A method and system for inversion of carbon emission intensity and tracing based on numerical simulation and reduced order model

Abstract

The application discloses a kind of based on numerical simulation and inversion carbon emission intensity and tracing back to source method and system of model, the method includes: constructing full-size three-dimensional map;According to map information and known environmental conditions, numerical simulation of pollutant diffusion;By intrinsic orthogonal decomposition method, the simulation data is reduced order;Based on reduced order model, carbon emission simulation database under the condition of multi-working condition variable environment is established;Real-time field measurement of carbon dioxide concentration and meteorological data, as a supplement to simulation database, the intensity of known emission source, the intensity and position of unknown emission source are calculated using source term estimation method, particle swarm algorithm and the like.The application can realize full-size reduction of limited area and high-precision small-scale carbon emission monitoring, and can effectively trace back to source, analyze intensity and predict potential emission source according to a small amount of field measured data.

Description

TECHNICAL FIELD

[0001] The present application relates to environmental monitoring, in particular to a method and system for inverting carbon emission intensity and tracing based on numerical simulation and reduced order model. BACKGROUND

[0002] With the rapid development of laser radar technology, it has become a reality to quickly generate high-precision maps. Such maps play an important role in urban planning, land management, resource exploration, disaster response, and other fields. By equipping unmanned mobile devices such as drones, remote control cars, and robot dogs with laser radars, it is possible to quickly and accurately measure features such as ground topography, buildings, and vegetation. In addition, small and high-precision pollution detection sensors have made great progress and are widely used.

[0003] The development of the above technology makes it possible to sample dynamic real-time carbon emission data based on fixed and mobile sensors and then predict the carbon emission intensity and location of strong point sources. This method combines sensor technology, data analysis, and spatial positioning systems to quickly obtain large-scale pollutant concentration data and accurately identify and locate emission sources through mathematical models and algorithms. However, this "bottom-up" inversion method still faces many challenges. The diffusion process of atmospheric pollutants often has strong dynamics and randomness, and pollution tracing requires a large amount of high-quality monitoring data. For this complex process, real-time data obtained from fixed and mobile sensors has a high degree of uncertainty and insufficiency. Existing tracing theories and methods are based on sufficient sensor distribution and ideal measurement environment, and the sparse, insufficient, and variable data in the actual process and the complex atmospheric environment pose a great challenge to the inversion and tracing of carbon emission sources. SUMMARY

[0004] The purpose of the present application is to provide a method and system for inverting carbon emission intensity and tracing based on numerical simulation and reduced order model with high accuracy and low dependence on a large amount of high-quality measured data, which can improve the quality and efficiency of sampling data, perform more accurate inversion process, and realize the intensity prediction and location tracing of carbon emission point sources.

[0005] Technical solution: The method for inverting carbon emission intensity and tracing based on numerical simulation and reduced order model according to the present application comprises:

[0006] (1) Construct a virtual three-dimensional map of the detected limited area by SLAM algorithm, collect carbon emission concentration data and meteorological data of the detected limited area;

[0007] (2) Based on the virtual three-dimensional map, combined with the meteorological data of the detected limited area and the intensity of the known emission source, the computational fluid dynamics numerical simulation is carried out to predict the velocity field, airflow field, carbon concentration value and distribution in the space, and a space carbon emission simulation database is generated; the data simulation results are used to guide the position, range and path of data collection;

[0008] (3) The most representative mode is extracted from the simulation database by the intrinsic orthogonal decomposition method, and the main dynamic characteristics are retained to obtain a reduced order model;

[0009] (4) The bionic algorithm is used to determine the patrol monitoring sampling path and collect carbon dioxide concentration data and meteorological data in real time;

[0010] (5) Based on the measured real-time carbon dioxide concentration data and meteorological data, combined with the simulation database supplementary data under the corresponding meteorological conditions, the intensity prediction and location tracing of the carbon emission point source of the traceable atmospheric pollution tracing problem are carried out.

