutional network. By embedding a plateau combustion efficiency correction factor and constructing a spatial correlation graph structure, it takes into account the interpretability of the physical model and the spatio-temporal correlation mining ability of data-driven, and significantly improves the spatio-temporal resolution and consistency of carbon emission field reconstruction.

[0021] 4. The present invention innovatively proposes a multi-factor permafrost degradation index that fuses surface deformation and spatio-temporal distribution of carbon emissions, breaks through the limitations of single remote sensing or ground monitoring means, realizes the dynamic correlation analysis of the permafrost melting process and methane release risk, and provides a quantitative evaluation tool for carbon cycle research in the plateau permafrost area.

[0022] 5. The present invention constructs an Arctic-plateau permafrost feature transfer warning model based on a transfer learning framework, and solves the problem of insufficient model generalization ability caused by scarce plateau monitoring data through a collaborative mechanism of pre-trained feature extraction and plateau data fine-tuning, significantly improving the robustness of methane flux prediction and the timeliness of warning signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic flowchart of the method of the present invention; Figure 2 It is a schematic structural diagram of the device of the present invention; Figure 3 It is a schematic structural diagram of the computer device of the present invention.

[0024] Among them, 100, atmospheric boundary layer dynamic modeling module; 200, multi-modal collaborative perception module; 300, data assimilation processing module; 400, permafrost degradation warning module; 40, computer device; 41, processor; 42, memory; 43, storage medium. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0026] Please refer to the attached Figure 1 , the present invention provides a method for monitoring and managing carbon emissions in plateau areas. In view of special environmental characteristics such as low air pressure, strong turbulence and permafrost degradation in the plateau, a monitoring system that combines a multi-modal perception network and a physical-data hybrid model is constructed to realize dynamic tracking of the carbon emission field and early warning of methane release risk.

[0027] As Figure 1 shown, the method for monitoring and managing carbon emissions in the plateau area may include the following steps: S1. Based on the environmental characteristics of low air pressure and strong turbulence in the plateau, construct an atmospheric boundary layer dynamic model that couples a non-hydrostatic balance equation and terrain gravity wave parameterization; S2. Through the dynamic cruise of the unmanned aerial vehicle group and the cooperation of ground stations of the multi-modal perception network, collect multi-source environmental data in real time; S3. Input the multi-source environmental data into a data assimilation algorithm that combines ensemble Kalman filter and graph convolutional network to generate the spatio-temporal distribution of the carbon emission field; S4. Based on the spatio-temporal distribution of the carbon emission field and satellite remote sensing data, construct a permafrost degradation index and input it into a transfer learning warning model, calculate the methane flux change rate in real time, and trigger a multi-level warning signal when it exceeds a preset threshold.

[0028] The following is a detailed description of each step in the method of the present invention, comprehensively elaborating on the specific implementation principles, technical details, and processes for each step.

[0029] For step S1, in this embodiment, by constructing a dynamic model of the atmospheric boundary layer adapted to the special environment of the plateau, accurate characterization of the turbulent diffusion process and combustion efficiency is achieved. Specifically, the construction of the dynamic model of the atmospheric boundary layer includes: In view of the significant vertical motion characteristics of the plateau atmosphere, a three-dimensional acceleration term is introduced to construct a non-hydrostatic equilibrium control equation to overcome the applicability limitations of the traditional hydrostatic equilibrium assumption in the low-pressure environment of the plateau. The control equation is expressed as: ; Where: represents the three-dimensional wind speed vector (unit: m / s), including the horizontal components , and the vertical component , which is used to describe the vertical airflow disturbance in the plateau atmosphere caused by the steep terrain; is the air pressure (unit: Pa); is the air density (unit: kg / m 3 ), which is obtained by interpolating real-time meteorological data; is the kinematic viscosity coefficient (unit: m 2 / s), which is obtained from the standard atmospheric parameter table; is the gravity term, and its vertical component expression is ; is the Coriolis force term, calculated as , where is the angular velocity vector of the Earth's rotation, and its direction is related to the geographical latitude.

