A method, system, device and storage medium for parameterizing and characterizing atmospheric OH radicals in a target area

Abstract

The application provides a target area atmospheric OH radical parameterization characterization method, system, equipment and storage medium, and the method comprises the following steps: S1, obtaining initial data of a target area based on an air quality model according to model input data containing a range of the target area, wherein the initial data comprises OH radicals and characteristic factors; S2, establishing an initial model of the relationship between the OH radicals and the characteristic factors based on machine learning; S3, performing feature alignment on the characteristic factors and ground observation data of the target area by using a transfer learning method, and obtaining an optimization model based on the initial model; S4, screening the characteristic factors in the optimization model to obtain key characteristic factors, and establishing an OH radical simplified model based on the key characteristic factors. The characterization method realizes rapid, accurate and simplified characterization of atmospheric OH radicals, and the required cost and difficulty are significantly lower than direct field observation, and the accuracy is higher than numerical simulation.

Description

Technical Field

[0001] This application relates to the field of atmospheric environment simulation, and in particular to a method, system, device and storage medium for parameterized characterization of atmospheric OH radicals in a target area. Background Technology

[0002] Since the implementation of the Action Plan for Air Pollution Prevention and Control, the amount of fine particulate matter (PM2.5) in my country's atmosphere has been increasing. 2.5 While PM2.5 concentrations decreased significantly, ozone (O3) pollution problems have not been effectively improved. According to the "Blue Book on Atmospheric Ozone Pollution Prevention and Control in China (2023)," PM2.5 concentrations in key cities across the country in 2022... 2.5 The annual average concentration was 29 μg / m³. 3 Compared to 2015, it decreased by 40%, while the 90th percentile of the daily maximum 8-hour average O3 (MDA8 O3) increased by 6.3 μg / (m²) from 2015 to 2019. 3 O3 levels remained high after 2019, indicating that my country still faces a severe challenge in controlling O3 pollution. O3 is produced through nitrogen oxides (NOx). x NO is generated through the photochemical reaction of NO and volatile organic compounds (VOCs), and its nonlinear response to precursors leads to highly complex emission reduction pathway selection. In this process, hydroxyl radicals (OH) act as the core of the atmospheric oxidation chain reaction, controlling NO emission. x And the conversion of VOCs to O3. Therefore, obtaining the long-term variation characteristics of OH free radicals is of great significance for exploring the causes of O3 pollution.

[0003] For regional and urban areas, observational experiments and model simulations are currently the two main techniques for characterizing atmospheric OH radical concentrations. Observational experiments are the most direct method for obtaining atmospheric OH concentration levels, using instruments such as laser-induced fluorescence (LIF), chemical ionization mass spectrometry (CIMS), and differential absorption spectroscopy (DOAS). However, the complexity and high cost of these instruments, along with their sensitivity to environmental conditions, contribute to their high observation costs. Furthermore, the extremely short lifetime and very low concentration of OH radicals make their observation challenging. The number of research teams both domestically and internationally possessing free radical observation techniques is relatively limited, and the accumulated data is currently scarce and fragmented.

[0004] Given the complexity of observing OH radicals, researchers have begun to explore model simulations to quantify their concentration levels. Commonly used model simulation methods include observational models and regional numerical models. However, due to the lack of understanding of the underlying mechanisms in the models and the uncertainty of emission inventories, different models produce significantly different simulation results for radicals, exhibiting high uncertainty. Studies have shown that the simulation uncertainty for radicals can reach over 100%.

[0005] In summary, for characterizing atmospheric OH at the urban scale, field observations suffer from insufficient spatiotemporal coverage due to high observation costs, while numerical simulations are subject to certain biases due to uncertainties in emissions and mechanisms. There is an urgent need to develop new methods for characterizing atmospheric OH radicals to achieve accurate and simplified characterization of atmospheric free radicals and to promote their nationwide application. Summary of the Invention

[0006] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a parameterized characterization method, system, device and storage medium for atmospheric OH radicals in a target area, in order to solve the technical problems of high cost, difficulty and scarcity of data, low temporal and spatial coverage and large deviation of OH radicals obtained by air quality models in the prior art.

[0007] To achieve the above and other related objectives, this application provides a parameterized characterization method for atmospheric OH radicals in a target area. The method includes: S1: Obtaining initial data for the target area based on an air quality model using model input data encompassing the target area. The initial data includes OH radicals and characteristic factors, whereby the characteristic factors include data on pollutants and meteorological factors. The model input data includes natural source emission inventories, anthropogenic source emission inventories, and meteorological fields. S2: Establishing an initial model of the relationship between OH radicals and characteristic factors based on machine learning. S3: Using transfer learning, aligning the characteristic factors with ground observation data of the target area to obtain an optimized model based on the initial model. S4: In the optimized model, selecting key characteristic factors; and establishing a simplified OH radical model based on these key characteristic factors. This characterization method achieves rapid, accurate, and simplified characterization of atmospheric OH radicals. Its cost and difficulty are significantly lower than direct field observations, and its accuracy is higher than numerical simulations.

[0008] In one embodiment of this application, the air quality model includes a meteorological field model and a chemical transport model.