[0011] Further, in step (1), the mobile device carries a laser radar, an inertial measurement module and a camera to obtain accurate distance measurement values and environmental images from the surrounding environment, carries a greenhouse gas monitoring module and a meteorological monitoring module to measure the carbon emission concentration data and meteorological data of the detected limited area; the real-time position of the mobile device and the structure information of the surrounding environment are fused by the SLAM algorithm to realize the modeling and positioning of the full-size scene.

[0012] Further, the greenhouse gas monitoring module adopts a small non-dispersive infrared carbon dioxide monitoring module, and the meteorological monitoring module adopts a small weather station.

[0013] Further, the mobile device includes a drone, a remote control car and a robot dog.

[0014] Further, in step (1), the CFD numerical simulation adopts a fixed steady-state point carbon emission source diffusion model, which is represented as follows:

[0015]

[0016] Wherein, c is the concentration of atmospheric pollutants measured by the sensor; (x, y, z) represents the position of the sensor in the three-dimensional coordinate system; q is the mass of pollutants generated by the emission source per unit time, i.e. the emission rate; u is the wind speed; σ y 、σ z , and σ

[0017] The atmospheric diffusion model adopts a convection diffusion model:

[0018]

[0019] in, This indicates the divergence of a vector field; K is the velocity vector; c S represents the diffusion coefficient matrix, which consists of the diffusion coefficients along the x, y, and z axes; c This indicates the emission rate of air pollutants.

[0020] Furthermore, step (3) includes:

[0021] Taking the velocity distribution in a velocity field as an example, the data of all grid points in the velocity field of a certain cross section are extracted and arranged in a specific sequence to represent a matrix X:

[0022] X = [x1, x2, ..., x n ]

[0023] Where, x i is the state vector at the i-th time point, and n is the number of time points;

[0024] Calculate the covariance matrix C of matrix X:

[0025] C = XX T / n

[0026] Among them, X T It is the transpose of X;

[0027] Perform eigenvalue decomposition on the covariance matrix C to obtain the eigenvalues ​​λ. i and the corresponding eigenvector φ i :

[0028] Cφ i =λ i φ i

[0029] Projecting the original data onto these feature vectors yields the modal coefficients a. i (t):

[0030]

[0031] Where x(t) is a data vector at a certain moment;

[0032] The original data can be reconstructed using a few main modes and their corresponding mode coefficients:

[0033]

[0034] Where r is the number of modes selected;

[0035] After the reduction, the complex velocity field containing a large amount of data can be represented by a few main modes and corresponding modal coefficients, and the temperature field, the pollutant concentration distribution field are the same.

[0036] Further, step (4) comprises: detecting the concentration change difference of the two positions to make a direction estimation;

[0037] Define C as the current concentration, C max is the maximum concentration from the initial to the nth position; V is the concentration change rate; V min is the set minimum change rate, V n is the concentration change rate of the nth step, define:

[0038] V n = (C n -C n-1 ) / C n-i

[0039] n is the nth sample to start searching; a is the angle with the current forward direction of the mobile device, d min is the set minimum step size, d is the current step size, k a , k d is the change factor related to V, which determines the change size of a and d, according to the dynamic stimulus response:

[0040] a n =k a a n-1 , d n =k d d n-1

[0041] When the repetition condition is not met, and C n changes very small relative to C max is the odor source, and ε is the absolute value of the difference between the two minimum values; the search strategy is as follows: first, the sensor moves randomly to search, with a given step size, at any angle, and advances without detecting gas, the step size remains unchanged, and randomly turns left or right by an angle a, until gas is detected, at which time the last two steps of the mobile device are determined according to the gas concentration change rate to determine whether to turn left or right, and the step size is changed; when the sensor data collection meets the stop condition, it is determined that the place is the gas source;

[0042] The step size and turning angle of the mobile device depend on the gas concentration change rate, and the execution steps are as follows:

[0043] First, determine whether the angle needs to be changed, the current step size is less than the set minimum step size, i.e. d n <d min ; or the current concentration change rate is less than the set minimum change rate, i.e. V n <V min; or the absolute value of the current concentration difference is less than the minimum value, that is, |C n -C max | < epsilon; the above three cases are determined not to convert the angle; the current step is greater than or equal to the set minimum step, the current concentration change rate is greater than or equal to the set minimum change rate, and the absolute value of the current concentration difference is greater than or equal to the minimum value, that is, d n >= d min , V n >= V min , and |C n >= C max , it is determined that the angle needs to be converted;

[0044] Then determine the conversion direction, if the gas concentration shows a decreasing trend, that is, V n < V n-1 , it is determined that the left turn is made at the n position, the turning angle is a n , and the step is d n ; if the gas concentration shows an increasing trend, that is, V n >= V n-1 , it is determined that the right turn is made at the n position, the turning angle is a n , and the step is d n .

[0045] Further, in step (5), the traceable atmospheric pollution traceability problem determination method is: assuming that the sensor measurement point data quantity is m, the number of emission sources is n, m < n; the coefficient matrix A = m*n is a rectangular matrix, and the matrix equation is solved:

[0046] A T c = A T Aq

[0047] Wherein, c is an array composed of m measurement values; q is an array composed of n emission source intensities;

[0048] When the above matrix equation has a unique solution, the atmospheric pollution traceability problem can be traced; if the above matrix equation is not only a unique solution, the sensor measurement point data m should be appropriately increased until the source parameter can be estimated.

[0049] Further, in step (5), the intensity prediction and position traceability of the carbon emission point source are performed through the source term estimation method, the particle swarm algorithm of the nonlinear traceability method, the Bayesian algorithm, the optimization algorithm or the improved Gaussian diffusion inversion model.

[0050] The system for inverting carbon emission intensity and traceability based on numerical simulation and reduced order model provided by the application comprises:

[0051] A data acquisition and storage module for recording and storing sampling data, including sampling path information, sampling point spatial position information, distance measurement data, environmental images, CO2 concentration, and meteorological data;

[0052] A data calculation module for processing the sampling data, establishing a high-precision three-dimensional digital map, establishing a simulation model, verifying simulation results, POD model reduction of the detected environmental concentration field, airflow field, and temperature field, establishing a simulation database, traceability judgment, planning a patrol monitoring sampling path, emission source intensity analysis and tracing;

[0053] A data visualization module for data statistics and analysis, with the data presented in the form of cloud maps and statistical charts.

[0054] Advantages: Compared with the prior art, the present application has the following significant advantages:

[0055] The present application examines the influence of meteorological factors on atmospheric pollutant diffusion through numerical simulation and real-time field sampling data verification. By adjusting the simulation parameters, it predicts carbon emissions under different working conditions and different meteorological factors, with high flexibility and wide coverage. The database based on real measurement and simulation further guides the sampling methods and methods (such as the position of fixed sampling sensors and the path of mobile sampling sensors), which is beneficial to improve the sampling efficiency and data quality, and thus reduces the dependence of general inversion algorithms on a large amount of high-quality data. Through the intrinsic orthogonal decomposition method, the simulation data is reduced, and the relationship between temperature field, velocity field, pollutant concentration distribution, and meteorological factors, carbon emission source intensity, and position is proposed, and a carbon emission database covering different meteorological conditions and different carbon emission point source intensities and positions is constructed. A small amount of measured data can be compared with the database to predict the intensity and position of the carbon emission point source. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is a flow chart of the method for inverting carbon emission intensity and tracing based on numerical simulation and reduced-order model provided by an embodiment of the present application;

[0057] Figure 2 is a three-dimensional digital map of the detected limited area in an embodiment of the present application;

[0058] Figure 3 is a numerical simulation result in an embodiment of the present application, wherein Figure 3 (a) is a carbon dioxide concentration distribution cloud map, Figure 3 (b) is a carbon dioxide flow field cloud map. DETAILED DESCRIPTION

[0059] The present application will be further described below with reference to the accompanying drawings.