[0030] is to reflect the enhanced effect of the complex terrain of the plateau on turbulent diffusion, and the terrain gravity wave parameterization method is used to correct the turbulent diffusion coefficient. Specifically, the influence of the terrain slope is quantified as a correction factor, and the expression is: ; Where: is the standard turbulent diffusion coefficient in the plain area (unit: m 2 / s), which is calculated based on the Monin-Obukhov similarity theory; is the terrain slope correction factor (dimensionless), which is calibrated by fitting the historical wind speed profile and terrain elevation data; is the spatial variation rate of the terrain in the horizontal direction (unit: m / m), which is generated by the first-order difference calculation of the digital elevation model (DEM).

[0031] In view of the combustion efficiency deviation caused by the plateau low pressure and the day-night temperature difference, a combustion efficiency correction factor is constructed to quantify the influence of environmental parameters. The expression of the correction factor is: ; where: = 101325 Pa is the standard atmospheric pressure; = 298 K is the standard temperature, which is used as the reference value; is the real-time monitored temperature (unit: K), which is obtained by the ground station sensor; is the real-time monitored pressure (unit: Pa), which is dynamically collected by the barometer carried by the unmanned aerial vehicle.

[0032] Through the correction factor, the theoretical combustion efficiency value is dynamically scaled according to the actual plateau environment to ensure the accuracy of the carbon emission source strength estimation in the subsequent model.

[0033] The non-hydrostatic balance equation is discretized and solved by the finite volume method. The spatial grid resolution is set to 10 m×10 m×5 m (horizontal×vertical), and the time step is dynamically adjusted according to the Courant-Friedrichs-Lewy (CFL) condition. During the solution process, the corrected turbulent diffusion coefficient and the combustion efficiency correction factor are embedded into the momentum equation and the scalar transport equation to realize the coupled calculation of the terrain effect and the combustion process. Specifically, in the pollutant diffusion equation, the diffusion term is calculated by using the corrected value, and the combustion source term is calibrated in real time by for the theoretical emission intensity.

[0034] The initial field of the model is generated by interpolating the historical meteorological reanalysis data in the plateau region. The boundary condition is set as an open boundary, and the data connection with the global climate model is realized through the nested grid technology. Preferably, the horizontal wind speed boundary is driven by the output result of the mesoscale meteorological model WRF, and the radiation boundary condition is adopted in the vertical direction to avoid reflection errors.

[0035] For step S2, in this embodiment, through the dynamic collaborative perception mechanism of the multi-modal perception network, environmental data acquisition with high spatio-temporal resolution is achieved. Specifically, the collaborative data acquisition process between the drone swarm and the ground stations includes: In this embodiment, the monitoring area division and path planning are as follows: Based on the Voronoi Diagram, the target monitoring area is spatially divided to generate several sub-regions to optimize the coverage efficiency of the drone swarm. The division process takes the maximization of information entropy as the objective function to quantify the monitoring value of each sub-region, and its mathematical expression is: ; Wherein, represents the carbon emission uncertainty probability (dimensionless) of the sub-region , which is jointly calculated from the predicted concentration field output by the atmospheric boundary layer dynamic model and historical monitoring data.

[0036] Specifically, the root mean square error between the predicted concentration field and historical data is mapped to an uncertainty probability distribution, and the high-uncertainty region corresponds to a higher information entropy weight, thereby guiding the drones to preferentially cruise in data-scarce regions.

[0037] In this embodiment, the specific implementation of the Voronoi diagram division is as follows: Initial seed points are generated by calculating the geographical coordinates of all ground stations in the monitoring area, and the sub-region boundaries are iteratively optimized based on Delaunay triangulation. The area weight of each sub-region is inversely proportional to its carbon emission prediction error, that is, the regions with larger prediction errors are divided into denser sub-grids. The uncertainty probability is calculated by the formula: ; Wherein, is the standard deviation of the concentration prediction of the sub-region , which is calculated from the ensemble member dispersion of the ensemble Kalman filter. By dynamically updating the positions of the seed points, the drone paths are adjusted in real time according to the model prediction uncertainty.