[0009] In one embodiment of this application, the data on pollutants and meteorological factors include all meteorological factor parameters, pollutant concentrations, and J-value parameters.

[0010] In one embodiment of this application, the natural source emission inventory uses MEGAN v3.0 output results; the anthropogenic source emission inventory is obtained by gridding activity levels and emission factors; the meteorological field is a global reanalysis meteorological simulation field; the meteorological factor data is obtained by WRFv3.9.1 simulation; and the pollutant concentration, OH radical, and Jvalue data are obtained by CMAQ v5.3.2 simulation.

[0011] The global reanalysis meteorological simulation field is a core infrastructure for modern meteorology and climate research. It combines massive amounts of observational data with numerical models through data assimilation technology to generate spatiotemporally continuous and physically consistent atmospheric state estimates.

[0012] In one embodiment of this application, the air quality model uses WRFv3.9.1 as the meteorological field model and CMAQv5.3.2 as the chemical transport model.

[0013] In one embodiment of this application, the air quality model is the WRF-CMAQ model. The WRF-CMAQ model simulates the OH radical concentration, total meteorological factor parameters, pollutant concentrations, and J-value parameters. WRF is a mesoscale meteorological model, and CMAQ is a mesoscale chemical transport model. WRF-CMAQ is a third-generation air quality model system strongly promoted by the U.S. Environmental Protection Agency (USEPA). It is designed based on the concept of "one atmosphere," considering all physical processes in the atmosphere and the chemical reaction processes of multiple species and configurations of pollutants. It considers processes such as chemical transport advection, gas-phase chemistry, plume treatment, dry sedimentation, and wet deposition; it also includes an aerosol module that can calculate aerosol transformation, providing multiple chemical mechanism options. It can be used for daily regional and urban-scale air quality forecasting, and can also be used to assess the effectiveness of pollutant emission reduction, predict the impact of environmental control strategies on air quality, and thus formulate optimal control plans.

[0014] The air quality model described in this application incorporates physical transport mechanisms (such as advection, diffusion, and deposition) and atmospheric chemical mechanisms (such as photochemical reactions and redox reactions), enabling it to quantitatively explain changes in atmospheric OH radical concentration through physicochemical processes. The extracted OH radical concentration exhibits a genuine physicochemical relationship with meteorological and pollutant data. Furthermore, the air quality model is not limited in its simulation area or time period, and can obtain full data on free radicals and related parameters for any city and time, thus compensating for the scarcity of atmospheric free radical observation data.

[0015] In one embodiment of this application, the CMAQ v5.3.2 uses the SAPRC07 gas phase chemical mechanism and the AERO6 aerosol mechanism, and employs 14 vertical layers.

[0016] In one embodiment of this application, the machine learning includes a decision tree algorithm. The decision tree algorithm has high selection accuracy and computational efficiency; therefore, the initial OH radical model established based on the data obtained from the air quality model, based on the above algorithm, has good physical and chemical properties.

[0017] In one embodiment of this application, the decision tree algorithm includes Random Forest (RF), Gradient Boosting (GB), and Extreme Gradient Boosting (XGBoost). The Random Forest (RF), Gradient Boosting (GB), and Extreme Gradient Boosting (XGBoost) algorithms offer high accuracy and computational efficiency in simulating the relationship between OH radical concentration and meteorological and pollutant conditions in this application.

[0018] In one embodiment of this application, the step of standardizing ground observation data is included before step S3. The standardization of ground observation data includes extracting ground observation data within a corresponding time period to retain only the data corresponding to the type of feature factor.

[0019] In one embodiment of this application, the alignment is achieved by minimizing the difference in feature factor distribution between the initial model and the ground observation data, thereby readjusting the weights of the feature factors in the initial model for machine learning. The transfer learning, by readjusting the feature factor weights in the initial model, avoids modeling failures caused by significant deviations between the simulated feature factors in the initial model and the feature factors in the ground observation data.

[0020] In one embodiment of this application, the ground observation data is obtained through field measurements of the target area.

[0021] Preferably, the measured data is data from a monitoring station.

[0022] In this application, the transfer learning does not change the physicochemical properties of the initial model, but only optimizes and adjusts the initial model parameters based on the systematic differences between air quality characteristic factors and observation characteristic factors, thus solving the problem of large deviations in the direct output data of the air quality model.

[0023] In one embodiment of this application, the screening method in S4 is a forward feature selection algorithm. The forward feature selection algorithm evaluates the importance and relative contribution of each feature factor to the optimization model. Based on the relative contribution of each feature factor, the correlation between multiple feature factors and the total feature factors reaches 0.85 or higher. Then, these multiple feature factors are key feature factors. The model corresponding to the key feature factors is the OH radical simplified model.

[0024] In one embodiment of this application, the characterization method further includes S5, which uses a transfer learning method to align key feature factors with ground observation data to further optimize the simplified OH radical model.

