Power plant carbon dioxide emission quantification method and system based on satellite observation data

Abstract

The invention provides a power plant carbon dioxide emission quantification method and system based on satellite observation data. The method comprises the following specific steps: extracting and screening observation XCO2 data with an inversion condition in a set space range around a target power plant, obtaining background XCO2 optimal estimation by adopting robust local regression, further obtaining a background baseline, obtaining robust observation enhanced XCO2 data, and obtaining the optimal estimation of the background XCO2. And inputting the estimated value into an emission inversion model to realize emission rate inversion at the transit moment of the power plant, and repeatedly iteratively operating for at least 1000 times to output an estimated value set of emission intensity to judge that the transit inversion result is effective, so that a matched quality control mechanism is formed. According to the method, manual intervention and experience setting dependence are reduced through automatic screening and robust modeling, inversion stability, calculation efficiency and engineering applicability are improved, and the method has obvious advantages in a large-range and long-time-sequence power plant COemission monitoring and list verification scene.

Description

Technical Field

[0001] This invention relates to the field of greenhouse gas emission monitoring and remote sensing quantitative inversion technology, specifically to a method and system for quantifying carbon dioxide emissions from power plants based on satellite observation data. This method utilizes column-average carbon dioxide concentration (XCO2) observation data acquired by a spaceborne hyperspectral passive remote sensing satellite, combined with meteorological reanalysis wind field data, to rapidly invert and determine the effectiveness of carbon dioxide emission intensity from fixed point sources such as power plants. Background Technology

[0002] Driven by the dual carbon targets and emission regulation requirements, accurate accounting and dynamic monitoring of carbon dioxide emissions from large stationary sources such as power plants are of great significance for emission inventory compilation, emission reduction effectiveness assessment, and abnormal emission identification.

[0003] Existing emission accounting methods mainly include: bottom-up statistical accounting based on fuel consumption and emission factors, online chimney monitoring (CEMS), and on-site sampling / verification. These methods generally have certain limitations in practical applications. For example, statistical accounting is heavily reliant on activity data and emission factors, suffers from delayed updates and accumulated uncertainty; online monitoring coverage is limited by construction and maintenance costs, and has limited independent third-party verification capabilities; on-site verification and airborne monitoring are costly and difficult to implement routine and large-scale synchronous assessments. With the operationalization of high-precision greenhouse gas remote sensing satellites such as OCO-2 / 3, using satellite observations to retrieve point source emission intensity from a "top-down" perspective has become an important supplement. However, conducting power plant CO2 emission retrieval based on strip / sparse footprint satellite observations still faces several key technical challenges: firstly, whether satellite overflight observations possess sufficient effective footprints and significantly enhanced signals requires objective screening criteria, but...

[0004] First, the sparse satellite overflight observation footprint and limited spatial coverage make it difficult to objectively determine the observation availability. Second, the background extraction is not robust due to spatial variations in background concentration, interference from other sources and sinks, and observation noise. The separation of the enhanced signal and the background signal is prone to bias, resulting in large separation errors for the enhanced signal. Third, wind field uncertainty and the selection of plume diffusion parameters have a significant impact on the inversion results, making the inversion process sensitive to wind direction deviation and initial parameter settings. It is prone to non-convergence or getting trapped in local optima, significantly affecting the stability of inversion based on models such as Gaussian plumes. Fourth, the lack of quality control and validity judgment mechanisms to match the inversion results easily leads to the mixed use of "invertible" and "non-invertible" cases, resulting in inconsistent inversion results and affecting the reliability and operational application of the results.

[0005] Therefore, there is an urgent need for a method for quantifying carbon dioxide emissions from power plants that is designed for passive satellite overflight observation scenarios, can filter observation availability, robustly separate background and enhancement, achieve stable inversion under limited prior conditions, and has effective discrimination. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a method and system for quantifying carbon dioxide emissions from power plants based on satellite observation data. This method utilizes column-average carbon dioxide concentration (XCO2) observation data acquired by a spaceborne hyperspectral passive remote sensing satellite, combined with meteorological reanalysis wind field data, to rapidly invert and determine the effectiveness of carbon dioxide emission intensity from fixed point sources such as power plants. This provides technical support for rapid verification of power plant emissions and also provides high-precision power plant emission information for the compilation of emission inventories.

[0007] To achieve the above-mentioned technical objectives, this invention provides a method for quantifying carbon dioxide emissions from power plants based on satellite observation data, the specific steps of which are as follows:

[0008] S1. Extract satellite observation XCO2 data within a defined spatial range around the target power plant; select observation XCO2 data that meet the inversion conditions based on constraints such as the number of effective detection footprints, the number of significantly enhanced footprints, and the consistency between the enhancement direction and the wind direction;

[0009] S2. Robust local regression is used to obtain the optimal estimate of background XCO2 from the observed XCO2 data selected in step S1, thereby obtaining the background baseline. The background baseline is then subtracted from the observed XCO2 data selected in step S1 to obtain robust observation-enhanced XCO2 data.