[0060] The embodiment of the present application provides a method for inverting carbon emission intensity and tracing source based on numerical simulation and reduced order model, which is based on reliable simulation database and effective low-order model, and can realize fast, convenient and real-time monitoring and prediction of environmental carbon concentration, carbon distribution, carbon source intensity prediction and positioning on the basis of ensuring accuracy. The method is a full-size reduction method for measuring the measured area, high-precision, small-scale actual measurement of carbon content and effective emission source tracing and evaluation, as shown in Figure 1 The method specifically comprises the following steps:

[0061] (1) In the measured limited area, a mobile device carrying a laser radar, an inertial measurement module, a camera and other sensors acquires accurate distance measurement values and environmental images from the surrounding environment. The distance measurement data collected by the laser radar and the environmental images acquired by the camera are transmitted to a ground processing station. The mobile device can be a drone, a remote control vehicle, a robot dog, etc. In the ground processing station, the acquired data is processed by a simultaneous localization and mapping method (SLAM algorithm) to generate a three-dimensional high-precision map, as shown in Figure 2 The SLAM algorithm can fuse the real-time position of the mobile device and the structural information of the surrounding environment to realize modeling and positioning of the full-size scene.

[0062] While scanning the terrain, the greenhouse gas monitoring module and the weather monitoring module carried by the mobile device also measure the carbon emission concentration data and the weather data of the detected limited area. The weather data includes wind direction, wind speed, air pressure, temperature, humidity, etc. The measured data is not only used as a reference for setting the boundary conditions in the initial simulation in step (2), but also used for verifying the simulation results. The carbon emission concentration data and the weather data acquired by the monitoring module are saved and transmitted to the ground processing station. Since the mobile device itself has a power source and a moving speed, for example, the propeller of a drone, in order to reduce the influence of the above interference factors on the weather data, the weather monitoring module can also be selected in a fixed mode. In this embodiment, the greenhouse gas monitoring module uses a small non-dispersive infrared (NDIR) carbon dioxide sensor, and the weather monitoring module uses a small weather station.

[0063] (2) Based on the virtual model shown in Figure 2 , combined with the weather data of the detected limited area and the intensity of the known emission source, a computational fluid dynamics (CFD) numerical simulation is performed to predict the velocity field, airflow field, carbon concentration value and distribution in space, and generate a spatial carbon emission simulation database. The boundary conditions of the numerical simulation include the intensity of the fixed emission source and the wind speed, temperature, etc. By comparing the wind speed and carbon dioxide concentration data of the same position observation point, if the relative error is <10%, the reliability of the model can be verified.

[0064] The result of data simulation is shown in Figure 3 , wherein Figure 3(a) shows the carbon dioxide concentration distribution cloud map, Figure 3 (b) shows the air flow field distribution map. According to Figure 3 (a) can know the approximate distribution range and high concentration area of pollutants, further guiding the range and position distribution of carbon dioxide sensor deployment. According to Figure 3 (b) can know the influence law of meteorological factors on the diffusion of atmospheric pollutants in the observed area, further guiding the moving sampling path of carbon dioxide sensor and the position deployment of small weather station.

[0065] Specifically, the CFD numerical simulation adopts a fixed steady point carbon emission source diffusion model (Gaussian plume model), which is expressed as follows:

[0066]

[0067] Where c is the measured atmospheric pollutant concentration of the sensor; (x, y, z) represents the position of the sensor in the three-dimensional coordinate system; q is the mass of pollutants generated by the emission source per unit time, i.e. the emission rate; u is the wind speed; σ y , σ z are the standard deviations of the pollutant concentration along the y-axis and z-axis directions, respectively;

[0068] The atmospheric diffusion model adopts a convection-diffusion model:

[0069]

[0070] Where, represents the divergence of the vector field; is the velocity vector; K c represents the diffusion coefficient matrix composed of diffusion coefficients in x, y, and z directions; S c represents the emission rate of atmospheric pollutants.