[0025] In one embodiment of this application, the characteristic factors include temperature, relative humidity, wind speed, sea level pressure, solar radiation, nitric oxide, nitrogen dioxide, and kOH at 2 × 10⁻⁶. 2 ~ 5×10 2ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and kOH at 5 × 10⁻⁶. 2 ~ 2.5×10 3 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and kOH at 2.5 × 10⁻⁶. 3 ~ 5×10 3 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and kOH at 5 × 10⁻⁶. 3 ~ 1×10 4 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and whose kOH concentration is greater than 1 × 10⁻⁶. 4 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, acetylene, ethylene, propylene, isoprene, and those with kOH < 7 × 10⁻⁶. 4 ppm -1 min -1 Alkenes, kOH > 7 × 10 4 ppm -1 min -1 Alkenes, benzene, toluene, kOH < 2×10 4 ppm -1 min -1 Aromatic hydrocarbons, m-xylene, p-xylene, o-xylene, kOH > 2×10 4 ppm -1 min -1 One or more of the following: aromatic hydrocarbons, formaldehyde, NO2 photolysis rate, maximum 8-hour daily concentration of O3, peroxyacetyl nitrate, carbon monoxide, inhalable particulate matter, and fine particulate matter.

[0026] In one embodiment of this application, the temperature and relative humidity are data at a height of 2m, and the wind speed is data at a height of 10m.

[0027] In one embodiment of this application, the kOH is at 2 × 10⁻⁶. 2 ~ 5×10 2 ppm -1 min -1 Ethane is an alkane and other non-aromatic compound that reacts only with OH.

[0028] In one embodiment of this application, the kOH is at 5 × 10⁻⁶. 2 ~ 2.5×10 3 ppm -1 min -1 The only alkanes and other non-aromatic compounds that react with OH are propane.

[0029] In one embodiment of this application, the kOH is at 2.5 × 10⁻⁶. 3 ~ 5×10 3 ppm -1 min -1 The alkanes and other non-aromatic compounds that react only with OH are selected from one or more of 2,3,4-trimethylpentane, isobutane, and n-butane.

[0030] In one embodiment of this application, the kOH is at 5 × 10⁻⁶. 3 ~ 1×10 4 ppm -1 min -1 The alkanes and other non-aromatic compounds that react only with OH are selected from one or more of 2,2-dimethylbutane, 2,3-dimethylbutane, 2,4-dimethylpentane, 2-methylpentane, 3-methylpentane, cyclopentane, isopentane, methylcyclopentane, n-heptane, n-hexane, and n-pentane.

[0031] In one embodiment of this application, the kOH is greater than 1 × 10⁻⁶. 4 ppm -1 min -1 The alkanes and other non-aromatic compounds that react only with OH are selected from one or more of the following: 2,4-trimethylpentane, 2-methylheptane, 2-methylhexane, 3-methylheptane, 3-methylhexane, cyclohexane, methylcyclohexane, n-decane, n-nonane, n-octane, and n-undecane.

[0032] In one embodiment of this application, the kOH < 7 × 10⁻⁶ 4 ppm -1 min -1 The olefin is selected from one or more of 1,3-butadiene, n-butene, and n-pentene.

[0033] In one embodiment of this application, the kOH > 7 × 10 4 ppm -1 min -1 The olefin is selected from one or more of 2-methyl-1-pentene, cis-2-butene, cis-2-pentene, styrene, trans-2-butene, and trans-2-pentene.

[0034] In one embodiment of this application, the kOH < 2 × 10⁻⁶ 4 ppm -1 min -1 The aromatic hydrocarbons are selected from one or more of ethylbenzene, isopropylbenzene, and n-propylbenzene.

[0035] To achieve the above and other related objectives, this application provides a parameterized characterization system for atmospheric OH radicals in a target region, comprising:

[0036] The initial data acquisition module is used to acquire initial data of the target area based on the air quality model, according to the model input data containing the target area range. The initial data includes OH radicals and characteristic factors, and the characteristic factors include data on pollutants and meteorological factors. The model input data includes natural source emission inventories, anthropogenic source emission inventories, and meteorological fields.

[0037] Initial Model Building Module: Used to establish an initial model for the relationship between OH radicals and feature factors based on machine learning;

[0038] The optimized model building module is used to align the feature factors with the ground observation data of the target area using the transfer learning method, and obtain an optimized model based on the initial model.

[0039] Simplified model building module: used to screen feature factors to obtain key feature factors in the optimization model; and to build a simplified OH radical model based on the key feature factors.

[0040] To achieve the above and other related objectives, this application provides a computer device, including: a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program stored in the memory to cause the device to perform the method described above.

[0041] To achieve the above and other related objectives, this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described above.

[0042] In summary, the method, system, device, and storage medium for parameterizing atmospheric OH radicals in a target region provided in this application have the following beneficial effects:

[0043] 1) The method described in this application can obtain the OH radical concentration results based on the characteristic factor data observed in any city and at any time period. It has high accuracy, low cost and low difficulty, and solves the problems of high difficulty and high cost of direct observation of OH, as well as large deviation of numerical simulation results of OH.

[0044] 2) The simplified model described in this application can characterize OH concentration based on a small number of key feature factors, and can be applied to time periods and regions with limited observation factors, providing key parameters for atmospheric oxidation regulation and control of secondary air pollutants. Attached Figure Description

[0045] Figure 1 The results shown are numerical simulations and machine learning initial and optimized models for predicting OH radicals in one embodiment of this application.