[0010] S3. Input the observation-enhanced XCO2 data extracted in step 2 into the emission inversion model, form reasonable boundary values ​​for unknown parameters and provide robust initial values ​​through Monte Carlo simulation, and use the trust region optimization algorithm to solve for emission intensity and diffusion-related parameters to realize emission rate inversion at the time of power plant passage.

[0011] S4. Iterate through step S3 at least 1000 times, outputting a set of estimated emission intensity values, evaluate the standard deviation and mean of the set, and if the proportion is less than 25%, then the transit inversion result is deemed valid, thus forming a supporting quality control mechanism.

[0012] A further technical solution of the present invention: In step S1, passive satellite overpass observation XCO2 data within a 60-kilometer radius around the emission source of the target power plant are selected. The number of effective detection footprints within this range should be greater than 150. The effective observation XCO2 data is extracted using QAflag or average kernel. The observation-enhanced XCO2 data is extracted from the effective observation XCO2 data using a standard Gaussian distribution. Combined with wind direction discrimination, the observation-enhanced XCO2 data that meet the inversion conditions are selected.

[0013] The specific method for extracting valid XCO2 observation data using QAflag is as follows: if QAflag is 0, it is considered a valid detection; if it is 1, it is considered an invalid detection. This QAflag is obtained based on the passive satellite concentration inversion average kernel AK. If the orbital data does not provide QAflag, the average kernel is used for judgment. If AK > 0.7, then QAflag = 0 is assigned.

[0014] A further technical solution of the present invention: In step S2, each observation position is obtained by minimizing the loss function Loss. Background concentration at The optimal estimate is obtained by averaging the optimal background concentration across all locations. Calculate the baseline CO2 background concentration near the monitored emission source according to formula ①. Then, calculate the enhancement concentration at each location according to formula ②. ;

[0015] ①;

[0016] Each observation location Background concentration at The optimal estimate is calculated using the following formula:

[0017] ③;

[0018] ④;

[0019] In the above formula, Y(x i ) indicates that in x i The CO2 concentration observed at a specific location, i.e., XCO2 observed by the satellite at that location; b(x i ) is x i Background XCO2 concentration at the location; s(x i The increase in CO2 concentration is due to CO2 emissions from power plants; e i The instrument measurement error is represented by n; n is the number of XCO2 observations. It characterizes the proportion of XCO2 enhancement caused by CO2 emissions from power plants in the measurements involved in the regression; express Measurement error at the location;

[0020] Through the robust extraction described above, the observation data from satellite remote sensing are... Distinguished as background concentration baseline Enhanced concentration at each observation point Subsequently, it participated in further emissions inversion.

[0021] A further technical solution of the present invention: In step S3, concentration inversion is performed by matching the simulated XCO2 concentration with the observed XCO2 concentration. The CO2 distribution characteristics downwind of the emission source are assumed to be a vertically integrated Gaussian plume distribution, then:

[0022] ⑤;

[0023] ⑥;

[0024] In the formula, The CO2 column concentration simulated by the Gaussian plume model. To enhance the CO2 column concentration in the Gaussian plume model simulation. The background baseline for CO2 column concentration simulated by the Gaussian plume model; M air M CO2 These represent the molar masses of atmospheric molecules and CO2 molecules, respectively; g is the acceleration due to gravity, in m / s². 2 ;P surf ρ is the surface pressure; w is the atmospheric water vapor column concentration, kg / m³. 2 Q represents carbon emission intensity, g / s; u represents wind speed, m / s; x0 represents 1000 meters; a and b represent σ. y The diffusion coefficient of (x);

[0025] Based on the trust region optimization algorithm, the unknown parameters (Q,u,a,b,...) involved in the Gaussian plume model are... To achieve an exact solution, no further assumptions are introduced; the parameters to be solved are unified to the parameter domain vector x = (Q, u, a, b, ... Monte Carlo simulations are used, and this process is repeated at least 10,000 times to construct a robust solution space for the unknown parameters. During these iterations, the boundaries of the solution space are obtained through Monte Carlo simulations to determine the maximum and minimum values ​​of each unknown parameter. The average value of the corresponding unknown parameters is used as the initial input x for the confidence region algorithm. k In the current initial input x k At this point, a subproblem is constructed with the least squares function as the objective, as shown in the following formula:

[0026] ⑦,

[0027] It is an approximation of the Hessian matrix of the objective function f. This represents the gradient of the objective function; If the change in the parameter vector to be solved is represented, then the search domain is defined as:

[0028] ⑧,

[0029] ⑨;

[0030] In the formula, α represents the step size;

[0031] The formula for the confidence region algorithm is as follows:

[0032] ⑩;

[0033] In the formula, f(x) is the fitness function, used to adapt the simulated XCO2 concentration to the observed XCO2 concentration;

[0034] These are equality constraints and inequality constraints; It is the set of unknown parameters (Q,u,a,b, ...) to be solved in the Gaussian plume model. ;

[0035] In this process This represents the maximum and minimum values ​​of the unknown parameters corresponding to 10,000 solutions in a genetic algorithm (GA).