[0071] (3) The most representative mode is extracted from the simulation database by the proper orthogonal decomposition method, and the main dynamic characteristics are retained to obtain the reduced order model;

[0072] Specifically, the model reduction technology based on the proper orthogonal decomposition method (POD) is combined with CFD simulation to model the detected scene, and the simulation data obtained, including velocity field, temperature field, and pollutant concentration field, are processed by dimension reduction. The POD (Proper Orthogonal Decomposition) method is a commonly used data dimension reduction technology, which extracts the main feature mode in the data and converts high-dimensional data into a representation in low-dimensional space, thereby simplifying calculation and analysis.

[0073] The result data of numerical simulation is embodied in small grids constituting its three-dimensional structure, such as the velocity distribution in the velocity field.

[0074] Extract the data in all grid points of a velocity field of a certain section and arrange them in a certain sequence to represent a matrix X:

[0075] X = [x1, x2,..., xn] n ]

[0076] where xi is the state vector at the i-th time point, and n is the number of time points. i

[0077] Calculate the covariance matrix C of matrix X:

[0078] C = XX T / n

[0079] where X T is the transpose matrix of X.

[0080] Perform eigenvalue decomposition on the covariance matrix C to obtain the eigenvalues λ i and the corresponding eigenvectors φ i :

[0081] Cφ i = λ i φ i

[0082] The eigenvectors φ i are called POD modes.

[0083] Project the original data onto these eigenvectors to obtain the modal coefficients a i (t):

[0084]

[0085] where x(t) is the data vector at a certain time.

[0086] Using a few major modes and the corresponding modal coefficients, the original data can be reconstructed:

[0087]

[0088] where r is the number of selected modes, and usually the first few major modes are selected.

[0089] ​After the above method is reduced, the complex velocity field containing a large amount of data can be represented by a few main modes and corresponding modal coefficients, and the temperature field, the pollutant concentration distribution field is the same. The modal coefficient depends on the boundary condition in the numerical simulation, that is, the pollutant emission source intensity and meteorological data in reality. Based on the above-mentioned low-order model, the carbon emission database covering carbon emission point sources of different intensities and positions and different meteorological conditions can be quickly and conveniently obtained. Because in the actual measurement process, the sensor cannot completely cover the detected area, the above-mentioned carbon emission database can be used as a supplement to the measured data, thereby reducing the dependence of the carbon emission point source intensity prediction and source tracing inversion algorithm on a large amount of high-quality data.

[0090] (4) For the inversion of carbon emission point source intensity and location tracing, only a small amount of carbon dioxide concentration data and meteorological data at a few typical positions need to be collected. For the fixed position deployment of carbon dioxide sensors and meteorological sensors, reference can be made to step (2). The patrol monitoring sampling path of the movable carbon dioxide sensor and meteorological sensor mounted on a drone or a remote control vehicle needs to use a bionic algorithm. The chemotactic mechanism is to compare two or more sensors at the same time to produce an instantaneous estimate of the same or opposite direction of the concentration gradient, and to estimate the direction according to the bionic principle of detecting the concentration change difference between two positions. This bionic odor source positioning algorithm does not rely on the absolute concentration of the sensor, can improve the search efficiency of the mobile device, and can not depend on the wind direction.

[0091] Define C as the current concentration, C max is the maximum concentration from the initial to the nth position; V is the concentration change rate; V min is the set minimum change rate, V n is the concentration change rate of the nth step, and is defined as:

[0092] V n = (C n -C n-1 ) / C n-i

[0093] n is the nth sample to start searching. a is the angle with the current forward direction of the mobile device, d min is the set minimum step size, d is the current step size, k a , k d is a change factor related to V, which determines the change size of a and d, according to the dynamic stimulus response:

[0094] a n = k a a n-1 , d n = k d d n-1

[0095] When the repetition condition is not met, and Cn Relative C max The change is very small, that is, the source of the taste, and ε is the absolute value of the difference between the two minimum values. The search strategy is as follows: first, the sensor moves to search randomly, and advances at a given step length, at any angle, without detecting gas, the step length remains unchanged, and is randomly turned left or right by an angle a until the gas is detected. At this time, according to the gas concentration change rate of the last two steps of the mobile device, it is determined whether to turn left or right, and the step length is changed; when the sensor collects data to meet the stopping condition, it is determined that the place is the gas source.