[0046] Figure 2 The image shows a comparison of OH predictions and observations using different methods in one embodiment of this application.

[0047] Figure 3 This is shown as an example of the importance analysis of different feature variables to OH machine learning in one embodiment of this application.

[0048] Figure 4 The results are shown as a comparison between the simplified OH model and the optimized OH model in one embodiment of this application.

[0049] Figure 5 The diagram shown is a flowchart of an atmospheric OH radical characterization method according to an embodiment of this application.

[0050] Figure 6 The diagram shown is a flowchart of an atmospheric OH radical characterization method according to an embodiment of this application.

[0051] Figure 7 The diagram shown is a structural schematic of an atmospheric OH radical characterization system according to an embodiment of this application.

[0052] Figure 8 The diagram shown is a structural schematic of a computer device according to an embodiment of this application. Detailed Implementation

[0053] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0054] It should be noted that in the following description, reference is made to the accompanying drawings, which illustrate several embodiments of this application. It should be understood that other embodiments may also be used, and changes in mechanical composition, structure, electrical system, and operation may be made without departing from the spirit and scope of this application. The following detailed description should not be considered limiting, and the scope of the embodiments of this application is defined only by the claims of the published patent. The terminology used herein is for describing particular embodiments only and is not intended to limit the scope of this application.

[0055] Throughout this specification, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," "fixing," and "holding" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0056] Furthermore, as used herein, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context indicates otherwise. The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data used are interchangeable where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. It should be further understood that the terms “comprising,” “including,” indicate the presence of the stated features, operations, elements, components, items, kinds, and / or groups, but do not exclude the presence, occurrence, or addition of one or more other features, operations, elements, components, items, kinds, and / or groups. The terms “or” and “and / or” as used herein are to be interpreted inclusively, or mean any one or any combination thereof. Thus, “A, B, or C” or “A, B, and / or C” means “any one of: A; B; C; A and B; A and C; B and C; A, B, and C.” An exception to this definition will only occur if the combination of elements, functions, or operations is inherently mutually exclusive in some way.

[0057] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the invention.

[0058] To address the existing problems, this application proposes a parameterized characterization method, system, device, and storage medium for atmospheric OH radicals in a target region. This addresses the technical issues in existing technologies where field observations suffer from insufficient spatiotemporal coverage due to high observation costs, and numerical simulations exhibit certain biases due to uncertainties in emissions, mechanisms, and other aspects.

[0059] The following embodiments of this application provide a specific method for characterizing urban OH radicals, such as... Figures 5-6 As shown, the specific steps include the following:

[0060] S1: Based on the model input data covering a large area of ​​the target region over a certain period of time, initial data for the target region is obtained based on the air quality model. The initial data includes OH radicals and characteristic factors, and the characteristic factors include data on pollutants and meteorological factors. The model input data includes natural source emission inventories, anthropogenic source emission inventories, and meteorological fields.

[0061] In one embodiment of this application, the air quality model includes a meteorological field model and a chemical transport model.

[0062] In one embodiment of this application, the air quality model uses WRFv3.9.1 as the meteorological field model and CMAQv5.3.2 as the chemical transport model.

[0063] In one embodiment of this application, the meteorological factor data is obtained by simulation using WRFv3.9.1; the pollutant concentration, OH radical and Jvalue data are obtained by simulation using CMAQ v5.3.2.

[0064] In one embodiment of this application, the air quality model is the WRF-CMAQ model. The WRF-CMAQ model simulates the OH radical concentration, total meteorological factor parameters, pollutant concentrations, and J-value parameters. WRF is a mesoscale meteorological model, and CMAQ is a mesoscale chemical transport model. WRF-CMAQ is a third-generation air quality model system strongly promoted by the U.S. Environmental Protection Agency (USEPA). It is designed based on the concept of "one atmosphere," considering all physical processes in the atmosphere and the chemical reaction processes of multiple species and configurations of pollutants. It considers processes such as chemical transport advection, gas-phase chemistry, plume treatment, dry sedimentation, and wet deposition; it also includes an aerosol module that can calculate aerosol transformation, providing multiple chemical mechanism options. It can be used for daily regional and urban-scale air quality forecasting, and can also be used to assess the effectiveness of pollutant emission reduction, predict the impact of environmental control strategies on air quality, and thus formulate optimal control plans.

[0065] The air quality model described in this application incorporates physical transport mechanisms (such as advection, diffusion, and deposition) and atmospheric chemical mechanisms (such as photochemical reactions and redox reactions), enabling it to quantitatively explain changes in atmospheric OH radical concentration through physicochemical processes. The extracted OH radical concentration exhibits a genuine physicochemical relationship with meteorological and pollutant data. Furthermore, the air quality model is not limited in its simulation area or time period, and can obtain full data on free radicals and related parameters for any city and time, thus compensating for the scarcity of atmospheric free radical observation data.

[0066] In one embodiment of this application, the CMAQ v5.3.2 uses the SAPRC07 gas phase chemical mechanism and the AERO6 aerosol mechanism, and employs 14 vertical layers.

[0067] In one embodiment of this application, the data on pollutants and meteorological factors include all meteorological factor parameters, pollutant concentrations, and J-value parameters.

