A method for source apportionment of atmospheric particulate matters refined to industry

Abstract

The application discloses a source analysis method for atmospheric particulate matters which is refined to an industry, and relates to the technical field of pollution source analysis, comprising: based on receptor data, a large category source is analyzed by a receptor model; a plurality of air quality models are used for simulation; urban scale particulate matter local emission and regional transportation proportion, primary source and secondary source redistribution relationship, urban emission and rural emission contribution, and different emission height contribution are obtained; based on the air quality model source analysis result, local and external contribution proportions of the large category source analyzed by the receptor model are distinguished, and a primary source category is refined and analyzed by using a source emission inventory. The receptor model provides a source category profile, the air quality model simulates a transmission and transformation process, and the emission inventory provides spatial distribution and intensity information of the source, which are complementary to each other, overcome the limitation of a single method, and make the analysis result have higher consistency and integrity in source category division, spatial attribution and process explanation.

Description

Technical Field

[0001] This invention relates to the field of pollution source analysis technology, and more specifically to a method for source analysis of atmospheric particulate matter that is refined to the industry level. Background Technology

[0002] Currently, the most commonly used source apportionment methods are receptor models, air quality models, and inventory methods, but each method has different limitations. Receptor models can effectively analyze open source contributions; they quantitatively analyze pollution source types without relying on detailed source strength information and meteorological fields. Air quality models quantitatively identify local and regional source contributions of pollutants, simulate pollutant transport evolution and dissipation processes, and predict pollution development trends. Source inventory methods are simple and easy to operate, but can only qualitatively or semi-quantitatively identify organized pollution sources. None of these three methods, used individually, can provide a complete and industry-specific analysis of various pollution sources.

[0003] However, with the deepening and refinement of air pollution prevention and control efforts, higher demands are being placed on the sustainability, accuracy, timeliness, and practicality of particulate matter source apportionment. Cities urgently need to address the relationship between local emissions and external transport; the refined sources of atmospheric particulate matter by species, region, industry, time period, or pollution process; the relationship between primary and secondary source redistribution; the proportion of urban and rural emissions; and the contributions of various emission methods (elevated sources, low-lying area sources, etc.). Changes in particulate matter composition and sources mean that traditional source apportionment techniques can no longer meet the current needs for precise and refined management decisions in air pollution prevention and control. Under the new circumstances, atmospheric particulate matter source apportionment needs to evolve with the times. Refined particulate matter source apportionment should not only include refined source apportionment of identified components, source types, industries, processes, spatial and temporal aspects of results; Therefore, how to provide a comprehensive and industry-specific source apportionment method for atmospheric particulate matter is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] In view of this, the present invention provides a source apportionment method for atmospheric particulate matter that is refined to an industry level, in order to solve the problems in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: This invention discloses a source apportionment method for atmospheric particulate matter with precision down to the industry level, comprising: Based on receptor data, the major sources analyzed by the receptor model are coal combustion sources, process sources, vehicle sources, dust sources, secondary sulfates, secondary nitrates, secondary organic matter, and other sources. Based on the city's high spatiotemporal resolution emission inventory, multiple air quality models were used for simulation, and the simulation results of each model were compared and verified with the monitoring data of each station as the benchmark. Using the air quality models, the proportion of local emissions and regional transport of particulate matter at the urban scale, the relationship between primary and secondary source redistribution, the contribution of urban emissions and rural emissions, and the contribution of different emission heights were obtained. Based on the source apportionment results of the air quality model, the proportion of local and external contributions of the major source categories identified by the receptor model is distinguished. The local contribution is further refined by analyzing the primary source categories using the source emission inventory. Dust sources are subdivided into road dust, construction dust and other dust, and coal-fired sources are subdivided into power plant coal-fired sources and residential coal-fired sources. Mobile sources are subdivided into gasoline vehicles, diesel vehicles and others, and industrial sources are subdivided into steel, cement, glass, coking, petrochemical and other industries.

[0006] Through the above technical solution, this invention organically combines the source identification capability of receptor models (such as PMF), the physicochemical mechanism simulation capability of air quality models, and the spatial distribution capability of high-resolution emission inventories to construct a multi-dimensional, multi-scale pollution source analysis framework. The receptor model provides source profiles, the air quality model simulates transport and transformation processes, and the emission inventory provides spatial distribution and intensity information of sources. The three complement each other, overcoming the limitations of a single method, and making the analysis results more consistent and complete in terms of source classification, spatial attribution, and process interpretation.

[0007] Preferably, in the above-mentioned source apportionment method for atmospheric particulate matter refined to an industry level, the specific steps for analyzing major source categories using the receptor model are as follows: Acquire particulate chemical composition data and gaseous precursor data for receptor sites; An improved PMF model is adopted, which introduces a similarity penalty term between source spectra into the objective function, and separates collinear first-order sources through adaptive iteration; We introduce mass balance weight and chemical species weight to perform dual-weight factor synergistic optimization of source contribution; Output the contribution values ​​of various sources and their time series changes.

[0008] The beneficial effects of this invention through the above technical solutions are as follows: By introducing an improved PMF model, a nonlinear secondary source generation model, and dual-weight factor synergistic optimization, the separation capability for collinear primary sources is enhanced, the physicochemical consistency of the secondary source generation process is improved, the accuracy and stability of source contribution estimation are optimized, and the continuity and reliability of the analysis results across spatiotemporal scales are ensured. A source spectrum similarity penalty term is introduced on the basis of the traditional PMF model, which enhances the separation capability for similar sources (such as coal-fired and industrial sources) by suppressing collinearity between source spectra. Mass balance weights ensure that the sum of source contributions and the measured particulate matter mass are closed, and chemical species weights ensure that the concentrations of characteristic species in each source spectrum are consistent with the measured composition. These dual constraints improve the physical rationality and chemical consistency of source contribution estimation.