[0038] In this embodiment, the dynamic adjustment process of the drone flight parameters is as follows: According to the pollutant concentration gradient monitored in real time, the flight altitude and cruising speed of the drones are dynamically adjusted to enhance the data acquisition pertinence in the vertical and horizontal directions: 1. Flight altitude adjustment The flight altitude of the drones is adaptively adjusted according to the vertical concentration gradient, and its calculation method is: ; Wherein: is the preset reference flight altitude, which is set based on the average terrain elevation of the monitoring area, specifically calculated by the sliding window mean filtering of the digital elevation model (DEM), and the window size is 1 km × 1 km.

[0039] is the height dynamic adjustment amplitude. To prevent the UAV from exceeding the safe flight range, the value range of is limited to 20% - 50% of the maximum climbing height of the UAV.

[0040] is the vertical pollutant concentration gradient, which is calculated in real time from the vertical profile data of the lidar carried by the UAV. The central difference method is used to calculate the concentration difference between adjacent height layers, and the formula is: ; where, is the vertical resolution of the lidar (usually 10 m).

[0041] 2. Cruise speed adjustment The horizontal cruise speed of the UAV is positively correlated with the magnitude of the horizontal concentration gradient, and its expression is: ; where: is the maximum designed cruise speed of the UAV; is the horizontal concentration gradient, which is calculated as: ; where, and are calculated by the difference of data at adjacent monitoring points on the horizontal flight trajectory of the UAV.

[0042] is the maximum value of the magnitude of the horizontal concentration gradient within the current monitoring period, which is reset at the beginning of each monitoring period and updated by the real-time data sharing of the UAV swarm in the whole area.

[0043] In this embodiment, the multi-source data fusion and spatio-temporal continuous field generation are as follows: Fuse the lidar data collected by the UAV swarm and the electrochemical sensor data of the ground station to generate a spatio-temporally continuous fused concentration field . The fusion process adopts a dynamic weighting strategy, and its calculation formula is: ; Wherein: is a spatiotemporally continuous fusion concentration field; is the lidar data carried by the unmanned aerial vehicle; is the temperature and humidity sensor data of the ground station; and are the dynamic weight coefficients of the unmanned aerial vehicle and ground station data respectively, and are dynamically adjusted according to the real-time accuracy index of the sensor, specifically: ; Wherein, is the lidar signal-to-noise ratio, is the calibration error of the electrochemical sensor (unit: %).

[0044] For the data missing problem in the uncovered area, the Kriging interpolation algorithm is used to generate a spatially continuous field, and the time series is completed by cubic spline interpolation in the time dimension.

[0045] Before data fusion, quality control is performed on the original collected data, specifically including: Based on the 3σ criterion (the Laida criterion), outliers of the sensor are removed; A time sliding window is used to smooth and filter the concentration data to eliminate the fluctuations caused by short-term environmental disturbances; For the spatiotemporal overlapping monitoring data of the unmanned aerial vehicle and the ground station, the consistency threshold is determined through cross-validation, and the conflicting data exceeding the threshold is removed.

[0046] For step S3, in this embodiment, through a hybrid data assimilation algorithm of ensemble Kalman filter and graph convolutional network, the physical model prediction and multi-source observation data are fused to generate a high spatiotemporal resolution carbon emission field distribution. Specifically, the data assimilation process includes: In this embodiment, the observation operator for the plateau environment is corrected as follows: A plateau environment combustion efficiency correction factor is embedded in the observation operator of the ensemble Kalman filter (EnKF) to construct a hybrid observation matrix adapted to the characteristics of low air pressure and strong turbulence. The mathematical expression of the observation matrix is: ; Wherein: is the physical observation matrix discretized based on the non-hydrostatic balance equation, describing the correlation between pollutant concentration and meteorological parameters under ideal conditions; is the monitoring point The combustion efficiency correction factor (defined as in step S1) is used to calibrate the systematic deviation of the combustion source strength in the plateau environment; is the total number of monitoring points.