[0068] In one embodiment of this application, the meteorological parameters include temperature, relative humidity, wind speed, sea level pressure, and solar radiation.

[0069] In one embodiment of this application, the temperature and relative humidity are data at a height of 2m, and the wind speed is data at a height of 10m.

[0070] In one embodiment of this application, the pollutant concentration includes nitric oxide, nitrogen dioxide, and kOH at a concentration of 2 × 10⁻⁶. 2 ~ 5×10 2 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and kOH at 5 × 10⁻⁶. 2 ~ 2.5×10 3 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and kOH at 2.5 × 10⁻⁶. 3 ~ 5×10 3 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and kOH at 5 × 10⁻⁶. 3 ~ 1×10 4 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and whose kOH concentration is greater than 1 × 10⁻⁶. 4 ppm -1 min -1Alkanes and other non-aromatic compounds that react only with OH, acetylene, ethylene, propylene, isoprene, and those with kOH < 7 × 10⁻⁶. 4 ppm -1 min -1 Alkenes, kOH > 7×10 4 ppm -1 min -1 Alkenes, benzene, toluene, kOH <2×10 4 ppm -1 min -1 Aromatic hydrocarbons, m-xylene, p-xylene, o-xylene, kOH > 2×10 4 ppm -1 min -1 Aromatic hydrocarbons, formaldehyde, maximum 8-hour daily concentration of O3, peroxyacetyl nitrate, carbon monoxide, inhalable particulate matter, and fine particulate matter.

[0071] In one embodiment of this application, the kOH is at 2 × 10⁻⁶. 2 ~ 5×10 2 ppm -1 min -1 Ethane is an alkane and other non-aromatic compound that reacts only with OH.

[0072] In one embodiment of this application, the kOH is at 5 × 10⁻⁶. 2 ~ 2.5×10 3 ppm -1 min -1 The only alkanes and other non-aromatic compounds that react with OH are propane.

[0073] In one embodiment of this application, the kOH is at 2.5 × 10⁻⁶. 3 ~ 5×10 3 ppm -1 min -1 The alkanes and other non-aromatic compounds that react only with OH are selected from one or more of 2,3,4-trimethylpentane, isobutane, and n-butane.

[0074] In one embodiment of this application, the kOH is at 5 × 10⁻⁶. 3 ~ 1×10 4 ppm -1 min -1The alkanes and other non-aromatic compounds that react only with OH are selected from one or more of 2,2-dimethylbutane, 2,3-dimethylbutane, 2,4-dimethylpentane, 2-methylpentane, 3-methylpentane, cyclopentane, isopentane, methylcyclopentane, n-heptane, n-hexane, and n-pentane.

[0075] In one embodiment of this application, the kOH is greater than 1 × 10⁻⁶. 4 ppm -1 min -1 The alkanes and other non-aromatic compounds that react only with OH are selected from one or more of the following: 2,4-trimethylpentane, 2-methylheptane, 2-methylhexane, 3-methylheptane, 3-methylhexane, cyclohexane, methylcyclohexane, n-decane, n-nonane, n-octane, and n-undecane.

[0076] In one embodiment of this application, the kOH < 7 × 10⁻⁶ 4 ppm -1 min -1 The olefin is selected from one or more of 1,3-butadiene, n-butene, and n-pentene.

[0077] In one embodiment of this application, the kOH > 7 × 10 4 ppm -1 min -1 The olefin is selected from one or more of 2-methyl-1-pentene, cis-2-butene, cis-2-pentene, styrene, trans-2-butene, and trans-2-pentene.

[0078] In one embodiment of this application, the kOH < 2 × 10⁻⁶ 4 ppm -1 min -1 The aromatic hydrocarbons are selected from one or more of ethylbenzene, isopropylbenzene, and n-propylbenzene.

[0079] In one embodiment of this application, the Jvalue parameter includes the NO2 photolysis rate.

[0080] In one embodiment of this application, the meteorological field is a global reanalysis meteorological simulation field; the natural source emission inventory uses MEGAN v3.0 output results; the anthropogenic source emission inventory is obtained by gridding activity levels and emission factors.

[0081] The global reanalysis meteorological simulation field is a core infrastructure for modern meteorology and climate research. It combines massive amounts of observational data with numerical models through data assimilation technology to generate spatiotemporally continuous and physically consistent atmospheric state estimates.

[0082] Air quality models encompass complete physical and chemical mechanisms, enabling quantitative explanations of atmospheric free radical concentration changes through physicochemical processes. Therefore, they can effectively characterize the relative changes between OH and various influencing factors. Furthermore, air quality models can output a full range of physicochemical factors, compensating for the scarcity of atmospheric free radical observational data. Based on these advantages, this embodiment first utilizes the WRF-CMAQ air quality model system to obtain a dataset of atmospheric free radicals and related influencing factors for subsequent machine learning.