[0036] when satisfy Error limit within the allowable range These are the unknown parameters of the final result;

[0037] The fitness function f is defined as follows: ⑪;

[0038] Where each k represents a single observation point in a satellite transit strip, k = (x, y); m represents the total number of valid observation points in the observation strip, obtained from step S1;

[0039] Its parameters include the emission inversion value Q required by the present invention, which represents the power plant emission measurement results at the moment of satellite transit.

[0040] A further technical solution of the present invention: In step S4, Monte Carlo simulation technology is used to perform 1,000 inversions on step S3, and then the standard deviation of the emission value Q is calculated. Calculation as follows ⑫;

[0041] in, The carbon dioxide emission value obtained from the i-th inversion; The average emissions value from 1000 inversions; N=1000; if If the emissions are less than 25%, the emissions inversion results are valid.

[0042] The preferred technical solution of this invention is as follows: In step S1, the effective observed XCO2 data is fitted using a standard Gaussian distribution function to obtain the approximate maximum position of the observed XCO2. The angle between the vector connecting this position and the emission source location and the unit north vector in the geodetic coordinate system is used to obtain the azimuth angle W1 of the approximate maximum XCO2 relative to the emission source. The average wind direction of ERA 5 obtained at different altitudes from 10 to 400 m is calculated as W2. Effective XCO2 data satisfying |w1-w2| less than 60° constitutes a transit observed XCO2 sequence with inversion conditions. The standard Gaussian distribution function is:

[0043] ⑬,

[0044] Where μ is the mean of the standard Gaussian distribution of the effective XCO2 data; The standard deviation of valid XCO2 data.

[0045] The preferred technical solution of this invention: For the calculation in step S2... and According to formula 14, the concentration of CO2 in a vertically integrated column can be converted. , Unit: g / m 2 The specific formula (14) is as follows:

[0046] ⑭.

[0047] In step S2 The specific calculation process is as follows:

[0048] ⑮;

[0049]

[0050] In the formula, Obtained by global least squares regression estimation, i.e., all All are involved in the calculation;

[0051] express The measurement error at that point can be determined through the measurement error information in the satellite's official documents.

[0052] The present invention also provides a power plant carbon dioxide emission quantification system based on satellite observation data, which is used to execute the power plant carbon dioxide emission quantification method based on satellite observation data as described in any one of claims 1 to 7; the system specifically includes a data acquisition and screening module, an enhanced concentration data extraction module, an emission inversion module, and an uncertainty analysis module;

[0053] The data acquisition and filtering module is used to extract satellite observation XCO2 data within a set spatial range around the target power plant; and to filter observation XCO2 data that meet the inversion conditions based on the number of effective detection footprints, the number of significantly enhanced footprints, and the consistency between the enhancement direction and the wind direction.

[0054] The enhanced concentration data extraction module is used to obtain the background baseline of the satellite observation samples, and further introduces a Gaussian mixture model to perform unsupervised clustering on the observation data, distinguishing between background observations less affected by emissions and enhanced observations more affected by emissions, so as to obtain more robust observation-enhanced XCO2 data.

[0055] The emission inversion module is used to receive the observed enhanced XCO2 data output by the enhanced concentration data extraction module, form reasonable boundary values ​​for unknown parameters through Monte Carlo simulation and provide robust initial values, and use the trust region optimization algorithm to solve for emission intensity and diffusion-related parameters, so as to realize the emission rate inversion at the time of power plant passage.

[0056] The uncertainty analysis module enhances the correlation index between the observed enhanced XCO2 data in the concentration data extraction module and the enhanced data simulated by the emission inversion module. When the correlation meets the threshold condition, the transit inversion result is determined to be valid; otherwise, it is determined to be invalid or needs to be removed, thus forming a supporting quality control mechanism.

[0057] A further technical solution of the present invention: the system further includes a processor and a memory, the memory being used to store program instructions, and the processor being used to call the stored instructions in the memory to execute the above-mentioned method for quantifying carbon dioxide emissions from power plants based on satellite observation data.

[0058] A further technical solution of the present invention: the system further includes a readable storage medium, on which a computer program is stored, and when the computer program is executed, it implements the above-mentioned method for quantifying carbon dioxide emissions from power plants based on satellite observation data.

[0059] This invention integrates column-average carbon dioxide concentration (XCO2) observation data acquired through spaceborne passive remote sensing with ERA5 meteorological reanalysis wind field data. Constrained by atmospheric transport, diffusion, and mass conservation mechanisms, it constructs an integrated technical process: "observation availability screening—robust background extraction—improved Gaussian plume inversion—validity judgment." Robust separation of background and enhanced signals is achieved through local robust regression and Gaussian mixture models (GMMs). Monte Carlo parameter boundary construction and trust region optimization are combined to enable rapid inversion and quantification of CO2 emission rates from power plant point sources. Simultaneously, a result validity judgment mechanism based on observation-simulation matching degree (correlation index) is introduced, forming a traceable quality control closed loop. This method can provide technical support for the quantification, verification, and regulation of CO2 point source emissions in the power industry, the identification of abnormal emissions, and the optimization of emission inventory reporting values ​​and the compilation of more refined facility-level point source inventories.

