Method for analyzing emission reduction sensitivity of atmospheric CO2 concentration level to PM2.5 pollution source

Abstract

A method for analyzing sensitivity of atmospheric CO2 concentration level to emission reduction of a PM2.5 pollution source relates to the field of collaborative emission reduction of atmospheric pollutants and greenhouse gases, and comprises the following steps: step 1, constructing a PM2.5 component and CO2 mixed input concentration matrix and an uncertainty matrix; 2, calculating pollution source contributions of PM2.5 and CO2 by adopting a specific factorization method; 3, constructing a carbon fluctuation index and a PM2.5 pollution source fluctuation index so as to establish a machine learning model training data set; step 4, constructing a nonlinear relationship between CO2 and PM2.5 pollution sources by adopting an XGBOOST algorithm, and predicting the CO2 concentration change by reducing the proportion of different PM2.5 pollution sources; 5, constructing a nonlinear relation between the PM2.5 pollution source and the CO2 pollution source by adopting an XGBOOST algorithm, and predicting the PM2.5 concentration change by reducing the proportion of different CO2 pollution sources; according to the method, a factorization method and a machine learning method are creatively combined, PM2.5 pollution source emission reduction and CO2 concentration are associated, the problem of separation between pollutant prevention and control and greenhouse gas prevention and control is solved, and the method has good application prospects in the aspects of air pollution prevention and control and climate improvement.

Description

Technical Field

[0001] The present invention belongs to the field of coordinated emission reduction of atmospheric pollutants and greenhouse gases, and specifically relates to a coupling PMF and XGBOOST model to analyze PM in the atmosphere. 2.5 and CO2 pollution source contribution and its assessment CO2 concentration on PM 2.5 Methods for reducing pollution source sensitivity. Background Art

[0002] Air pollution and the greenhouse effect have a wide range of impacts on humans and the living environment. 2.5 It is an important pollutant in the atmosphere, and CO2 is the main greenhouse gas. 2.5 As CO2 and PM2.5 share the same origin, it is crucial to design policies that can effectively reduce air pollution while also reducing carbon emissions. Therefore, it is necessary to study the effects of CO2 concentration levels on PM2.5. 2.5 Pollution source reduction sensitivity analysis is carried out to determine the optimal control sources that have both pollution reduction and carbon reduction effects. The receptor model is now used to conduct PM 2.5 Common methods of source analysis include CMB, PMF, etc. After reasonable calculation, the contribution of pollution sources to PM can be accurately obtained. 2.5 However, due to the increasing prominence of CO2, only PM 2.5 It is not enough to optimally control pollution sources. It is very important to identify the optimal pollution sources that also have a carbon reduction effect.

[0003] In the face of the above-mentioned increase in control requirements, the main problems with current technical methods are: (1) Due to the complexity of the atmospheric environment, it is difficult for existing receptor models to achieve CO2 and PM 2.5 (2) Some tracing methods require complex computing power; (3) CO2 and PM 2.5 There are complex nonlinear relationships between sources, making it difficult to conduct sensitivity analysis. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method for determining the atmospheric CO2 concentration level on PM 2.5 The method of sensitivity analysis of pollution source emission reduction comprehensively utilizes the particulate matter components and CO2 monitoring data of multiple periods, not only for PM 2.5 The sources of CO2 and PM2.5 emissions are monitored and analyzed, and machine learning methods are used to identify the optimal PM2.5 emissions that can simultaneously reduce CO2 concentrations. 2.5 source.

[0005] The present invention provides a method for determining the atmospheric CO2 concentration level on PM 2.5 The sensitivity analysis method for pollution source emission reduction adopts the following technical solutions:

[0006] (1) Build PM 2.5 Component and CO2 mixed input concentration matrix and uncertainty matrix

[0007] Reduce CO2 observations by 10 5 -10 6 Times and PM 2.5 The component values ​​are combined to form a mixed input matrix and factorized to obtain PM 2.5 The pollution source contribution results of CO2 are more accurate. The calculation steps and formulas of the mixed input matrix are as follows:

[0008] 1) CO2 concentration weighting:

[0009]

[0010] in, is the weighted CO2 concentration, unit: p mg / m 3 , p is the weighting coefficient, and its value is set to 10 5 -10 6 When the PM is about 2.5 and CO2 source contribution is relatively accurate; is the initial CO2 concentration, in mg / m 3 .