[0009] Preferably, in the above-mentioned source apportionment method for atmospheric particulate matter refined to an industry level, the objective function formula for introducing a source spectrum similarity penalty term is as follows: ; in, The objective function is... Total number of receptor samples; The total number of chemical species involved in the analysis; For the first The first sample The residuals of each species; It is uncertain; The adaptive penalty coefficient varies with the number of iterations. ; For source spectrum and The square of the direction cosine reflects the angle between the spectra; and The source spectrum vector; This represents the total number of primary source factors.

[0010] Through the above technical solution, the beneficial effects of the present invention are as follows: by introducing a direction cosine penalty term based on the source spectrum angle into the objective function and combining it with an adaptive penalty coefficient, the ability to distinguish factors with high source spectrum similarity (such as coal-fired and industrial sources) is significantly improved, the risk of factor mixing and misjudgment is reduced, and the physical interpretability of factor analysis is enhanced.

[0011] Based on precursor concentrations and atmospheric oxidation potential, generation functions for secondary sulfates, nitrates, and organic matter are constructed. This allows the estimation of secondary sources to no longer rely on linear assumptions but instead reflects actual chemical reaction kinetics. In this industry-refined source apportionment method for atmospheric particulate matter, the secondary source generation model is as follows: The model for secondary sulfate formation is as follows: ; in, Represents atmospheric oxidation potential; The conversion efficiency coefficient. The reaction rate constant is obtained by combining nonlinear least squares with receptor data. This refers to the concentration of sulfur dioxide precursors. Air mass aging time; The model for secondary nitrate formation is as follows: ; in, It is the equilibrium constant. The concentration of nitric acid gas is determined by Obtained by oxidation, This represents the ammonium salt concentration, reflecting the chemical equilibrium constraints in the formation of ammonium nitrate. The model for the formation of secondary organic matter is as follows: ; Among them, This refers to the concentration of various volatile organic compound precursors. Here is the oxidation rate constant. This is the yield coefficient; for Total number of species classifications.

[0012] Through the above technical solution, the beneficial effects of the present invention are as follows: nonlinear generation models based on precursor concentration, oxidation potential and chemical equilibrium are constructed for secondary sulfate, secondary nitrate and secondary organic matter respectively, which can more accurately reflect the generation mechanism and environmental dependence of secondary particulate matter, and enhance the rationality of the mechanism and dynamic response capability of the secondary source contribution estimation.

[0013] Preferably, in the above-mentioned source apportionment method for atmospheric particulate matter refined to an industry level, mass balance weight and chemical species weight are introduced to perform dual-weight factor synergistic optimization of source contributions, with the final optimization objective being: ; in, and These are the weighting coefficients; The objective function for introducing a similarity penalty term between source spectra; For quality balance weights; Weights for chemical species; Constraints ; Measured total mass concentration of particulate matter; No. The contribution concentration of each source class; ; No. A collection of characteristic species from a single source; No. The first sample Measured concentrations for each species; No. The first of the sources Quality score of each species.

[0014] Through the above technical solution, the beneficial effects of the present invention are as follows: by introducing mass balance weight and chemical species weight, a dual-constraint optimization objective is constructed, which ensures the rationality of the source contribution estimation results in terms of mass closure and chemical composition consistency, and improves the physical reliability and stability of the analysis results at the chemical species level.

[0015] Preferably, in the above-mentioned source apportionment method for atmospheric particulate matter refined to an industry level, multiple air quality models are used for simulation, and the simulation results of each model are compared and verified based on station monitoring data. Multi-model weight fusion and bias correction are introduced. Using hourly concentration data from ground monitoring stations as a benchmark, the bias weights of each model at each station and at each time are calculated to construct dynamic fusion coefficients. The benchmark concentration field after dynamic fusion of multiple models is as follows: ; in, The total number of air quality models participating in the fusion; Spatial location coordinates; For time indexing; For the first The model is in location time Simulated concentration; For the first The model is in location ,time The fusion weights are determined by minimizing the root mean square error (RMSE) with respect to the monitored values: ; ; in For monitoring site collection, Monitoring site index To monitor concentration.

[0016] Through the above technical solution, the beneficial effects of the present invention are as follows: by adopting a dynamic fusion weight and bias correction method, the fusion coefficient of multiple air quality models is optimized based on monitoring data, which improves the spatiotemporal accuracy and consistency of simulation results, reduces the impact of systematic bias of a single model, and provides a high-precision reference concentration field for subsequent source contribution analysis.

[0017] Preferably, in the above-mentioned source apportionment method for atmospheric particulate matter refined to the industry level, an air quality model is used to obtain the proportions of local emissions and regional transport of particulate matter at the urban scale: A combination of marker-based species tracking and perturbation methods was used to define local contributions. With regional contribution : ; ; in The local source contribution concentration is simulated after only shutting down external emission sources. The baseline simulated concentration; Introducing dynamic atmospheric transport-reaction coupling coefficient Correcting the error caused by the superposition of contributions due to nonlinear chemical reactions: ; The secondary allocation correction of the source contribution of the nonlinear chemical process is achieved by fitting an offline chemical box model.

[0018] Through the above technical solution, the beneficial effects of the present invention are as follows: by combining the marker species tracking method and the perturbation method, and introducing a nonlinear transport-reaction coupling coefficient, it is possible to accurately distinguish the contribution of local and regional transport to the receptor point concentration, overcome the error of the linear superposition assumption in nonlinear chemical processes, and improve the accuracy and mechanism consistency of regional transport analysis.