[0047] Diagonal matrix Each element of corresponds to the real-time environmental correction coefficient of a specific monitoring point, ensuring that the observation operator dynamically reflects the spatial heterogeneity of the actual combustion efficiency.

[0048] In this embodiment, the spatio-temporal feature extraction of the graph convolutional network is as follows: Construct the graph structure of the graph convolutional network (GCN) , where the node set represents the monitoring points, and the edge set characterizes the spatial correlation strength between the monitoring points.

[0049] The node feature vector includes the real-time concentration gradient (unit: kg / m 3 / m), temperature (unit: K), and air pressure (unit: Pa). The edge weight is calculated by the following formula: ; where: and are the spatial coordinates of the monitoring points and ; is the Euclidean distance (unit: m) between node and , calculated through GPS coordinates; is the distance attenuation coefficient (unit: m), which controls the spatial correlation range. Preferably, takes a value of 1.5 times the average monitoring point spacing in the monitoring area.

[0050] and are the concentration gradient vectors of the monitoring points and ; is the difference in the horizontal concentration gradient magnitude (unit: kg / m between node and 3 / m), reflecting the spatial consistency of the pollutant diffusion direction; is the maximum value of the horizontal concentration gradient magnitude (unit: kg / m in the current monitoring period3 / m) for normalization processing.

[0051] The graph convolutional network aggregates node features through a multi-layer message passing mechanism and outputs a spatio-temporal correlation feature matrix , whose dimension is ( is the number of nodes, is the feature dimension). Preferably, a gated graph convolutional layer (GGCN) is used to suppress noise interference, and the physical interpretability of the original monitoring data is retained through skip connections.

[0052] In this embodiment, the state variable update and the assimilation field generation are as follows: The spatio-temporal correlation features output by the graph convolutional network are input into the ensemble Kalman filter to update the state variable and generate a high-precision carbon emission field. The specific process is as follows: 1. Prediction state variable calculation Based on the forward simulation of the atmospheric boundary layer dynamic model, a predicted state variable is generated, which includes a concentration field , a three-dimensional wind speed field and a turbulent diffusion coefficient . The prediction error covariance matrix is calculated through the ensemble member dispersion, and the expression is: ; where is the number of ensemble members, is the ensemble mean.

[0053] 2. Kalman gain matrix calculation The Kalman gain matrix is used to balance the weights of the model prediction and the observed data, and its calculation formula is: ; where is the observation noise covariance matrix, which is dynamically calculated from the sensor accuracy index. Preferably, different noise variances are set for the lidar and the electrochemical sensor respectively, and is constructed through block matrix combination.

[0054] 3. Analysis state variable update The predicted value is corrected using the Kalman gain matrix to obtain the assimilated analysis state variable : ; where is the observed data vector, which includes the real-time monitoring concentration values of the unmanned aerial vehicle and the ground station.

[0055] 4. Output of spatio-temporal distribution field Extract the concentration field data from the analyzed state variables to generate a high-resolution carbon emission field that is spatio-temporally continuous . Preferably, the spatial resolution is 10m×10m, the time resolution is 1 hour, and the discrete grid data is converted into a continuous field through the bilinear interpolation algorithm

[0056] For step S4, in this embodiment, by fusing the spatio-temporal distribution of the carbon emission field and satellite remote sensing data, a permafrost degradation index is constructed and a transfer learning early warning model is driven to realize the dynamic assessment and multi-level early warning of methane release risk. Specifically, the permafrost degradation early warning process includes In this embodiment, the permafrost degradation index is constructed as follows Based on the coupled analysis of satellite remote sensing surface deformation data and the spatio-temporal distribution of the carbon emission field, the synergistic effect between permafrost degradation and carbon emission release is quantified. The calculation formula of the permafrost degradation index is ; where is the surface deformation amount (unit: m) extracted by the differential interferometric measurement technology (D-InSAR) of synthetic aperture radar (SAR) satellites, and the surface vertical deformation is solved by the phase unwrapping algorithm. The time baseline is 14 days, and the spatial resolution is better than 10m×10m