[0060] Currently, high-precision CO2 point source inversion data with consistent methodology and reproducible quality control at the facility scale remains relatively scarce. This invention can generate comparable emission estimation sequences at the satellite transit scale, compensating for the lack of existing data supply. Compared to traditional inversion processes that rely on manual selection of plumes, subjective background settings, or strong dependence on initial parameter values, this invention reduces human intervention and reliance on experience-based settings through automated screening and robust modeling, improving inversion stability, computational efficiency, and engineering applicability. It has significant advantages in scenarios involving large-scale, long-term power plant CO2 emission monitoring and inventory verification. Attached Figure Description

[0062] Figure 1 This is an overall flowchart of the present invention;

[0063] Figure 2 The method described in the examples is used to retrieve satellite cloud images of observation points for power plant carbon dioxide emissions.

[0064] Figure 3 It is a fitting curve between simulated data and observed data when inverting the carbon dioxide emissions of power plants using the method in the examples.

[0065] Specific implementation methods

[0066] The present invention will be further described below with reference to embodiments. The technical solutions shown in the following embodiments are specific applications of the present invention and are not intended to limit the scope of the claimed invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present invention.

[0067] To obtain the optimal method for top-down point source emission monitoring and quantification, and to achieve accurate understanding of carbon dioxide emission rates and dynamic monitoring of emission sources, this invention proposes using passive satellite observation data and 10 m and 100 m u and v wind speed data from the European Centre for Mesoscale Weather Forecasting. First, it determines whether the satellite transit observation data for the power plant area to be analyzed supports further analysis. Then, a robust algorithm is used to separate background and enhanced concentrations. An improved Gaussian plume model is used to invert the point source emission rate during satellite transit. Finally, the correlation between observed and simulated concentrations is used to determine whether the emission inversion is valid.

[0068] Example 1 provides a method for quantifying carbon dioxide emissions from power plants based on satellite observation data. See the technical flowchart of this invention. Figure 1 Specifically, it includes the following steps:

[0069] S1. Extract satellite observation XCO2 data within a defined spatial range around the target power plant; select observation XCO2 data that meet the inversion conditions based on constraints such as the number of effective detection footprints, the number of significantly enhanced footprints, and the consistency between the enhancement direction and the wind direction;

[0070] First, in Example 1, passive satellite transit trajectory data within a 60-kilometer radius of the target emission source were selected as potentially usable XCO2 data. The number of effective detection footprints within this range should be greater than 150. The definition of effective detection is derived from the QAflag during passive satellite inversion; a QAflag of 0 indicates effective detection, while a value of 1 indicates invalid detection. This QAflag is obtained based on the passive satellite concentration inversion average kernel (AK). If the orbital data does not provide a QAflag, this invention uses the average kernel for judgment; if AK > 0.7, QAflag is assigned the value 0.

[0071] Secondly, in the effective observation of XCO2 data, it should be ensured that there are more than 20 footprint points with XCO2 concentrations exceeding 2σ. c , σ c The standard deviation of the XCO2 data for the selected range.

[0072] Subsequently, the XCO2 data selected in the above steps were fitted using the standard Gaussian distribution function according to formula ①, where μ is the mean of the standard Gaussian distribution, and... The standard deviation is used to calculate the azimuth angle (w1) between the fitted μ (approximately representing the location of the maximum XCO2 value) and the emission source location. Specifically, a standard Gaussian distribution function is used to fit the effective observed XCO2 data to obtain the approximate maximum XCO2 location. The angle between the vector connecting this location and the emission source location and the unit north vector in the geodetic coordinate system is used to obtain the azimuth angle W1 of the approximate maximum XCO2 value relative to the emission source. The average wind direction of the ERA 5 obtained at different heights from 10 to 400 m is calculated as w2, and |w1-w2| should be less than 60°. If all the above conditions are met, the invention considers the detection to be effective and subsequent operations can be performed. The standard Gaussian distribution function is as follows:

[0073] ①.

[0074] S2. Robust local regression is used to obtain the optimal estimate of background XCO2 from the XCO2 data filtered in step S1, thus obtaining the background baseline. The background baseline is then subtracted from the XCO2 data filtered in step S1 to obtain robust observation-enhanced XCO2 data. In Example 1, to effectively extract the enhanced concentration of CO2 caused by power plant emissions for further emission verification, a system based on local robust regression and the GMM algorithm was designed to extract both enhanced and non-enhanced CO2 concentrations, thereby accurately extracting the enhanced CO2 concentration under usable orbits.