[0019] Preferably, in the above-mentioned source apportionment method for atmospheric particulate matter refined to an industry level, a primary-secondary redistribution matrix is ​​established to decompose the total particulate matter concentration into: ; ; in, Direct contribution from primary emissions; Secondary generation contribution; For the first Secondary organic aerosols generated from precursor-like substances This refers to the concentration of secondary inorganic aerosols; In the distribution relationship model, a transformation elasticity coefficient is introduced. Describing precursors To secondary particulate matter components Conversion efficiency varies with environmental conditions: ; in, This is a function of temperature influence. The effect function of relative humidity; Ozone effect function.

[0020] Through the above technical solution, the beneficial effects of the present invention are as follows: by constructing a primary-secondary redistribution matrix and introducing a conversion elasticity coefficient, the environmental sensitivity modeling of the secondary particulate matter generation process is realized, which can dynamically reflect the influence of factors such as temperature, humidity, and ozone on the secondary generation efficiency, and improve the environmental adaptability and mechanism explanation ability of the secondary source contribution estimation.

[0021] Preferably, in the aforementioned method for source apportionment of atmospheric particulate matter refined to an industry level, urban source contributions are defined based on land use type and the spatial distribution of emission sources. Contribution to rural areas : ; ; in The concentration of urban source contribution was obtained by source region labeling, retaining only emission sources within urban administrative boundaries and closing emission sources in rural areas; the urban-rural transmission reciprocity coefficient was also introduced. This describes the mutual contribution of urban and rural emissions to atmospheric transport: ; in, The contribution of urban sources to urban points themselves. The contribution of rural resources to rural areas; Transmission coefficient from urban source to rural receiver; Transmission coefficient from rural sources to urban receivers; Through the above technical solution, the beneficial effects of this invention are as follows: by using the source region labeling method and the urban-rural transport reciprocity coefficient, the separation and interactive influence quantification of urban and rural source contributions are realized, revealing the feedback mechanism of atmospheric transport between urban and rural areas, and providing scientific support for regional collaborative governance and differentiated management between urban and rural areas.

[0022] Preferably, in the above-mentioned source apportionment method for atmospheric particulate matter refined to an industry level, the contribution analysis of different emission heights stratifies the emission sources according to their height. , , The contribution of each height layer is calculated using the layer-by-layer closure method: ; in, To retain only the first Simulated concentrations after shutting down emission sources at other altitude levels; Introducing a vertical contribution attenuation function This describes the attenuation of the contribution of emissions at different altitudes to ground concentrations with distance: ;in For characteristic attenuation height, The decay exponent is obtained through multi-layer model simulation and highly coupled inversion with the boundary layer.

[0023] The beneficial effects of the present invention through the above technical solution are as follows: by using the layer-by-layer closure method and the vertical contribution attenuation function, the contribution of emissions at different height levels to ground concentration and its attenuation law with distance are quantified, which enhances the ability to identify the impact of elevated sources and near-ground sources, and provides technical support for vertical structure optimization and the formulation of emission reduction strategies for elevated sources.

[0024] As can be seen from the above technical solutions, compared with the prior art, this invention discloses a source apportionment method for atmospheric particulate matter that is refined to the industry level. Based on a technical architecture of multi-source data fusion and multi-model collaboration, this industry-level pollution source apportionment method significantly improves the ability to separate collinear primary sources, the accuracy of the mechanism characterization of secondary particulate matter generation processes, the mass closure and chemical consistency of source contribution estimation, the spatiotemporal accuracy of the benchmark concentration field, the nonlinear decoupling ability of local and regional transport, the quantification ability of urban and rural source interaction, and the vertical analysis ability of contributions at different emission heights through a series of innovative mechanisms such as improved receptor models, nonlinear secondary source generation models, dual-weight factor optimization, multi-model dynamic fusion, transport-reaction coupling correction, urban-rural transport reciprocity coefficient, and vertical contribution attenuation function. Thus, it achieves a full-chain technical breakthrough from source identification, spatial attribution, process analysis to result correction, providing high-resolution and high-reliability scientific support for precise source tracing and collaborative governance of regional air pollution. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0026] Figure 1 A flowchart of the method provided by the present invention; Figure 2 This is a schematic diagram showing the distribution of national and municipal monitoring stations provided by the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] This invention discloses a source apportionment method for atmospheric particulate matter with industry-specific precision, such as... Figure 1 As shown, the specific steps are as follows: Step 1: Pollution source analysis based on receptor model; 1. Data Acquisition and Preprocessing: Several receptor sampling points were set up in the target city to collect atmospheric particulate matter samples and obtain data on the chemical composition of particulate matter and gaseous precursors.

[0029] Specifically, taking into full account the current air pollution situation in Urumqi and Ili Kazakh Autonomous Prefecture, and based on the distribution of national and municipal monitoring stations (such as... Figure 2 Based on the administrative divisions, pollution source distribution, and development plans, it is proposed to set up 6 sampling points in Urumqi City, located in Midong District, Toutunhe District, Xinshi District, Tianshan District, Shuimogou District, and Urumqi County. In Ili Kazakh Autonomous Prefecture, it is proposed to set up 12 sampling points: one receptor sampling point each at two national control stations (Yining City), and one sampling point each in the other 10 counties and cities (Kuitun City, Khorgos City, Yining County, Chabuchar Xibe Autonomous County, Huocheng County, Gongliu County, Xinyuan County, Zhaosu County, Tekes County, and Nileke County). A total of 18 sampling points will be set up to collect particulate matter samples. The collected samples will undergo chemical composition analysis, including: Carbon composition: Organic carbon (OC) and elemental carbon (EC) were determined using a thermoluminescent carbon analyzer. Water-soluble ions: determined using ion chromatography. , , , , , , wait; Inorganic elements: Al, Si, Ca, Fe, Ti, Pb, Zn, Cu, As, etc. were determined by inductively coupled plasma mass spectrometry (ICP-MS). Gaseous precursors: Simultaneous collection of concentration data for gaseous pollutants such as SO2, NO2, O3, and VOCs.