[0083] In a specific embodiment of this application, the input data for the air quality model in S1 all include model input data within the historical time period of the target area. This model input data is fuzzy data and indirect data. In this embodiment, the meteorological field model is WRFv3.9.1, and the chemical transport model uses CMAQ v5.3.2. The CMAQ v5.3.2 model uses the SAPRC07 gas phase chemical mechanism and the AERO6 aerosol mechanism, employing 14 vertical layers. This embodiment uses a three-layer nested grid with resolutions of 36 km, 12 km, and 4 km, respectively. The anthropogenic emission inventory input uses data from the Tsinghua University MEIC team from 2017 to 2020, and the natural source emissions use the output results of MEGAN v3.0. Model input data including the target area refers to model input data including the target area and a larger area. For example, if the target area is Jing'an District of Shanghai, then the model input data including the target area can be one of the following: model input data of Jing'an District of Shanghai, model input data of Jing'an District of Shanghai and several nearby districts, model input data of Shanghai, or model input data of Shanghai and surrounding cities. The historical time period includes data from 2 to 10 years.

[0084] To ensure the extracted data samples cover multiple emission levels and weather types, this embodiment conducts hourly continuous simulations from 2017 to 2020. The primary target city in this embodiment is Shanghai, and the observation data comes from the Shanghai Academy of Environmental Sciences Superstation (SAES) results. Therefore, this embodiment extracts OH radicals (target factor) and influencing factors (characteristic factors) from the SAES grid. The influencing factors include five meteorological elements and NO... x The model extracts 32 characteristic factors, including 2 (VOCs), 19 (jNO2), 1 (JNO2), and 5 (other pollutants). The original data was at a 1-hour resolution, and was processed into daily data based on the characteristics of each factor, including three processing methods: daily average, daily maximum, and daily maximum 8-hour moving average. Table 1 shows all 33 factors that the model needs to extract and their related descriptions.

[0085] Table 1. Various factors extracted by WRF-CMAQ and their descriptions

[0086]

[0087]

[0088] S2: An initial model for the relationship between OH radicals and characteristic factors is established based on machine learning.

[0089] In one embodiment of this application, the machine learning includes a decision tree algorithm. The decision tree algorithm has high selection accuracy and computational efficiency; therefore, the initial OH radical model established based on the data obtained from the air quality model, based on the above algorithm, has good physical and chemical properties.

[0090] In one embodiment of this application, the decision tree algorithm includes Random Forest (RF), Gradient Boosting (GB), and Extreme Gradient Boosting (XGBoost). The Random Forest (RF), Gradient Boosting (GB), and Extreme Gradient Boosting (XGBoost) algorithms offer high accuracy and computational efficiency in simulating the relationship between OH radical concentration and meteorological and pollutant conditions in this application.

[0091] To achieve the goal of rapidly predicting atmospheric free radicals based on influencing factors, this embodiment establishes a mathematical relationship between atmospheric free radicals and influencing factors that is independent of physicochemical processes. This embodiment selects a machine learning algorithm based on decision tree theory, which is commonly used in the field of atmospheric environment due to its intuitive feature importance assessment and high computational efficiency. First, the training capabilities of commonly used decision tree algorithms are evaluated, specifically including Random Forest (RF), Gradient Boosting (GB), and Extreme Gradient Boosting (XGBoost), and the algorithm with superior simulation accuracy and computational efficiency for this study is selected.

[0092] In a specific embodiment of this application, based on the extracted OH radical target factors and characteristic factor data, this embodiment establishes a preliminary characterization method for atmospheric free radicals based on the decision tree algorithm. The specific modeling process is as follows:

[0093] Selecting a machine learning model: The simulation performance and efficiency of commonly used machine learning models are compared, and the results are shown in Table 2.

[0094] Table 2 Comparison of simulation performance and efficiency of commonly used machine learning models

[0095] Model [R 2 ]]> MAE RMSE Time consumed / s RF 0.94 100.19 143.96 24.51 GB 0.96 76.97 120.65 73.46 XGBoost 0.96 75.00 117.12 7.93

[0096] In this embodiment, GB and XGBoost have roughly the same accuracy, significantly higher than the RF model, while XGBoost runs about 10 times faster than the GB model. Considering both simulation bias and computational efficiency, this embodiment selects XGBoost as the subsequent machine learning model.

[0097] Data preprocessing: The model data from which 33 factors were extracted were subjected to time series alignment and standardization for use as machine learning input. The first three years (75%) of the 2017-2020 dataset were selected as the training set and 2020 as the test set (25%) to balance the model's generalization ability and adaptability to long-term changes.

[0098] Parameter tuning: During the learning process, the hyperparameters were tuned using the Optuna Python package through Bayesian optimization; and randomized 12-fold cross-validation was carried out to further optimize the model parameters.

[0099] S3: Using transfer learning, the feature factors are aligned with the ground observation data of the target area, and an optimized model is obtained based on the initial model.

[0100] In one embodiment of this application, the step of standardizing ground observation data is included before step S3. The standardization of ground observation data includes extracting ground observation data within a corresponding time period to retain only the data corresponding to the type of feature factor.

[0101] In one embodiment of this application, the alignment is achieved by minimizing the difference in feature factor distribution between the initial model and the ground observation data, thereby readjusting the weights of the feature factors in the initial model for machine learning. The transfer learning, by readjusting the feature factor weights in the initial model, avoids modeling failures caused by significant deviations between the simulated feature factors in the initial model and the feature factors in the ground observation data.