[0057] is the spatio-temporal mean value of the spatio-temporal distribution of the carbon emission field in the permafrost degradation area, is the carbon emission field The spatio-temporal mean value in the permafrost degradation area (unit: kg / m 3 ), and its calculation method is ; where is the area of the permafrost degradation area (unit: m 2 ), delimited by overlay analysis of the historical permafrost distribution map and the current surface deformation anomaly area is the monitoring period duration (unit: s), set synchronously with the satellite revisit period

[0058] is the soil temperature monitored by the ground station (unit: K) is monitored in real time by a thermistor sensor buried at the ground station. The sensor depth is 50 cm, and the data sampling interval is 1 hour is the critical temperature of permafrost melting, calibrated through laboratory thermodynamics experiments. The experimental samples are taken from the typical permafrost borehole cores in the monitoring area

[0059] In this embodiment, the transfer learning early warning model is constructed as follows: Based on the pre-trained Arctic permafrost feature extraction network and the plateau monitoring data fine-tuning mechanism, a methane flux prediction model is constructed: 1. Arctic feature pre-training: Use a pre-trained convolutional neural network (CNN) to extract the temperature-humidity spatio-temporal features of the Arctic permafrost region . The input data is historical Arctic satellite infrared data (unit: K) and microwave radar humidity data (unit: %), with a spatial resolution of 1 km × 1 km and a time span of 10 years. The CNN contains 5 convolutional layers and 3 max pooling layers, and the output feature dimension is 256.

[0060] 2. Plateau feature fusion prediction: Construct a long short-term memory network (LSTM) to fuse the permafrost degradation index and the dynamic characteristics of the carbon emission field, and output the predicted methane flux value (unit: kg / m 2 / s). Specifically, the input features include: : Permafrost degradation index (unit: m 2 ·kg -1 ·K -1 ); : The maximum value of the temporal change rate of the carbon emission concentration in the permafrost region (unit: kg / m 3 / s), calculated from the temporal derivative of the carbon emission field and discretized using the central difference method.

[0061] The LSTM network contains 2 hidden layers, with 128 neurons in each layer, and the output layer is mapped to the methane flux value through a fully connected layer. The prediction formula is expressed as: ; Among them, the CNN output is used as the attention weight to dynamically adjust the dependence of the LSTM prediction result on the Arctic features.

[0062] In this embodiment, the multi-level early warning signal is triggered as follows: Compare the cumulative effect of the methane flux change rate with the dynamic threshold to trigger a hierarchical early warning: Cumulative increment calculation: In consecutive monitoring periods, calculate the cumulative increment of the predicted methane flux value : ; Dynamic threshold setting: Critical flux threshold Dynamically adjusted according to the carbon sink capacity of the plateau region, and the calculation method is as follows: ; Wherein, is the baseline value of methane flux in the same historical period (unit: kg / m 2 / s), calculated by the moving average of 5-year monitoring data; is the safety factor (dimensionless), with a value range of 1.2 - 1.5, set according to the evaluation results of the vulnerability of the regional ecosystem.

[0063] When , according to the exceeding ratio trigger multi-level early warnings: Level 1 early warning (1.0 < r ≤ 1.5): Start data review and manual inspection; Level 2 early warning (1.5 < r ≤ 2.0): Restrict industrial emissions in the surrounding area and issue public warnings; Level 3 early warning (r > 2.0): Start the emergency emission reduction plan and ecological restoration measures.