[0030] All data underwent quality control, outliers were removed, and missing values ​​were supplemented using interpolation.

[0031] 2. Initial source spectrum library construction: Collect typical pollution source composition spectra from the local area and surrounding regions to construct an initial source spectrum library, including: Coal source spectrum: Collect particulate matter samples from coal-fired sources such as power plants, industrial boilers, and heating boilers, and analyze their chemical composition characteristics; Process source spectrum: For key industries such as steel, cement, and chemical, emission samples are collected from each stage of the production process; Motor vehicle emission profile: Collect motor vehicle emission samples from locations such as tunnels, parking lots, and roadsides to distinguish between gasoline and diesel vehicles; Dust source spectrum: Collect samples of road dust, construction dust, soil dust, etc.

[0032] Each source spectrum is expressed as a mass fraction, that is, the proportion of each chemical species in the particulate matter emitted from the source.

[0033] 3. Analysis of the improved PMF model: An improved PMF (positive definite matrix factorization) model was used to analyze the receptor data. This model introduces a similarity penalty term between source spectra into the objective function of the traditional PMF model to solve the problem of difficulty in distinguishing collinear source spectra.

[0034] (1) Construction of the objective function; The objective function formula for introducing the source spectrum similarity penalty term is as follows: ; in, The objective function is... Total number of receptor samples; The total number of chemical species involved in the analysis; For the first The first sample The residuals of each species; It is uncertain; The adaptive penalty coefficient varies with the number of iterations. ; For source spectrum and The square of the direction cosine reflects the angle between the spectra; and The source spectrum vector; This represents the total number of primary source factors. The design principle of the penalty term is: when two source spectra are highly similar, When the value approaches 0, the penalty term is relatively small; when there is a difference between the two source spectra, the penalty term increases, thereby driving the model to separate collinear source spectra during the iteration process.

[0035] The adaptive penalty coefficient λ varies with the number of iterations: ; in, This represents the number of iterations. This is the initial penalty coefficient, with a value ranging from 0.1 to 10; The attenuation coefficient ranges from 0.01 to 0.1. Adaptive design ensures strong source spectrum separation in the early stages of iteration, gradually relaxing the constraints in later stages to guarantee convergence to the global optimum.

[0036] (2) Model solution; The objective function is optimized and solved using the alternating least squares method. The specific iterative steps are as follows: ① Initialize the source spectrum matrix F and contribution matrix G; ② Fix F, update G; ③ Fix G, update F; ④ Calculate the Q value of the objective function; ⑤ Repeat steps ② to ④ until the Q value converges or the maximum number of iterations is reached (e.g., 500 times).

[0037] After each iteration, the source spectrum matrix is ​​normalized to ensure that the magnitude of each source spectrum vector is 1.

[0038] 4. Nonlinear quadratic source generation model; A nonlinear secondary source generation model based on precursor concentration and atmospheric oxidation potential was constructed to calculate the contributions of secondary sulfate, secondary nitrate and secondary organic matter.

[0039] (1) Secondary sulfate formation model; The formation of secondary sulfate is affected by The model expression for the influence of oxidation rate and atmospheric oxidation capacity is as follows: ; in, Represents atmospheric oxidation potential; This refers to the concentration of sulfur dioxide precursors. The air mass aging time, in hours, is calculated using a backward trajectory model. The conversion efficiency coefficient; is the reaction rate constant.

[0040] and The solution is obtained by combining nonlinear least squares method with receptor data. Specifically, the measured secondary sulfate concentration is used as the dependent variable, and... As the independent variable, the fit is determined. and The optimal value.

[0041] (2) Secondary nitrate formation model; The formation of secondary nitrates is controlled by the chemical equilibrium between nitric acid gas and ammonium salt, and the model expression is as follows: ; in, This is the equilibrium constant, with a value ranging from 0.1 to 10. The concentration of nitric acid gas is determined by... Obtained by oxidation, Oxidation rate is used with Similar function form; This represents the ammonium salt concentration, obtained directly from receptor data.

[0042] This model reflects the chemical equilibrium constraints of ammonium nitrate formation: when ammonium salt is sufficient, nitrate formation is controlled by the concentration of nitric acid gas; when ammonium salt is insufficient, nitrate formation is limited by the concentration of ammonium salt.

[0043] (3) Secondary organic matter generation model; The formation of secondary organic aerosols (SOA) involves the oxidation process of various VOCs precursors, and the model expression is as follows: ; in, This refers to the concentrations of various volatile organic compound precursors, categorized by species (e.g., alkanes, alkenes, aromatic hydrocarbons, etc.). The oxidation rate constant is obtained with reference to atmospheric chemical mechanisms (such as the MCM mechanism); The yield coefficient was determined by fitting receptor data. for Total number of species classifications.

[0044] The oxidation rate constants and yield coefficients of various VOCs were calibrated using an offline chemical chamber model (such as the F0AM model), and the parameters were optimized by combining the measured SOA concentrations at the acceptor sites.