[0075] Local robust regression is a commonly used method for obtaining background concentration baselines and is widely applied in spectral signal processing. Compared to directly extracting enhancement concentrations, background concentrations are smoother and easier to extract. In this step, the background concentration is first extracted as a baseline using local robust regression, and then the enhancement concentration is obtained by subtracting the baseline from the overall concentration observed by satellite. The details are as follows:

[0076] The XCO2 data downwind of the power plant monitored by passive satellites can be represented by Equation ②. Under this condition, the loss function (Loss) can be minimized to obtain the data for each observation location. Background concentration at Optimal estimate.

[0077] ②;

[0078] In formula ②Y(x i ) indicates that in x i The CO2 concentration observed at a specific location, i.e., XCO2 observed by the satellite at that location; b(x i ) is x i Background XCO2 concentration at s(x) i The value of e is the increase in CO2 concentration caused by CO2 emissions from power plants.i This is due to instrument measurement error;

[0079] Minimize the loss function (Loss) as shown in formula ③:

[0080] ③;

[0081] In formula ③, n represents the number of XCO2 observations. The proportion of XCO2 enhancement due to power plant CO2 emissions in the regression measurements, representing the overall impact, can be calculated using formula ⑤:

[0082] ④,

[0083] ⑤;

[0084] In formula ④, Obtained by global least squares regression estimation, i.e., all All are involved in the calculation; express The measurement error at that point can be determined through the measurement error information in the official satellite documentation.

[0085] Robust local regression can be used to obtain the values ​​at each measurement location. Optimal estimation of background concentration

[0086] The optimal estimate of background concentration across all locations is obtained by averaging the background concentration at all locations. This allows for the preliminary acquisition of a baseline of CO2 background concentration near the monitored emission source. Formula 6 is defined as the enhanced concentration extraction value at each position, while Formula 7 is the enhanced concentration extraction value at each position.

[0087] ⑥,

[0088] ⑦;

[0089] The robust extraction steps described above can be used to extract observation data from satellite remote sensing. Distinguished as background concentration baseline Enhanced concentration at each observation point This data is then used for further emission inversion. The satellite-observed XCO2 data selected in step S1 is then input into step 2 to obtain robust enhanced XCO2 sequence values.

[0090] S3. Input the observation-enhanced XCO2 data extracted in step 2 into the emission inversion model. Use Monte Carlo simulation to establish reasonable boundary values ​​for unknown parameters and provide robust initial values. Employ a trust region optimization algorithm to solve for emission intensity and diffusion-related parameters, thereby achieving emission rate inversion at the moment the power plant passes through. In Example 1, an increase in CO2 concentration in the downwind region of the power plant will cause a larger XCO2 value in the corresponding region based on satellite observation data. In this example, we use atmospheric transport simulation to perform concentration inversion by matching the simulated concentration with the observed concentration. The CO2 distribution characteristics downwind of the emission source are assumed to be a vertically integral Gaussian plume distribution, as shown in formulas ⑧-⑩:

[0091] ⑧,

[0092] ⑨

[0093] In the formula, The CO2 column concentration simulated by the Gaussian plume model. To enhance the CO2 column concentration in the Gaussian plume model simulation. The background baseline for CO2 column concentration simulated by the Gaussian plume model; M air M CO2 These represent the molar masses of atmospheric molecules and CO2 molecules, respectively; g is the acceleration due to gravity, in m / s². 2 ;P surf ρ is the surface pressure; w is the atmospheric water vapor column concentration, kg / m³. 2 Q represents carbon emission intensity, g / s; u represents wind speed, m / s; x0 represents 1000 meters; a and b represent σ. y The diffusion coefficient of (x).

[0094] Since the XCO2 measured by the satellite is essentially the mixing ratio of CO2 column concentration to dry air, while the simulated values ​​obtained from formulas ⑧ and ⑨ are CO2 column concentrations. and According to formula 10, it can be converted into the vertical integral CO2 column concentration. , (Unit: g / m) 2 ),

[0095] ⑩,

[0096] The embodiment employs a trust region-based optimization algorithm to optimize the unknown parameters (Q,u,a,b,...) involved in the Gaussian plume model. We perform an exact solution without introducing any further assumptions. The fitness function f is defined by formula ⑪:

[0097] ⑪

[0098] This invention employs Monte Carlo simulation, repeating the process 10,000 times to construct a robust solution space for the unknown parameters. During these iterations, the boundaries of the solution space are obtained through Monte Carlo simulation to determine the maximum and minimum values ​​of each unknown parameter. The average values ​​of the corresponding unknown parameters are used as the initial input conditions for the confidence region algorithm. The confidence region algorithm is an iterative algorithm for solving nonlinear constrained optimization problems, and its detailed principle is shown in formula ⑫:

[0099] ⑫

[0100] f(x) is the fitness function in formula 11, used to adapt the simulated column concentration to the observed column concentration; These are equality constraints and inequality constraints; It is the set of unknown parameters (Q,u,a,b, ...) to be solved in the Gaussian plume model. .