[0096] The step length and turning angle of the mobile device depend on the gas concentration change rate, and the execution steps are as follows:

[0097] First, it is determined whether the angle needs to be converted. The current step length is less than the set minimum step length, that is, d n <d min ; or the current concentration change rate is less than the set minimum change rate, that is, V n <V min ; or the absolute value of the current concentration difference is less than the minimum value, that is, |C n -C max | < ε. The above three cases are determined not to change the angle. The current step length is greater than or equal to the set minimum step length, the current concentration change rate is greater than or equal to the set minimum change rate, and the absolute value of the current concentration difference is greater than or equal to the minimum value, that is, d n ≥ d min , V n ≥ V min , and |C n -C max | ≥ ε, it is determined that the angle needs to be converted.

[0098] Then determine the conversion direction. If the gas concentration shows a decreasing trend, that is, V n <V n-1 , it is determined to turn left at the n position, the turning angle is a n , and the step length is d n ; if the gas concentration shows an increasing trend, that is, V n ≥ V n-1 , it is determined to turn right at the n position, the turning angle is a n , and the step length is d n .

[0099] (5) Further, based on the measured real-time carbon dioxide concentration data and meteorological data, combined with the simulation database supplementary data under the corresponding meteorological conditions, the intensity of the carbon emission point source is inverted and the location is traced.

[0100] First, the traceability is judged: the traceability of the atmospheric pollution traceability problem is discussed, that is, whether the source parameters can be estimated:

[0101] Suppose the sensor measurement data is m, the number of emission sources is n, m < n, the coefficient matrix A = m*n is a rectangular matrix, and the matrix equation is solved:

[0102] A T c = A T Aq

[0103] Wherein, c is an array of m measurement values; q is an array of n emission source strengths;

[0104] When the above matrix equation has a unique solution, the atmospheric pollution source tracing problem can be traced. If the above matrix equation is not only a unique solution, the amount of sensor measurement data m should be appropriately increased until the source parameters can be estimated.

[0105] Further, for the traceable problem, based on the above measured data and supplementary data from the database, the intensity prediction and location tracing of the carbon emission point source are carried out by using methods including but not limited to source term estimation method, nonlinear source tracing method particle swarm algorithm, Bayesian algorithm, optimization algorithm, improved Gaussian diffusion inversion model, etc.

[0106] The embodiment of the present application also provides a system for inversion of carbon emission intensity and source tracing based on numerical simulation and reduced order model, comprising:

[0107] A data acquisition and storage module for recording and storing sampling data, the sampling data including sampling path information, sampling point spatial position information, distance measurement data, environmental image, CO2 concentration, meteorological data, etc.

[0108] A data calculation module for processing the sampling data, establishing a high-precision three-dimensional digital map, establishing a simulation model, verifying the simulation results, POD model reduction of the detected environmental concentration field, airflow field and temperature field, establishing a simulation database, traceability judgment, planning a patrol monitoring sampling path, emission source intensity analysis and source tracing, etc.

[0109] A data visualization module for data statistics and analysis, the data being presented in the form of cloud charts, statistical charts, etc.

[0110] In summary, the method for inversion of carbon emission intensity and source tracing based on numerical simulation and reduced order model provided by the present application has the advantages of convenience, real-time and high efficiency. The method combines sensor technology, model reduction of proper orthogonal decomposition (POD), data analysis and spatial positioning system, can conveniently obtain pollutant data under different point source intensities and different meteorological conditions, including airflow velocity distribution, temperature distribution, concentration distribution, etc., and can perform real-time and accurate intensity inversion identification and location tracing of the emission source through existing mathematical models and algorithms. The method proposes simulation data as a supplement to the measured data, which can calculate the overall carbon content, fixed carbon emission source intensity and potential emission source tracing using a small amount of measured data.