[0064] Generally speaking, the present invention first constructs a dynamic model of the atmospheric boundary layer coupling the non-hydrostatic balance equation and terrain gravity wave parameterization based on the characteristics of the plateau low-pressure and strong turbulence environment, and realizes the accurate calculation of local combustion efficiency through three-dimensional wind speed field modeling and correction of the turbulent diffusion coefficient; then, through the dynamic cruise of the unmanned aerial vehicle (UAV) swarm and the cooperation of ground stations to collect multi-source environmental data, the flight altitude and speed of the UAV are adjusted in real time according to the concentration gradient, and the lidar and ground sensor data are fused to generate a spatio-temporally continuous monitoring field; then, the multi-source data is input into a data assimilation algorithm that combines the ensemble Kalman filter and graph convolutional network. By embedding a plateau combustion efficiency correction factor and constructing a spatial correlation graph structure, the spatio-temporal distribution of a high-resolution carbon emission field is generated; finally, the carbon emission field and satellite remote sensing surface deformation data are fused to construct a permafrost degradation index, driving a transfer learning early warning model to calculate the methane flux change rate in real time, and triggering multi-level early warning signals according to dynamic thresholds, forming a full-chain closed-loop monitoring system from environmental perception, model assimilation to risk assessment.

[0065] The monitoring and management system for carbon emissions in the plateau region described below can be referred to in correspondence with the monitoring and management method for carbon emissions in the plateau region described above.

[0066] Please refer to the appendix Figure 2 , the present invention also provides a monitoring and management system for carbon emissions in the plateau region, including: An atmospheric boundary layer dynamic modeling module 100, configured to construct a dynamic model of the atmospheric boundary layer coupling the non-hydrostatic balance equation and terrain gravity wave parameterization based on the characteristics of the plateau low-pressure and strong turbulence environment, correct the turbulent diffusion coefficient, and calculate the local combustion efficiency; The multimodal collaborative perception module 200 is used to perform dynamic cruise and data collection tasks through a multimodal perception network composed of an unmanned aerial vehicle (UAV) swarm and a ground station, wherein the UAV path is adjusted in real time based on the concentration field uncertainty distribution output by the atmospheric boundary layer dynamic model; The data assimilation processing module 300 is used to input multi-source environmental data into a data assimilation algorithm that combines ensemble Kalman filtering and graph convolutional network, fuse physical model predictions and sensor observation data, and generate a carbon emission field distribution result with high spatio-temporal resolution; The permafrost degradation warning module 400 is used to fuse the spatio-temporal distribution of the carbon emission field and satellite remote sensing surface deformation data, construct a permafrost degradation index and input it into a transfer learning model, and trigger multi-level warning signals according to the methane flux change rate.

[0067] The system of this embodiment can be used to execute the above method embodiment, and its principle and technical effects are similar, so they will not be elaborated here.

[0068] Please refer to the appendix Figure 3 The present invention also provides a computer device 40, including: a processor 41 and a memory 42. The memory 42 stores a computer program executable by the processor. When the computer program is executed by the processor, the above method is executed.

[0069] The present invention also provides a storage medium 43. A computer program is stored on the storage medium 43. When the computer program is run by the processor 41, the above method is executed.

[0070] Among them, the storage medium 43 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM for short), electrically erasable programmable read-only memory (EEPROM for short), erasable programmable read-only memory (EPROM for short), programmable read-only memory (PROM for short), read-only memory (ROM for short), magnetic memory, flash memory, magnetic disk or optical disc.

[0071] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and managing carbon emissions in plateau areas, characterized in that, It includes the following steps: Based on the characteristics of the plateau low-pressure and strong turbulence environment, construct a dynamic model of the atmospheric boundary layer that couples the non-hydrostatic balance equation and terrain gravity wave parameterization, correct the turbulent diffusion coefficient, and calculate the local combustion efficiency; Through the dynamic cruise of the unmanned aerial vehicle (UAV) swarm and the cooperation of ground stations by the multi-modal perception network, collect multi-source environmental data in real time, where the UAV path is dynamically adjusted based on the uncertainty distribution of the concentration field output by the atmospheric boundary layer dynamic model; Input the multi-source environmental data into a data assimilation algorithm that combines the ensemble Kalman filter and graph convolutional network, fuse the physical model prediction and sensor observation data, and generate the spatio-temporal distribution of the carbon emission field; Based on the spatio-temporal distribution of the carbon emission field and satellite remote sensing data, construct a permafrost degradation index and input it into a transfer learning early warning model, calculate the methane flux change rate in real time, and trigger multi-level warning signals when it exceeds the preset threshold.