[0011] 2)PM 2.5 Component and CO2 mixed input matrix construction:

[0012]

[0013] Among them, X mix is the PM calculated in the input PMF 2.5 Components mixed with CO2 input matrix data; For PM 2.5 The component time series matrix data, is the time series data of CO2.

[0014] (2) Calculation of PM 2.5 and CO2 pollution sources

[0015] PM 2.5Chemical components include elements, ions, carbon, etc. Water-soluble ions include NH4+, Na+, K+, Ca2+, Mg2+, SO42-, NO3- and Cl-; carbon components include OC and EC; elements include Ca, Si, Al, Fe, K, V, Cr, Mn, Co, Ni, Cu, Zn, Ga, As, Se, Ag, Cd, Sn, Sb, Ba, Au, Hg, Tl, Pb and Bi components. Species can be selected according to specific circumstances. The concentration matrix and uncertainty matrix are included in the positive definite matrix factorization PMF model for decomposition. All species types are set to Strong, the number of runs is set to 20-30 times, and the number of factors is 6-9. The factor spectrum matrix and factor contribution matrix are optimized through rotation calculation. The factors are identified as different source categories according to the chemical identification components in the factor spectrum matrix. The factor contribution matrix is ​​used to compare with the atmospheric PM 2.5 Perform multiple linear regression to calculate the effect of pollution sources on atmospheric PM 2.5 The contribution of pollution sources to CO2 is obtained by multiplying the factor contribution matrix with the CO2 source spectrum matrix.

[0016] (3) Constructing the carbon fluctuation index and PM 2.5 Pollution source fluctuation index

[0017] Multiply the CO2 concentration by the PM 2.5 The carbon fluctuation index is constructed by the ratio of PM 2.5 The pollution source concentration is multiplied by 100 and divided by the CO2 pollution source concentration to construct PM 2.5 Pollution source volatility index.

[0018] (4) Machine learning method to calculate the effect of CO2 concentration on PM 2.5 Sensitivity of pollution source reduction

[0019] The XGBOOST algorithm is used to establish the carbon fluctuation index and PM 2.5 The nonlinear relationship of the pollution source fluctuation index, n_estimators is selected as 500-1000, max_depth is selected as 3-20, the training set is selected as 70-80%, and the test set is selected as 20-30%. 2.5 The pollution source fluctuation index is reduced by 10% each time it increases within the range of 10% to 50%, and the corresponding carbon fluctuation index is predicted to calculate the effect of CO2 concentration on PM 2.5 Sensitivity of pollution source reduction.

[0020] (5) Machine learning method to calculate PM 2.5 Sensitivity of concentration to CO2 pollution source reduction

[0021] Using XGBOOST algorithm to build PM 2.5The nonlinear relationship between the fluctuation index and the CO2 pollution source fluctuation index, n_estimators is selected as 500-1000, max_depth is selected as 3-20, the training set is selected as 70-80%, and the test set is selected as 20-30%. The CO2 pollution source fluctuation index is reduced by 10% each time in the range of 10% to 50%, and the corresponding PM is predicted. 2.5 Fluctuation index to calculate the effect of CO2 concentration on PM 2.5 Sensitivity of pollution source reduction.