[0045] 5. Collaborative optimization using dual weighting factors; By introducing mass balance weights and chemical species weights, source contributions are synergistically optimized using dual weighting factors. The final optimization objective is: ; in, and These are the weighting coefficients; The objective function for introducing a similarity penalty term between source spectra; For quality balance weights; Weights for chemical species; The mass balance weight is defined as: ; in, This represents the measured total mass concentration of particulate matter. For the first The contribution concentration of each source class. The mass balance weight reflects the consistency between the sum of source contributions and the measured total mass, with the following constraints: This ensures that the relative error in quality balance does not exceed 20%.

[0046] Chemical species weight is defined as: ; in, For the first A set of characteristic species from a single source (e.g., the characteristic species of a coal-fired source are As, Se). wait); For the first The first sample Measured concentrations for each species; For the first The first of the sources Quality score of each species; The number of elements in the characteristic species set. The weight reflects the degree of matching between the source spectrum characteristic species and the measured concentration, ideally approaching 1.

[0047] Dual-weight collaborative optimization is carried out iteratively: during the PMF model iteration process, each time the source spectrum and contribution matrix are updated, the current mass balance weight and chemical species weight are calculated and added as a penalty term to the objective function, driving the model to optimize in the direction of satisfying mass balance and chemical species matching.

[0048] 6. Output results; The model outputs the contribution concentrations and time-series changes of various pollution sources, including: Concentration and time series of coal-fired power contribution; Concentration and time series of source contributions in the process; Concentration and time series of contributions from motor vehicle sources; Dust source contribution concentration and time series; Contribution concentrations and time series of secondary sulfate, secondary nitrate, and secondary organic matter; Contribution of concentration and time series from other sources (such as biomass combustion, sea salt, etc.).

[0049] Step 2: Simulation and verification based on the air quality model; 1. Construction of high spatiotemporal resolution emission inventories; Based on pollution source emission survey data from the city and surrounding areas, a high spatiotemporal resolution emission inventory was constructed. The spatial resolution of the emission inventory was set to 1 km × 1 km, and the temporal resolution was set to hours, covering [a certain area / region]. , , , , , , Major pollutants such as...

[0050] The emissions inventory was compiled using a combination of bottom-up and top-down approaches. Industrial sources are allocated based on the company's actual emission data and geographical location; Traffic sources are allocated based on road network data and traffic flow data, according to road type and time period; Dust sources are allocated based on land use type and meteorological conditions; Residential sources are allocated based on population density and energy consumption data.

[0051] 2. Simulation using multiple air quality models; At least three mainstream air quality models were selected for simulation, including but not limited to: CMAQ model (Community Multiscale Air Quality Model): The third-generation air quality model developed by the U.S. Environmental Protection Agency, which uses the CB6r3 gas phase chemical mechanism and the AERO7 aerosol mechanism; CAMx model (Comprehensive Air Quality Model with Extensions): A comprehensive air quality model developed by Ramboll, Inc. in the United States, which uses the CB6 gas phase chemical mechanism and the CF aerosol mechanism. WRF-Chem model: A fully online coupled meteorological-chemical model employing the MOZART gas-phase chemistry mechanism and the GOCART aerosol mechanism.

[0052] Each model uses the same meteorological field input (provided by the WRF model) and the same emission inventory, with a simulation period covering at least one year, and outputs a surface concentration field (spatial resolution 3km×3km, temporal resolution hour) and a three-dimensional concentration field (vertically layered with at least 15 layers).

[0053] 3. Dynamic fusion of multiple models and bias correction; Based on hourly concentration data from ground monitoring stations, the simulation results of each model are dynamically fused and biased.

[0054] The baseline concentration field after dynamic fusion of multiple models is defined as follows: ; in, The total number of air quality models participating in the fusion; Spatial location coordinates; For time indexing; For the first The model is in location time Simulated concentration; For the first The model is in location ,time The fusion weights are determined by minimizing the root mean square error (RMSE) with respect to the monitored values: ; ; in For monitoring site collection, Monitoring site index To monitor concentration.

[0055] This optimization problem is a quadratic programming problem, which can be solved using the Lagrange multiplier method or gradient descent method. For grids without monitoring stations, spatial interpolation methods (such as the inverse distance weighting method) are used to extrapolate the weights from neighboring stations.

[0056] The deviation correction adopts a rolling window method: the weights are calculated based on the data of the past 72 hours and applied to the simulation correction of the next 24 hours. The window is updated every 24 hours to achieve dynamic adaptive correction.

[0057] Step 3: Analysis of pollution sources at the city scale; 1. Analysis of the ratio of local emissions to regional transport; We used a combination of marker species tracking and perturbation methods to analyze the contribution ratio of local emissions to regional transport.

[0058] Marked species tracking method: Marked species are set in the air quality model to mark pollutants from local emission sources (within the administrative boundary of the target city) and regional emission sources (outside the boundary), respectively. The transport, transformation and deposition processes of marked species are tracked to directly calculate the local contribution and the regional contribution.

[0059] Perturbation method: Design two sets of sensitivity simulation experiments: Baseline simulation: Simulations are performed using a complete emissions inventory to obtain... ; Local contribution simulation: Simulations were conducted by shutting down external emission sources and retaining only local emission sources to obtain... .