2. The method for monitoring and managing carbon emissions in plateau areas according to claim 1, wherein The construction of the atmospheric boundary layer dynamic model includes: Construct a non-hydrostatic balance equation by introducing the vertical acceleration term to describe the non-hydrostatic effect of the plateau atmosphere, and its control equation is: ； Among them, is the three-dimensional wind speed vector, is the air pressure, is the air density, is the kinematic viscosity coefficient, is the gravity term, is the Coriolis force term; Correction of turbulent diffusion coefficient based on topographic gravity wave parameterization method , the expression is: ； Among them, is the turbulent diffusion coefficient in the plain area, is the terrain slope correction factor, is the terrain elevation, is the spatial change rate of the terrain in the horizontal direction; Combined with the plateau ambient temperature and air pressure , calculate the localized combustion efficiency correction factor , whose expression is: ； in, is standard atmospheric pressure, is the standard temperature.

3. The method for monitoring and managing carbon emissions in plateau areas according to claim 1, characterized in that, The step of collecting multi-source environmental data in real time through the dynamic cruise of the UAV swarm and the cooperation of ground stations by the multi-modal perception network includes: Divide the monitoring area spatially based on the Voronoi diagram, and determine the initial cruise path of the UAV swarm with the maximization of information entropy as the objective function. The objective function of the maximization of information entropy is expressed as: ； Among them, is the carbon emission uncertainty probability of the area , which is jointly calculated from the concentration field prediction result output by the atmospheric boundary layer dynamic model and historical monitoring data; Dynamically adjust the UAV flight altitude according to the vertically obtained concentration gradient in real time, and the adjustment formula is: ； Among them, is the flight altitude of the drone, is the preset reference flight altitude, is the amplitude of dynamic altitude adjustment, is the concentration gradient of pollutants in the vertical direction; At the same time, adjust the UAV cruise speed according to the horizontally obtained concentration gradient, and the expression is: ； Among them, is the cruising speed of the drone, is the maximum designed cruising speed of the drone, is the concentration gradient in the horizontal direction, is the maximum value of the horizontal concentration gradient modulus within the current monitoring period; Fuse the lidar data carried by the UAV and the temperature and humidity sensor data of the ground station to generate a spatio-temporally continuous fused concentration field, and the calculation formula is: ； Among them, is a spatiotemporally continuous fusion concentration field, is the lidar data carried by the unmanned aerial vehicle, is the temperature and humidity sensor data of the ground station, and are the dynamic weight coefficients of the unmanned aerial vehicle and the ground station data, respectively.

4. The method for monitoring and managing carbon emissions in plateau areas according to claim 1, characterized in that The step of inputting the multi-source environmental data into a data assimilation algorithm that combines the ensemble Kalman filter and graph convolutional network, fusing the physical model prediction and sensor observation data, and generating the spatio-temporal distribution of the carbon emission field includes: Embed a correction factor for the combustion efficiency of the plateau environment in the observation operator of the ensemble Kalman filter; Construct the graph structure of the graph convolutional network , where is a set of nodes, and the node represents a monitoring point. The node feature vector includes the real-time concentration gradient, temperature, and air pressure is a set of edges, and the edge weights are jointly calculated by the spatial distance and the concentration gradient difference. The expression is as follows: ； Among them, is the edge weight between nodes and ; is the spatial coordinate of the monitoring point ; is the Euclidean distance between nodes and ; is the distance attenuation coefficient, is the concentration gradient vector of the monitoring point ; is the difference in the horizontal direction concentration gradient modulus between nodes and ; is the maximum value of the horizontal direction concentration gradient modulus within the current monitoring period; Input the spatio-temporal correlation features output by the graph convolutional network into the ensemble Kalman filter to update the state variables, and its formula is: ； Among them, is the assimilated state variable, is the predicted state variable, is the observed data, is the observation operator matrix, is the Kalman gain matrix; Finally, output the spatio-temporal distribution of the carbon emission field.