[0022] A method of inducing atmospheric CO2 concentration level to PM 2.5 The method of sensitivity analysis of pollution source emission reduction has the following advantages: creatively combining factor decomposition method and machine learning method to 2.5 Linking pollution source emission reduction with CO2 concentration solves the problem of separation between pollutant prevention and control and greenhouse gas prevention and control, and has good application prospects in air pollution prevention and control and climate improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The effect of atmospheric CO2 concentration on PM 2.5 Technical flow chart of the pollution source emission reduction sensitivity analysis method;

[0024] Figure 2 PM obtained by the present invention 2.5 Source contribution and CO2 source contribution result graph;

[0025] Figure 3 The atmospheric CO2 concentration level of the present invention is the effect of PM 2.5 Pollution source reduction sensitivity results diagram;

[0026] Figure 4 The atmospheric PM 2.5 Sensitivity graph of concentration level to CO2 pollution source reduction; DETAILED DESCRIPTION

[0027] The following example shows the effect of atmospheric CO2 concentration level on PM 2.5 The pollution source emission reduction sensitivity analysis method is further described in detail so that those skilled in the art can implement it by referring to the text of the specification. Figure 1 The effect of atmospheric CO2 concentration on PM 2.5 The specific process of the method for sensitivity analysis of pollution source emission reduction is as follows:

[0028] 1PM 2.5 Component and CO2 mixed input concentration matrix and uncertainty matrix construction

[0029] Reduce CO2 observations by 10 5-10 6 times, using online monitoring of PM 2.5 Composition and CO2 concentration data, each PM 2.5 The components and CO2 are placed in the table header, and the species concentration data at each monitoring moment is completed as a row, thus forming PM 2.5 Components and CO2 mixed input concentration matrix. PM 2.5 The concentration of water-soluble ions in the chemical components was multiplied by 0.2; the carbon components were multiplied by 0.4; and the elemental components and CO2 were multiplied by 0.6 to construct the uncertainty matrix.

[0030] 2 Calculate PM 2.5 and CO2 pollution sources

[0031] The concentration matrix and uncertainty matrix were incorporated into the positive definite matrix factorization (PMF) model for decomposition. All species types were set to Strong, the number of runs was set to 20-30 times, and the number of factors was 6-9. The factor spectrum matrix and factor contribution matrix were optimized by rotation calculation. The factors were identified as different source categories according to the chemical identification components in the factor spectrum matrix. The factor contribution matrix was used to compare with the atmospheric PM 2.5 Perform multiple linear regression to calculate the effect of pollution sources on atmospheric PM 2.5 The contribution of the pollution source to CO2 is obtained by multiplying the factor contribution matrix with the CO2 source spectrum matrix. The result is as follows: Figure 2 shown.

[0032] 3 Construction of Carbon Fluctuation Index and PM 2.5 Pollution source fluctuation index

[0033] Multiply the CO2 concentration by the PM 2.5 The carbon fluctuation index is constructed by the ratio of PM 2.5 The pollution source concentration is multiplied by 100 and divided by the CO2 pollution source concentration to construct PM 2.5 Pollution source volatility index.

[0034] 4 Machine learning method to calculate the effect of CO2 concentration on PM 2.5 Sensitivity of pollution source reduction

[0035] The XGBOOST algorithm is used to establish the carbon fluctuation index and PM 2.5 The nonlinear relationship of the pollution source fluctuation index, n_estimators is selected as 500-1000, max_depth is selected as 3-20, the training set is selected as 70-80%, and the test set is selected as 20-30%. 2.5 The pollution source fluctuation index is reduced by 10% each time it increases within the range of 10% to 50%, and the corresponding carbon fluctuation index is predicted to calculate the effect of CO2 concentration on PM 2.5 The results are as follows: Figure 3 shown.

[0036] 5 Machine Learning Methods to Calculate PM 2.5 Sensitivity of concentration to CO2 pollution source reduction

[0037] Using XGBOOST algorithm to build PM 2.5 The nonlinear relationship between the fluctuation index and the CO2 pollution source fluctuation index, n_estimators is selected as 500-1000, max_depth is selected as 3-20, the training set is selected as 70-80%, and the test set is selected as 20-30%. The CO2 pollution source fluctuation index is reduced by 10% each time in the range of 10% to 50%, and the corresponding PM is predicted. 2.5 Fluctuation index to calculate the effect of CO2 concentration on PM 2.5 The results are as follows: Figure 4 shown.