[0101] In this process, This represents the maximum and minimum values ​​of the unknown parameters corresponding to 10,000 solutions in a genetic algorithm (GA).

[0102] The average value of the corresponding unknown parameters is used as the initial input x of the trust region algorithm. k In the current initial input x k At this point, a subproblem is constructed with the least squares function as the objective, as shown in formula 13.

[0103] ⑬

[0104] It is an approximation of the Hessian matrix of the objective function f. This represents the gradient of the objective function; Let T represent the change in the parameter vector to be solved, and let T be the time of each real-time observation; then the search domain is defined as:

[0105] ⑭

[0106] ⑮

[0107] α represents the step size, when +1 satisfies Error limit within the allowable range +1 represents the unknown parameter of the final result, which includes the emission inversion value Q required by the present invention. This value represents the power plant emission measurement result at the instant the satellite passes overhead.

[0108] The XCO2 enhancement sequence extracted in step S2 is used as input for step S3. The emission inversion value Q is obtained through the emission inversion model developed in step S3.

[0109] S4. Iterate through step S3 1000 times, outputting a set of estimated emission intensity values. Evaluate the standard deviation and mean of the set. If the proportion is less than 25%, the transit inversion result is considered valid, thus forming a corresponding quality control mechanism. To evaluate the validity of the emission inversion results in step S3, the embodiment includes an uncertainty analysis module to assess the robustness of the inversion results. Due to the characteristics of the heuristic search algorithm in S3, the parameter estimation results have slight differences in each iteration, causing the Q (emission) estimate to fluctuate within a certain range. This invention uses Monte Carlo simulation technology to perform 1000 inversions (i.e., repeat step S3 1000 times) on each DQ-1 transit case, and then calculates the standard deviation of the emission value Q. As shown in formula 16:

[0110] 16;

[0111] in, The carbon dioxide emission value obtained from the i-th inversion; The average emissions value from 1000 inversions; N=1000; if If the emissions are less than 25%, the emissions inversion results are valid.

[0112] Appendix Figure 2 Figure b is a satellite cloud image of a specific observation point used in Example 1 to quantify carbon dioxide emissions from power plants based on satellite observation data. Figure b is a schematic diagram of Figure a. Figure 3 The figure shows the intermediate process of using the algorithm designed in the application to invert the instantaneous emission rate of a power plant when a passive satellite passes over it. The light-colored dots in the figure represent the measured XCO2 data of the passive satellite, and the dots in the dark-colored areas represent the XCO2 data fitted by the model. By inputting these two data into step S3, the emission rate of the power plant can be obtained (the optimal solution for the emission rate can be obtained when the simulated XCO2 best matches the measured XCO2).

[0113] Example 2 provides a power plant carbon dioxide emission quantification system based on satellite observation data, used to execute the power plant carbon dioxide emission quantification method based on satellite observation data in Example 1; the system specifically includes a data acquisition and screening module, an enhanced concentration data extraction module, an emission inversion module, and an uncertainty analysis module;

[0114] The data acquisition and filtering module is used to extract satellite observation XCO2 data within a set spatial range around the target power plant; and to filter observation XCO2 data that meet the inversion conditions based on the number of effective detection footprints, the number of significantly enhanced footprints, and the consistency between the enhancement direction and the wind direction.

[0115] The enhanced concentration data extraction module is used to obtain the background baseline of the satellite observation samples, and further introduces a Gaussian mixture model to perform unsupervised clustering on the observation data, distinguishing between background observations less affected by emissions and enhanced observations more affected by emissions, so as to obtain more robust observation-enhanced XCO2 data.

[0116] The emission inversion module is used to receive the observed enhanced XCO2 data output by the enhanced concentration data extraction module, form reasonable boundary values ​​for unknown parameters through Monte Carlo simulation and provide robust initial values, and use the trust region optimization algorithm to solve for emission intensity and diffusion-related parameters, so as to realize the emission rate inversion at the time of power plant passage.

[0117] The uncertainty analysis module enhances the correlation index between the observed enhanced XCO2 data in the concentration data extraction module and the enhanced data simulated by the emission inversion module. When the correlation meets the threshold condition, the transit inversion result is determined to be valid; otherwise, it is determined to be invalid or needs to be removed, thus forming a supporting quality control mechanism.

[0118] Example 3 provides a power plant carbon dioxide emission quantification system based on satellite observation data. In addition to Example 2, it also includes a processor and a memory. The memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the power plant carbon dioxide emission quantification method based on satellite observation data in Example 1.

[0119] Example 4 provides a power plant carbon dioxide emission quantification system based on satellite observation data. In addition to Example 2, it also includes a readable storage medium on which a computer program is stored. When the computer program is executed, it implements the power plant carbon dioxide emission quantification method based on satellite observation data in Example 1.