5. The method for monitoring and managing carbon emissions in plateau areas according to claim 4, wherein The calculation formula of the Kalman gain matrix is: ； Among them, is the prediction error covariance matrix, is the observation noise covariance matrix; The said observation operator matrix is constructed as follows: ； Among them, is the observation matrix based on the physical model, is a diagonal matrix, and its diagonal elements are the combustion efficiency correction factors of each monitoring point ; is the total number of monitoring points.

6. The method for monitoring and managing carbon emissions in plateau areas according to claim 1, wherein The step of constructing a permafrost degradation index based on the spatio-temporal distribution of the carbon emission field and satellite remote sensing data, inputting it into a transfer learning early warning model, calculating the methane flux change rate in real time, and triggering multi-level warning signals when it exceeds the preset threshold includes: Obtain the surface deformation amount based on satellite remote sensing data, and combine it with the spatio-temporal distribution of the carbon emission field to calculate the permafrost degradation index, and its expression is: ； Among them, is the permafrost degradation index, is the surface deformation amount, is the spatio-temporal mean of the spatio-temporal distribution of the carbon emission field in the permafrost degradation area, is the soil temperature monitored by the ground station, is the critical temperature of permafrost melting, which is calibrated through experiments on historical permafrost samples; Input the permafrost degradation index into the input transfer learning early warning model to calculate the change rate of methane flux in real time. When the change rate exceeds the preset threshold, multi-level warning signals are triggered, and the threshold is dynamically adjusted according to the carbon sink capacity of the plateau region.

7. The method for monitoring and managing carbon emissions in plateau areas according to claim 6, wherein, The transfer learning early warning model includes: A pre-trained convolutional neural network extracts the temperature-humidity spatio-temporal features of the Arctic permafrost region, and the input features are the historical satellite infrared data and microwave radar humidity data of the Arctic region; Construct a long short-term memory network to fuse the permafrost degradation index and the spatio-temporal distribution data of the carbon emission field, and output the methane flux prediction value; When the system detects that the cumulative increase in the predicted methane flux value exceeds the dynamically calculated critical flux threshold within a continuously set warning and monitoring period, it will automatically trigger warning signals at corresponding levels according to the proportion exceeding the threshold.

8. A carbon emission monitoring and management system for plateau areas, which is used to execute the method according to any one of claims 1-7, characterized in that, Including: An atmospheric boundary layer dynamic modeling module, which is used to construct a dynamic model of the atmospheric boundary layer that couples the non-hydrostatic balance equation and terrain gravity wave parameterization based on the characteristics of the plateau low pressure and strong turbulence environment, and corrects the turbulent diffusion coefficient and calculates the local combustion efficiency; A multi-modal collaborative perception module, which is used to perform dynamic cruise and data acquisition tasks through a multi-modal perception network composed of an unmanned aerial vehicle (UAV) swarm and ground stations, and the UAV path is adjusted in real time based on the concentration field uncertainty distribution output by the atmospheric boundary layer dynamic model; A data assimilation processing module, which is used to input multi-source environmental data into a data assimilation algorithm that hybridizes the ensemble Kalman filter and graph convolutional network, fuses the physical model prediction and sensor observation data, and generates a high spatio-temporal resolution carbon emission field distribution result; A permafrost degradation warning module, which is used to fuse the spatio-temporal distribution of the carbon emission field and satellite remote sensing surface deformation data, construct a permafrost degradation index and input it into a transfer learning model, and trigger multi-level warning signals according to the methane flux change rate.

9. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1-7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method according to any one of claims 1-7.

Images