Claims

1. Effect of atmospheric CO2 concentration level on PM 2.5 The method for sensitivity analysis of pollution source emission reduction is characterized by The method is implemented by the following steps: 1) Build PM 2.5 Component and CO2 mixed input concentration matrix and uncertainty matrix, reducing the CO2 observation value by 10 5 -10 6 Times and PM 2.5 The component values ​​together form the input data concentration matrix, PM 2.5 The component values ​​and CO2 concentrations were multiplied by 0.2-0.6 times as the input data uncertainty matrix; 2) Calculate PM 2.5 The pollution source contributions of CO2 and CO2 are analyzed by incorporating the concentration matrix and uncertainty matrix into the positive definite matrix factorization (PMF) model for decomposition. All species types are set to Strong, the number of runs is set to 20-30 times, and the number of factors is 6-9. When all factors can be explained and the Q / Qexp value is between 0.75-1.25, PM 2.5 and CO2 pollution source contributions; 3) Constructing carbon fluctuation index and PM 2.5 Pollution source fluctuation index, multiplying CO2 concentration by PM 2.5 The carbon fluctuation index is constructed by the ratio of PM 2.5 The pollution source concentration is multiplied by 100 and divided by the CO2 pollution source concentration to construct PM 2.5 Pollution source fluctuation index; 4) CO2 concentration on PM 2.5 The sensitivity of pollution source emission reduction is calculated by using the XGBOOST algorithm to establish the carbon fluctuation index and PM 2.5 The nonlinear relationship of the pollution source fluctuation index, n_estimators is selected as 500-1000, max_depth is selected as 3-20, the training set is selected as 70-80%, and the test set is selected as 20-30%. 2.5 The pollution source fluctuation index is reduced by 10% each time it increases within the range of 10% to 50%, and the corresponding carbon fluctuation index is predicted to calculate the effect of CO2 concentration on PM 2.5 Sensitivity of pollution source reduction; 5) PM 2.5 The sensitivity of concentration to CO2 pollution source reduction is calculated by using the same method to construct PM 2.5 Fluctuation index and carbon source fluctuation index to calculate PM 2.5 Sensitivity of concentration to CO2 pollution source reduction.

2. The effect of atmospheric CO2 concentration level on PM 2.5 The method for sensitivity analysis of pollution source emission reduction is characterized by: Reduce CO2 observations by 10 5 -10 6 Times and PM 2.5 The component values ​​are combined to form a mixed input matrix and factorized to obtain PM 2.5 The pollution source contribution results of CO2 are more accurate. The calculation steps and formulas of the mixed input matrix are as follows: 1) CO2 concentration weighting: in, is the weighted CO2 concentration, unit: p mg / m 3 , p is the weighting coefficient, and its value is set to 10 5 -10 6 When the PM is about 2.5 and CO2 source contribution is relatively accurate; is the initial CO2 concentration, in mg / m 3 ; 2)PM 2.5 Component and CO2 mixed input matrix construction: Among them, X mix is the PM calculated in the input PMF 2.5 Components mixed with CO2 input matrix data; For PM 2.5 The component time series matrix data, is the time series data of CO2.

3. The effect of atmospheric CO2 concentration level on PM 2.5 The method for sensitivity analysis of pollution source emission reduction is characterized by: Multiply the CO2 concentration by the PM 2.5 The carbon fluctuation index is constructed by the ratio of CO2 concentration to its average concentration, which better reflects the level of CO2 change and has more physical significance than simply using CO2 concentration.

4. The effect of atmospheric CO2 concentration level on PM 2.5 The method for sensitivity analysis of pollution source emission reduction is characterized by: PM 2.5 The pollution source concentration is multiplied by 100 and divided by the CO2 pollution source concentration to construct PM 2.5 Pollution source fluctuation index, which optimizes the contribution of CO2 pollution sources to PM 2.5 Interference from pollution sources.

Images