[0120] The above description is merely one embodiment of the present invention, and while it is detailed and specific, it should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A method for quantifying carbon dioxide emissions from power plants based on satellite observation data, characterized in that, The specific steps are as follows: S1. Extract satellite observation XCO2 data within a defined spatial range around the target power plant; select observation XCO2 data that meet the inversion conditions based on constraints such as the number of effective detection footprints, the number of significantly enhanced footprints, and the consistency between the enhancement direction and the wind direction; S2. Robust local regression is used to obtain the optimal estimate of background XCO2 from the observed XCO2 data selected in step S1, thereby obtaining the background baseline. The background baseline is then subtracted from the observed XCO2 data selected in step S1 to obtain robust observation-enhanced XCO2 data. S3. Input the observation-enhanced XCO2 data extracted in step 2 into the emission inversion model, form reasonable boundary values ​​for unknown parameters and provide robust initial values ​​through Monte Carlo simulation, and use the trust region optimization algorithm to solve for emission intensity and diffusion-related parameters to realize emission rate inversion at the time of power plant passage. S4. Iterate through step S3 at least 1000 times, outputting a set of estimated emission intensity values, evaluate the standard deviation and mean of the set, and if the proportion is less than 25%, then the transit inversion result is deemed valid, thus forming a supporting quality control mechanism.

2. The method for quantifying carbon dioxide emissions from power plants based on satellite observation data according to claim 1, characterized in that: In step S1, passive satellite overpass observation XCO2 data within a 60-kilometer radius around the emission source of the target power plant are selected. The number of effective detection footprints within this range should be greater than 150. The effective observation XCO2 data is extracted using QAflag or average kernel. The observation-enhanced XCO2 data is extracted from the effective observation XCO2 data using standard Gaussian distribution. Combined with wind direction discrimination, the observation-enhanced XCO2 data that meet the inversion conditions are selected. The specific method for extracting valid XCO2 observation data using QAflag is as follows: if QAflag is 0, it is considered a valid detection; if it is 1, it is considered an invalid detection. This QAflag is obtained based on the passive satellite concentration inversion average kernel AK. If the orbital data does not provide QAflag, the average kernel is used for judgment. If AK > 0.7, then QAflag = 0 is assigned.

3. A method for quantifying carbon dioxide emissions from power plants based on satellite observation data according to claim 1 or 2, characterized in that: In step S2, each observation location is obtained by minimizing the loss function Loss. Background concentration at The optimal estimate is obtained by averaging the optimal background concentration across all locations. Calculate the baseline CO2 background concentration near the monitored emission source according to formula ①. Then, calculate the enhancement concentration at each location according to formula ②. ; ① ； ②； Each observation location Background concentration at The optimal estimate is calculated using the following formula: ③； ④ ； In the above formula, Y(x i ) indicates that in x i The CO2 concentration observed at a specific location, i.e., XCO2 observed by the satellite at that location; b(x i ) is x i Background XCO2 concentration at the location; s(x i The increase in CO2 concentration is due to CO2 emissions from power plants; e i The instrument measurement error is represented by n; n is the number of XCO2 observations. It characterizes the proportion of XCO2 enhancement caused by CO2 emissions from power plants in the measurements involved in the regression; express Measurement error at the location; Through the robust extraction described above, the observation data from satellite remote sensing are... Distinguished as background concentration baseline Enhanced concentration at each observation point Subsequently, it participated in further emissions inversion.

4. A method for quantifying carbon dioxide emissions from power plants based on satellite observation data according to claim 1 or 2, characterized in that: In step S3, concentration inversion is performed by matching the simulated XCO2 concentration with the observed XCO2 concentration. The CO2 distribution characteristics downwind of the emission source are assumed to be a vertically integrated Gaussian plume distribution, then: ⑤ ； ⑥ ； In the formula, The CO2 column concentration simulated by the Gaussian plume model. To enhance the CO2 column concentration in the Gaussian plume model simulation. The background baseline for CO2 column concentration simulated by the Gaussian plume model; M air M CO2 These represent the molar masses of atmospheric molecules and CO2 molecules, respectively; g is the acceleration due to gravity, in m / s². 2 ;P surf ρ is the surface pressure; w is the atmospheric water vapor column concentration, kg / m³. 2 Q represents carbon emission intensity, g / s; u represents wind speed, m / s; x0 represents 1000 meters; a and b represent σ. y The diffusion coefficient of (x); Based on the trust region optimization algorithm, the unknown parameters (Q,u,a,b,...) involved in the Gaussian plume model are... To achieve an exact solution, no further assumptions are introduced; the parameters to be solved are unified to the parameter domain vector x = (Q, u, a, b, ... Monte Carlo simulations are employed, and this process is repeated at least 10,000 times to construct a robust solution space for the unknown parameters. During these iterations, the boundaries of the solution space are obtained through Monte Carlo simulations to determine the maximum and minimum values ​​of each unknown parameter. The average value of the corresponding unknown parameters is used as the initial input x for the confidence region algorithm. k In the current initial input x k At this point, a subproblem is constructed with the least squares function as the objective, as shown in the following formula: ⑦， It is an approximation of the Hessian matrix of the objective function f. This represents the gradient of the objective function; If the change in the parameter vector to be solved is represented, then the search domain is defined as: ⑧ ， ⑨； In the formula, α represents the step size; The formula for the confidence region algorithm is as follows: ⑩； In the formula, f(x) is the fitness function, used to adapt the simulated XCO2 concentration to the observed XCO2 concentration; These are equality constraints and inequality constraints; It is the set of unknown parameters (Q,u,a,b, ...) to be solved in the Gaussian plume model. ; In this process This represents the maximum and minimum values ​​of the unknown parameters corresponding to 10,000 solutions in a genetic algorithm (GA). when satisfy Error limit within the allowable range These are the unknown parameters of the final result; The fitness function f is defined as follows: ⑪; Where each k represents a single observation point in a satellite transit strip, k = (x, y); m represents the total number of valid observation points in the observation strip, obtained from step S1; Its parameters include the emission inversion value Q required by the present invention, which represents the power plant emission measurement results at the moment of satellite transit.