[0102] In one embodiment of this application, the ground observation data is obtained through field measurements of the target area.

[0103] Preferably, the measured data is data from a monitoring station.

[0104] In this application, the transfer learning does not change the physicochemical properties of the initial model, but only optimizes and adjusts the initial model parameters based on the systematic differences between air quality characteristic factors and observation characteristic factors, thus solving the problem of large deviations in the direct output data of the air quality model.

[0105] Although air quality models can effectively characterize the relative changes between OH radicals and various characteristic factors, the significant uncertainties in the atmospheric pollution source emission inventory input into the air quality model often lead to certain deviations between the absolute values ​​of pollutant concentrations and observation results. This is especially true for short-lived substances such as atmospheric free radicals, where the simulation deviations are even greater, often reaching orders of magnitude.

[0106] In a specific embodiment of this application, to avoid the aforementioned problems, the next step of this embodiment is to align the feature factors with the ground observation data of the target area, and optimize the characterization model of OH radicals through transfer learning. The feature factors of the ground observation data are consistent with the simulated feature factors (i.e., the feature factors shown in Table 1), and the observation data covers a longer period than the model input data, from 2017 to 2024. In addition, OH number concentration observation data accumulated in Shanghai from November 8, 2019 to December 4, 2019, are not used for machine learning due to the limited data sample, but only for model validation. The model optimization includes the following steps:

[0107] Data preprocessing includes standardizing ground observation data and matching it with simulation time labels, and removing samples from the ground observation data that are missing feature factors compared to those in the initial model. The feature factors of the ground observation data are divided into two segments: the first segment, from 2017 to 2019, overlaps with the model training period and is used for transfer learning to establish an optimized model that integrates the feature factors of the initial model and the ground observation data; the second segment, from 2020 to 2024, uses the feature factors of the ground observation data and the optimized model to directly predict the OH radical concentration during this period.

[0108] Transfer learning: The initial free radical model established above is adapted to the ground observation data of the target area using a feature distribution alignment method. The adaptation factors are the feature factors of the initial model from 2017 to 2019 and the feature factors of the ground observation data of the target area. By readjusting the weights of the feature factors in the initial model in machine learning, the difference in the distribution of feature factors in the initial model and the ground observation data is minimized, avoiding modeling failure due to large deviations between the feature factors simulated in the initial model and those in the ground observation data. An optimized atmospheric free radical characterization model is then established by retraining based on the weighted feature factors. For the prediction of OH free radicals from 2020 to 2024, the feature factors from the ground observation data are directly substituted into the optimized model for calculation. The above-mentioned initial OH characterization model is constructed based on air quality model data, and the final optimized OH characterization model is obtained by fusing ground observation data and model input data. In this embodiment, as... Figure 1 As shown, the optimized model represents a significantly lower concentration of OH than the CMAQ air quality model. Compared with the initial model, the summer high values ​​are similar, but the autumn and winter OH concentrations are higher than those in the initial model.

[0109] Model Validation: The accuracy of the optimized model was verified based on OH observation results from November 8, 2019 to December 4, 2019. Simultaneously, the OH free radical results from air quality simulations were compared to determine whether machine learning methods were more accurate than numerical simulations. For this embodiment, as... Figure 2As shown, the CMAQ air quality model results are significantly higher than observed data, while the initial model results are lower than observed data. Overall, the optimized model's characterization of OH is closest to the observed results.

[0110] S4: In the optimization model, feature factors are screened to obtain key feature factors; based on the key feature factors, a simplified model of OH radicals is established.

[0111] In one embodiment of this application, the screening method in S4 is a forward feature selection algorithm. The forward feature selection algorithm evaluates the importance and relative contribution of each feature factor to the optimization model. Based on the relative contribution of each feature factor, the correlation between multiple feature factors and the total feature factors reaches 0.85 or higher. Then, these multiple feature factors are key feature factors. The model corresponding to the key feature factors is the OH radical simplified model.

[0112] Because decision tree algorithms are resistant to overfitting, the above atmospheric OH radical characterization model did not perform feature factor screening and directly input 32 full feature factors as the final model result. However, for other cities or regions, there may be a problem of fewer observed feature factor types, making it impossible to obtain all variables. For Shanghai, there may also be situations where some feature factors are missing or have outliers. Therefore, this embodiment proposes to use forward feature selection (FFS) to screen out the variables that contribute the most to the OH characterization model. The FFS algorithm uses all possible variable combinations to train the model, identifies the best initial model, and then iteratively adds predictive variables until no more variables can improve the model performance. The variables selected at this point are the factors necessary for the parameterized characterization of OH radicals. This model is a simplified characterization model, applicable to time periods and regions with limited observed factors.

[0113] In a specific embodiment of this application, as can be seen from this embodiment, as Figure 3 As shown, temperature (T), radiation (SRAD), and photolysis constant (J) NO2 Ozone concentration (MDA8 O3) and atmospheric pressure (PSFC) are the five most important feature factors in OH machine learning, with a cumulative R0. 2 It can reach 0.86 of the total eigenfactor. For example... Figure 4 As shown, this embodiment utilizes the aforementioned five key characteristic factors to re-establish a simplified OH radical model. Comparison with the optimized OH model reveals a high correlation between the two. R0 2 The performance can reach over 0.99, indicating that the performance of the simplified OH model established in this embodiment after feature screening of key factors is basically comparable to that of the optimized OH model established with all feature factors.