Claims

1. A method for inversion of carbon intensity and tracing based on numerical simulation and reduced order model, characterized in that, The method comprises the following steps: (1) constructing a virtual three-dimensional map of the detected limited area by a SLAM algorithm, collecting carbon emission concentration data and meteorological data of the detected limited area; (2) based on the virtual three-dimensional map, combining the meteorological data of the detected limited area and the intensity of the known emission source, performing a computational fluid dynamics numerical simulation to predict the velocity field, airflow field, carbon concentration value and distribution in the space, and generating a spatial carbon emission simulation database; using the data simulation results to guide the position, range and path of data collection; (3) extracting the most representative modes from the simulation database by an intrinsic orthogonal decomposition method, and retaining the main dynamic characteristics to obtain a reduced-order model; (4) determining a patrol monitoring sampling path based on a bionic algorithm and collecting carbon dioxide concentration data and meteorological data in real time; (5) based on the measured real-time carbon dioxide concentration data and meteorological data, combining the simulation database supplementary data under the corresponding meteorological conditions, predicting the intensity and location of the carbon emission point source for the traceable atmospheric pollution traceability problem; In step (1), a mobile device carries a laser radar, an inertial measurement module and a camera to obtain accurate distance measurement values and environment images from the surrounding environment, carries a greenhouse gas monitoring module and a meteorological monitoring module to measure the carbon emission concentration data and meteorological data of the detected limited area, and fuses the real-time position of the mobile device and the structure information of the surrounding environment by a SLAM algorithm to realize modeling and positioning of a full-size scene; In step (1), the CFD numerical simulation adopts a fixed steady-state point carbon emission source diffusion model, which is expressed as follows: where c is the measured concentration of atmospheric pollutants by the sensor; (x, y, z) represents the position of the sensor in the three-dimensional coordinate system; q is the mass of pollutants generated by the emission source per unit time, i.e., the emission rate; u is the wind speed; σ y , σ z are the standard deviations of the pollutant concentration along the y-axis and z-axis directions, respectively. The atmospheric diffusion model adopts a convection-diffusion model: wherein, denotes the divergence of a vector field; is the velocity vector; K c denotes a diffusion coefficient matrix consisting of diffusion coefficients in each of the x, y, z directions; S c denotes the emission rate of atmospheric pollutants; Step (3) comprises: Taking the velocity distribution in the velocity field as an example, the data in all grid points of a certain cross section of the velocity field are extracted and arranged in a certain sequence to represent a matrix X: X = [x1, x2,..., x i ,...,x N ] where x i is the state vector at the i-th time point, and N is the number of time points. The covariance matrix C of the matrix X is calculated: C=XX T / N where X T is the transpose of X; Eigenvalue decomposition of the covariance matrix C yields the eigenvalues λ i and the corresponding eigenvectors φ i : Cφ i = λ i φ i Projecting the raw data onto these feature vectors gives the modal coefficients a i (t): Wherein, x(t) is a data vector at a certain time; The original data are reconstructed by using a few main modes and corresponding modal coefficients: Wherein, r is the number of selected modes; After reduction, the complex velocity field containing a large amount of data is represented by a few main modes and corresponding modal coefficients, and the temperature field and the pollutant concentration distribution field are the same; Step (4) comprises: detecting the concentration change difference of two positions to estimate the direction; C is the current concentration, C max is the maximum concentration from the initial to the nth position; V is the rate of change of concentration; V min is the set minimum rate of change, V n is the rate of change of concentration at the nth position, defined as: V n = (C n -C n-1 ) / C n-i a is the angle with the current direction of the mobile device, d min is the minimum step size, d is the current step size, k a , k d is the change factor related to V, which determines the change size of a and d, according to the dynamic stimulation response: a n = k a a n-1 d n = k d d n-1 When the repetition condition is not met, and C n Relative to C max When the change is very small, that is, the source of the smell, and ε is the absolute value of the difference between the two minimum values; the search strategy is as follows: first, the sensor moves randomly to search, with a given step size, at any angle, and advances without detecting gas. The step size remains unchanged, and the sensor randomly turns left or right by an angle a until gas is detected. At this time, the last two steps of the mobile device are used to determine whether to turn left or right according to the gas concentration change rate, and the step size is changed. When the sensor collects data that meets the stopping condition, it is determined that the sensor is the source of the gas. The step length and turning angle of the mobile device depend on the gas concentration change rate, and the steps are performed as follows: Firstly, it is judged whether the angle needs to be converted. The current step is less than the set minimum step, i.e. d n <d min ; or the current concentration change rate is less than the set minimum change rate, i.e. V n <V min ; or the absolute value of the current concentration difference is less than the minimum value, i.e. │C n -C max │<ε; the above three cases are determined not to need to convert the angle; the current step is greater than or equal to the set minimum step, the current concentration change rate is greater than or equal to the set minimum change rate, and the absolute value of the current concentration difference is greater than or equal to the minimum value, i.e. d n ≥d min , V n ≥V min , and │C n -C max │≥ε, it is determined that the angle needs to be converted; Then the turning direction is judged. If the gas concentration presents a decreasing trend, i.e. V n < V n-1 , it is determined that a left turn is made at the n position with a turning angle of a n and the vehicle advances d n ; if the gas concentration presents an increasing trend, i.e. V n ≥ V n-1 , it is determined that a right turn is made at the n position with a turning angle of a n and the vehicle advances d n . In step (5), the traceable atmospheric pollution traceability problem is determined as follows: assuming that the sensor measurement point data amount is m, the number of emission sources is L, m < L; the coefficient matrix A = m*L is a rectangular matrix, and the matrix equation is solved as follows: A T c = A T Aq Wherein, c is an array composed of m measurement values; q is an array composed of L emission source intensities; When the above matrix equation has a unique solution, the atmospheric pollution traceability problem is traceable; if the above matrix equation does not have only a unique solution, the sensor measurement point data m should be appropriately increased until the source parameters can be estimated.