5. A method for quantifying carbon dioxide emissions from power plants based on satellite observation data according to claim 1 or 2, characterized in that: In step S4, Monte Carlo simulation is used to perform 1,000 inversions on step S3, and then the standard deviation of the emission value Q is calculated. Calculation as follows ⑫; in, The carbon dioxide emission value obtained from the i-th inversion; The average emissions value from 1000 inversions; N=1000; if If the emissions are less than 25%, the emissions inversion results are valid.

6. The method for quantifying carbon dioxide emissions from power plants based on satellite observation data according to claim 2, characterized in that: In step S1, the effective observed XCO2 data are fitted using a standard Gaussian distribution function to obtain the approximate maximum position of the observed XCO2. The azimuth angle W1 of the approximate maximum XCO2 value relative to the emission source is then obtained using the angle between the vector connecting this position and the emission source location and the unit north vector in the geodetic coordinate system. The average wind direction of ERA 5 obtained at different altitudes from 10 to 400 m is calculated as W2. Effective XCO2 data satisfying |w1-w2| less than 60° constitutes a transit observed XCO2 sequence that meets the inversion conditions. The standard Gaussian distribution function is: ⑬， Where μ is the mean of the standard Gaussian distribution of the effective XCO2 data; The standard deviation of the valid XCO2 data.

7. The method for quantifying carbon dioxide emissions from power plants based on satellite observation data according to claim 3, characterized in that: For the calculation in step S2 and According to formula 14, the concentration of CO2 in a vertically integrated column can be converted. , Unit: g / m 2 The specific formula (14) is as follows: ⑭ ； In step S2 The specific calculation process is as follows: ⑮ ； In the formula, Obtained by global least squares regression estimation, i.e., all All are involved in the calculation; express The measurement error at that point can be determined through the measurement error information in the satellite's official documents.

8. A power plant carbon dioxide emission quantification system based on satellite observation data, characterized in that: The system is used to execute the power plant carbon dioxide emission quantification method based on satellite observation data according to any one of claims 1 to 7; the system specifically includes a data acquisition and screening module, an enhanced concentration data extraction module, an emission inversion module, and an uncertainty analysis module; The data acquisition and filtering module is used to extract satellite observation XCO2 data within a set spatial range around the target power plant; and to filter observation XCO2 data that meet the inversion conditions based on the number of effective detection footprints, the number of significantly enhanced footprints, and the consistency between the enhancement direction and the wind direction. The enhanced concentration data extraction module is used to obtain the background baseline of the satellite observation samples, and further introduces a Gaussian mixture model to perform unsupervised clustering on the observation data, distinguishing between background observations less affected by emissions and enhanced observations more affected by emissions, so as to obtain more robust observation-enhanced XCO2 data. The emission inversion module is used to receive the observed enhanced XCO2 data output by the enhanced concentration data extraction module, form reasonable boundary values ​​for unknown parameters through Monte Carlo simulation and provide robust initial values, and use the trust region optimization algorithm to solve for emission intensity and diffusion-related parameters, so as to realize the emission rate inversion at the time of power plant passage. The uncertainty analysis module enhances the correlation index between the observed enhanced XCO2 data in the concentration data extraction module and the enhanced data simulated by the emission inversion module. When the correlation meets the threshold condition, the transit inversion result is determined to be valid; otherwise, it is determined to be invalid or needs to be removed, thus forming a supporting quality control mechanism.

9. A power plant carbon dioxide emission quantification system based on satellite observation data according to claim 8, characterized in that: The system further includes a processor and a memory, the memory being used to store program instructions, and the processor being used to call the stored instructions in the memory to execute the power plant carbon dioxide emission quantification method based on satellite observation data as described in any one of claims 1-7.

10. A power plant carbon dioxide emission quantification system based on satellite observation data according to claim 8, characterized in that: The system further includes a readable storage medium storing a computer program, which, when executed, implements a method for quantifying carbon dioxide emissions from power plants based on satellite observation data as described in any one of claims 1-7.

Images