[0060] Local contribution and regional contribution are defined as follows: ; ; in The local source contribution concentration is simulated after only shutting down external emission sources. The baseline simulated concentration; Due to the nonlinearity of atmospheric chemical processes, + and There is a discrepancy between them. To correct this discrepancy, a dynamic atmospheric transport-reaction coupling coefficient is introduced. :

[0061] The determination was achieved through fitting an offline chemical chamber model. Specifically, the pollutant concentrations from local and regional emissions were input into a zero-dimensional chemical chamber model to simulate their nonlinear interaction and fit the model. The relationship function between meteorological conditions (temperature, humidity, and light intensity).

[0062] 2. The redistribution relationship between primary and secondary sources; A primary-secondary redistribution matrix is ​​established to decompose the total particulate matter concentration into the direct contribution of primary emissions and the contribution of secondary generation.

[0063] The basic decomposition relationship is: ; ; in, Direct contribution from primary emissions; Secondary generation contribution; For the first Secondary organic aerosols generated from precursor-like substances This refers to the concentration of secondary inorganic aerosols; in, The direct contribution to primary emissions is obtained by summing the contributions from primary sources (coal combustion, process, motor vehicle, dust, etc.) analyzed by the receptor model; Contribute to secondary generation; For the first Secondary organic aerosols generated from precursors (aromatic hydrocarbons, alkenes, alkanes, etc.); This refers to the concentration of secondary inorganic aerosols (sulfates, nitrates, ammonium salts); and This is the allocation coefficient.

[0064] To further characterize the transformation efficiency of secondary generation as environmental conditions change, a transformation elasticity coefficient is introduced. :

[0065] in, This is the baseline value for conversion efficiency under standard conditions; This is a function of temperature influence. The effect of relative humidity is a function for sulfates and nitrates. It is usually a piecewise function, and the conversion efficiency is enhanced at higher relative humidity; This is the ozone effect function, reflecting the contribution of oxidation capacity.

[0066] The specific parameters of each influence function were determined by inversion from indoor smoke chamber experiments or field observation data.

[0067] 3. Analysis of the contributions of urban and rural emissions; Based on land use type and spatial distribution of emission sources, the contributions of urban and rural emissions are analyzed.

[0068] Based on land use classification data (such as MODIS land use products), spatial grids are divided into urban grids and rural grids. Urban grids are defined as those within urban administrative boundaries where the impermeable surface accounts for more than 50%; rural grids are defined as those outside urban administrative boundaries where the impermeable surface accounts for less than 20%.

[0069] The urban source contribution and the rural source contribution are defined as follows: ; ; in The concentration of urban source contribution was obtained by source region labeling, retaining only emission sources within urban administrative boundaries and closing emission sources in rural areas; the urban-rural transmission reciprocity coefficient was also introduced. This describes the mutual contribution of urban and rural emissions to atmospheric transport: ; in, The contribution of urban sources to urban points themselves. The contribution of rural resources to rural areas; Transmission coefficient from urban source to rural receiver; Transmission coefficient from rural sources to urban receivers; The transmission coefficient is determined in the following way: ; ; in, The contribution concentration of urban sources to rural receptor sites. The contribution concentrations of rural sources to urban receptor sites were obtained directly through source region labeling.

[0070] 4. Analysis of the contribution of different emission heights; The emission sources were stratified by height to analyze the contribution of different emission heights to ground concentration.

[0071] The emission sources are stratified by height as follows: H1 layer: 0~50m (near ground emission, mainly including motor vehicles, dust, and low-rise sources); H2 layer: 50~200m (medium and low-rise power sources, mainly including industrial chimneys and heating boilers); H3 layer: 200~500m (elevated source, mainly including power plant chimneys and large industrial chimneys); H4 level: above 500m (ultra-high-rise source, mainly including high-altitude emission facilities).

[0072] The contribution of each height layer was calculated using the layer-by-layer closure method: ; To retain only the first Simulated concentrations after shutting down emission sources at other altitude levels; Introducing a vertical contribution attenuation function This describes the attenuation of the contribution of emissions at different altitudes to ground concentrations with distance: ;in For characteristic attenuation height, The decay exponent is obtained through multi-layer model simulation and highly coupled inversion with the boundary layer.

[0073] Specifically, for emission sources at different altitudes, their ground-level concentration contributions are simulated, and the model is fitted. Function, Determine and The value of . Studies have shown that under stable boundary layer conditions, Smaller The larger value indicates that the elevated source contributes less to the ground surface; under convective boundary layer conditions, Larger The smaller value indicates that the elevated source can affect the ground concentration through turbulent mixing.

[0074] Step 4: Multi-source data fusion and analysis; By fusing the results of receptor model analysis with those of air quality model analysis, a more refined analysis of pollution sources can be achieved.

[0075] The specific fusion method is as follows: 1. Alignment of major source contributions: The major source categories (coal combustion, process, motor vehicle, dust, secondary sulfate, secondary nitrate, and secondary organic matter) resolved by the receptor model serve as the basic classification framework.

[0076] 2. Local / External Contribution Decomposition: Using the source apportionment results of the air quality model, the local contribution ratio and external contribution ratio of each major source category analyzed by the receptor model are further decomposed.

[0077] 3. Primary / Secondary redistribution: For secondary sources (sulfates, nitrates, organic matter), the redistribution relationship between primary and secondary sources is analyzed using an air quality model to further distinguish the source composition of their precursors.

[0078] 4. Urban / Rural Contribution Decomposition: Using the urban and rural contribution analysis results, the contribution of each type of source is further decomposed into urban and rural sources.

[0079] 5. Decomposition of emission height contribution: Using the emission height contribution analysis results, the contribution of each type of source is further decomposed into different emission heights.