2. The method for inversion of carbon intensity and attribution based on numerical simulation and reduced order model according to claim 1, characterized in that, The greenhouse gas monitoring module adopts a small non-dispersive infrared carbon dioxide monitoring module, and the meteorological monitoring module adopts a small weather station.

3. The method for inversion of carbon intensity and attribution based on numerical simulation and reduced order model according to claim 1, characterized in that, The mobile device comprises a drone, a remote control vehicle and a robot dog.

4. The method for inversion of carbon intensity and attribution based on numerical simulation and reduced order model according to claim 1, characterized in that, In step (5), the intensity prediction and location tracing of the carbon emission point source are performed by a source item estimation method, a nonlinear tracing method particle swarm algorithm, a Bayesian algorithm, an optimization algorithm, or an improved Gaussian diffusion inversion model.

5. A system for inversion of carbon intensity and tracing based on numerical simulation and reduced order model, implementing the method for inversion of carbon intensity and tracing based on numerical simulation and reduced order model of claim 1, characterized in that, The system for inverting carbon emission intensity and tracing based on numerical simulation and reduced order model comprises: a data acquisition and storage module for recording and storing sampling data, wherein the sampling data comprises sampling path information, sampling point spatial position information, distance measurement data, environmental images, CO2 concentration, and meteorological data; a data calculation module for processing the sampling data, establishing a high-precision three-dimensional digital map, establishing a simulation model, verifying simulation results, reducing the order of a POD model of a detected environmental concentration field, an airflow field, and a temperature field, establishing a simulation database, judging traceability, planning a patrol monitoring sampling path, analyzing emission source intensity, and tracing; a data visualization module for data statistics and analysis, wherein the data is presented in the form of a cloud chart and a statistical chart.

Images