[0080] 6. Use the source emission inventory to conduct a detailed analysis of primary sources, subdividing dust sources into road dust, construction dust and other dust, coal-fired sources into power plant coal-fired sources and residential coal-fired sources; mobile sources into gasoline vehicles, diesel vehicles and others, and industrial sources into steel, cement, glass, coking, petrochemical and other industries.

[0081] The final output includes multi-dimensional and multi-level pollution source analysis results, including: The absolute contribution concentration and relative contribution ratio of various pollution sources; Local and external contributions from various pollution sources; Secondary generation contributions corresponding to various primary sources; The urban and rural contributions of various pollution sources; The varying degrees of contribution from different pollution sources; Various pollution sources are identified into different subcategories. Dust sources are subdivided into road dust, construction dust, and other dust; coal-fired sources are subdivided into power plant coal-fired sources and residential coal-fired sources; mobile sources are subdivided into gasoline vehicles, diesel vehicles, and others; and industrial sources are subdivided into steel, cement, glass, coking, petrochemical, and other industries.

[0082] The time series variation and spatial distribution characteristics of the above results.

[0083] Example 2:

[0084] This embodiment is basically the same as Embodiment 1, except that: In step one, the receiver sampling points are set up using a dense sampling method, with sampling points set up in different functional areas such as the urban core area, industrial area, residential area, main traffic artery, and background area. Each functional area has no less than 3 sampling points to reflect the spatial differences within the city.

[0085] In step two, the selection of multiple air quality models further includes regional scale models and urban scale models. The regional scale models are CMAQ and CAMx, and the urban scale models are ADMS-Urban and AERMOD. Through the coupling of multi-scale models, the full-scale simulation from regional transport to local urban diffusion is realized.

[0086] In step three, the analysis of the ratio of local emissions to regional transport further incorporates isotope tracing technology. By analyzing the ratios of stable isotopes such as lead, strontium, and neodymium in particulate matter and combining them with an isotope source library, the accuracy of the model analysis results is independently verified.

[0087] Example 3:

[0088] This embodiment is basically the same as Embodiment 1, except that: In step one, the adaptive penalty coefficient λ of the improved PMF model is hyperparameter tuned using Bayesian optimization to minimize the model cross-validation error. The optimal combination of λ0 and α is automatically searched to improve the robustness of the model's analysis.

[0089] In step three, the parameters z0 and β of the vertical contribution attenuation function are mapped to meteorological parameters (boundary layer height, wind speed, turbulence intensity) using machine learning methods (such as random forest), so as to realize the dynamic adjustment of the attenuation function under different meteorological conditions.

[0090] Technical effectiveness verification: To verify the technical effectiveness of the present invention, an application verification was conducted for one year in a typical city (population of about 5 million, industrial structure mainly based on steel and chemical industries).

[0091] The verification results show that: 1. Improved source resolution accuracy: Compared with the traditional PMF model, the improved PMF model improves the distinguishability of collinear source spectra (such as coal-fired sources and process sources) by about 35% and reduces cross-validation error by about 20%.

[0092] 2. Enhanced secondary source analysis capability: After introducing a nonlinear secondary source generation model, the correlation coefficients between the analysis results of secondary sulfate, secondary nitrate and secondary organic matter and the estimation results based on tracers (such as the ratio of sulfate to sulfur dioxide) reached 0.92, 0.89 and 0.85, respectively, which are significantly better than the traditional methods of 0.78, 0.72 and 0.68.

[0093] 3. Significant effect of multi-model fusion: The root mean square error (RMSE) between the baseline concentration field and the monitored values ​​after dynamic fusion is reduced by about 25% compared with the single optimal model, and the correlation coefficient is improved by about 0.08.

[0094] 4. Multi-dimensional analysis capability: It enables simultaneous analysis of multiple dimensions such as local / regional, primary / secondary, urban / rural, and different altitudes, providing comprehensive support for the formulation of refined pollution control strategies.

[0095] 5. Practical Application Results: Based on the analysis results of this method, a city implemented clean energy replacement for coal-fired boilers (contributing 18% of the total emissions) and restricted high-emission vehicles from motor vehicle sources (contributing 22% of the total emissions). After one year, the city's annual average PM2.5 concentration decreased by about 15%, with the contributions from coal-fired and motor vehicle sources decreasing by about 25% and 18%, respectively, verifying the accuracy and practicality of the analysis results.

[0096] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0097] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A source apportionment method for atmospheric particulate matter refined to an industry level, characterized in that, include: Based on receptor data, the major sources analyzed by the receptor model are coal combustion sources, process sources, vehicle sources, dust sources, secondary sulfates, secondary nitrates, secondary organic matter, and other sources. Based on the city's high spatiotemporal resolution emission inventory, multiple air quality models were used for simulation, and the simulation results of each model were compared and verified with the monitoring data of each station as the benchmark. Using the air quality models, the proportion of local emissions and regional transport of particulate matter at the urban scale, the relationship between primary and secondary source redistribution, the contribution of urban emissions and rural emissions, and the contribution of different emission heights were obtained. Based on the source apportionment results of the air quality model, the proportion of local and external contributions of the major source categories identified by the receptor model is distinguished. The local contribution is further refined by analyzing the primary source categories using the source emission inventory. Dust sources are subdivided into road dust, construction dust and other dust, and coal-fired sources are subdivided into power plant coal-fired sources and residential coal-fired sources. Mobile sources are subdivided into gasoline vehicles, diesel vehicles and others, and industrial sources are subdivided into steel, cement, glass, coking, petrochemical and other industries.

2. The source apportionment method for atmospheric particulate matter refined to an industry level according to claim 1, characterized in that, The specific steps for analyzing major source classes using the receptor model are as follows: We acquired particulate chemical composition data and gaseous precursor data for receptor sites; we adopted an improved PMF model, introduced a source spectrum similarity penalty term into the objective function, and separated collinear primary sources through adaptive iteration. We introduce mass balance weight and chemical species weight to perform dual-weight factor synergistic optimization of source contribution; Output the contribution values ​​of various sources and their time series changes.

3. The source apportionment method for atmospheric particulate matter refined to an industry level according to claim 2, characterized in that, The objective function formula for introducing the source spectrum similarity penalty term is as follows: ； in, The objective function is... Total number of receptor samples; The total number of chemical species involved in the analysis; For the first The first sample The residuals of each species; It is uncertain; The adaptive penalty coefficient varies with the number of iterations. ; For source spectrum and The square of the direction cosine reflects the angle between the spectra; and The source spectrum vector; This represents the total number of primary source factors.

4. The source apportionment method for atmospheric particulate matter refined to an industry level according to claim 3, characterized in that, Based on the air quality model, the secondary source generation model is as follows: The model for secondary sulfate formation is as follows: ； in, Represents atmospheric oxidation potential; The conversion efficiency coefficient. The reaction rate constant is obtained by combining nonlinear least squares with receptor data. This refers to the concentration of sulfur dioxide precursors. Air mass aging time; The model for secondary nitrate formation is as follows: ; Among them, It is the equilibrium constant. The concentration of nitric acid gas is determined by Obtained by oxidation, This represents the ammonium salt concentration, reflecting the chemical equilibrium constraints in the formation of ammonium nitrate. The model for the formation of secondary organic matter is as follows: ； Among them, This refers to the concentration of various volatile organic compound precursors. Here is the oxidation rate constant. This is the yield coefficient; for Total number of species classifications.

5. The source apportionment method for atmospheric particulate matter refined to an industry level according to claim 4, characterized in that, By introducing mass balance weights and chemical species weights, source contributions are synergistically optimized using dual weighting factors. The final optimization objective is: ; in, and These are the weighting coefficients; The objective function for introducing a similarity penalty term between source spectra; For quality balance weights; Weights for chemical species; Constraints ; Measured total mass concentration of particulate matter; No. The contribution concentration of each source class; ； No. A collection of characteristic species from a single source; No. The first sample Measured concentrations for each species; No. The first of the sources Quality score of each species.

6. The source apportionment method for atmospheric particulate matter refined to an industry level according to claim 1, characterized in that, Multiple air quality models were used for simulation, and the simulation results of each model were compared and verified against the monitoring data of each station. Multi-model weight fusion and bias correction were introduced. Using the hourly concentration data of ground monitoring stations as the benchmark, the bias weight of each model at each station and at each time was calculated to construct a dynamic fusion coefficient. The benchmark concentration field after dynamic fusion of multiple models is as follows: ; in, The total number of air quality models participating in the fusion; Spatial location coordinates; For time indexing; For the first The model is in location time Simulated concentration; For the first The model is in location ,time The fusion weights are determined by minimizing the root mean square error (RMSE) with respect to the monitored values: ； ； in For monitoring site collection, Monitoring site index To monitor concentration.

7. The source apportionment method for atmospheric particulate matter refined to an industry level according to claim 1, characterized in that, Using air quality models, the proportions of local particulate matter emissions and regional transport at the urban scale were obtained: A combination of marker-based species tracking and perturbation methods was used to define local contributions. With regional contribution : ； ； in The local source contribution concentration is simulated after only shutting down external emission sources. The baseline simulated concentration; Introducing dynamic atmospheric transport-reaction coupling coefficient Correcting the error caused by the superposition of contributions due to nonlinear chemical reactions: ; The secondary allocation correction of the source contribution of the nonlinear chemical process is achieved by fitting an offline chemical box model.

8. The source apportionment method for atmospheric particulate matter refined to an industry level according to claim 1, characterized in that, Establish a primary-secondary redistribution matrix to decompose the total particulate matter concentration into: ； ； in, Direct contribution from primary emissions; Secondary generation contribution; For the first Secondary organic aerosols generated from precursor-like substances This refers to the concentration of secondary inorganic aerosols; In the distribution relationship model, a transformation elasticity coefficient is introduced. Describing precursors To secondary particulate matter components Conversion efficiency varies with environmental conditions: ; in, This is a function of temperature influence. The effect function of relative humidity; Ozone effect function.

9. The source apportionment method for atmospheric particulate matter refined to an industry level according to claim 1, characterized in that, Define urban source contribution based on land use type and spatial distribution of emission sources. Contribution to rural areas : ； ； in The concentration of urban source contribution was obtained by source region labeling, retaining only emission sources within urban administrative boundaries and closing emission sources in rural areas; the urban-rural transmission reciprocity coefficient was also introduced. This describes the mutual contribution of urban and rural emissions to atmospheric transport: ； in, The contribution of urban sources to urban points themselves. The contribution of rural resources to rural areas; Transmission coefficient from urban source to rural receiver; Transmission coefficient from rural sources to urban receivers.

10. The source apportionment method for atmospheric particulate matter refined to an industry level according to claim 1, characterized in that, Analysis of the contribution of different emission heights, stratifying emission sources by height. , , The contribution of each height layer is calculated using the layer-by-layer closure method: ; in, To retain only the first Simulated concentrations after shutting down emission sources at other altitude levels; Introducing a vertical contribution attenuation function This describes the attenuation of the contribution of emissions at different altitudes to ground concentrations with distance: ;in For characteristic attenuation height, The decay exponent is obtained through multi-layer model simulation and highly coupled inversion with the boundary layer.

Images