[0114] This embodiment also specifically provides, for example, Figure 6The diagram also provides a flowchart of a parameterized characterization method for atmospheric OH radicals in a target region.

[0115] In summary, the parameterized characterization method for atmospheric OH radicals in a target area provided in this application can obtain OH radical concentration results based on an optimized model using characteristic factor data observed in any city at any time period. It has high accuracy, low cost and low difficulty, and can well meet research and management needs.

[0116] This embodiment also specifically provides, for example, Figure 7 The diagram also provides a structural schematic of a parameterized characterization system for atmospheric OH radicals in a target region. The atmospheric OH radical parameterized characterization system 10 includes:

[0117] The initial data acquisition module 11 is used to acquire initial data of the target area based on the air quality model according to the model input data containing the target area range. The initial data includes OH radicals and characteristic factors, and the characteristic factors include data of pollutants and meteorological factors. The model input data includes natural source emission inventories, anthropogenic source emission inventories and meteorological fields.

[0118] Initial Model Building Module 12: Used to establish an initial model for the relationship between OH radicals and feature factors based on machine learning;

[0119] Optimization model construction module 13: used to use transfer learning method to align the feature factors with the ground observation data of the target area and obtain an optimized model based on the initial model;

[0120] Simplified Model Building Module 14: Used to screen feature factors to obtain key feature factors in the optimization model; and to build a simplified OH radical model based on the key feature factors.

[0121] It should be understood that the division of the various modules in the above system is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software through processing element calls; they can be fully implemented in hardware; or some modules can be implemented through processing element calls in software, while others are implemented in hardware. Moreover, these modules can be fully or partially integrated together, or implemented independently. The processing element mentioned here can be an integrated circuit with signal processing capabilities. During implementation, each step of the above method or each of the above modules can be completed through the integrated logic circuits in the hardware of the processor element or through software instructions.

[0122] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more digital signal processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together to form a system-on-a-chip (SOC).

[0123] This embodiment also provides a computer device for parameterizing and characterizing atmospheric OH free radicals in a target area.

[0124] like Figure 8 The diagram shown illustrates the structure of a computer device 20 according to an embodiment of this application. The computer device 20 includes a memory 21 and a processor 22; the memory 21 stores computer instructions; the processor 22 executes the computer instructions to implement... Figure 5 The method described.

[0125] In some embodiments, the number of the memory 21 and the processor 22 in the computer device 20 can be one or more, while Figure 8 Each example is taken as an instance.

[0126] In one embodiment of this application, the processor 22 in the computer device 20 will operate as follows: Figure 5 The steps described involve loading one or more instructions corresponding to the process of an application into memory 21, and then having the processor 22 run the application stored in memory 21, thereby achieving the following: Figure 5 The method described.

[0127] For example, the memory 21 is used to store model programs and model input data; the processor 22 is used to execute the model programs stored in the memory 21, so that the computer device 20 performs actions such as... Figure 5 The method shown.

[0128] The memory 21 may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. The memory 21 stores an operating system and operating instructions, executable modules or data structures, or subsets thereof, or extended sets thereof. The operating instructions may include various operation instructions used to implement various operations. The operating system may include various system programs used to implement various basic services and handle hardware-based tasks.

[0129] The processor 22 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0130] In some specific applications, the various components of the computer device 20 are coupled together through a bus system, which may include, in addition to a data bus, a power bus, a control bus, and a status signal bus.

[0131] In one embodiment of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the following: Figure 5 The method described.

[0132] For example, the computer-readable storage medium stores model programs and related data.

[0133] As will be understood by those skilled in the art, the embodiments implementing the above-described system and its unit functions can be accomplished using hardware related to a computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the embodiments including the above-described system and its unit functions; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0134] In summary, the method, system, equipment, and storage medium for parameterizing OH radicals in the urban atmosphere of a target area provided in this application can obtain OH radical concentration results based on an optimized model using characteristic factor data observed in any city at any time period. This method is highly accurate, cost-effective, and relatively easy to implement, solving the problems of high difficulty and cost in direct OH observation, as well as large deviations in numerical simulation OH results.

[0135] This application effectively overcomes the various shortcomings of the prior art and has high industrial application value.

[0136] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A parameterized characterization method for atmospheric OH radicals in a target region, characterized in that, The method includes: S1: Based on the model input data covering the target area, obtain initial data for the target area using an air quality model. The initial data includes OH radicals and characteristic factors, which include pollutant and meteorological factor data. The model input data includes natural source emission inventories, anthropogenic source emission inventories, and meteorological fields. S2: An initial model for the relationship between OH radicals and characteristic factors is established based on machine learning; S3: Using transfer learning, the feature factors are aligned with the ground observation data of the target area, and an optimized model is obtained based on the initial model; S4: In the optimization model, feature factors are screened to obtain key feature factors; based on the key feature factors, a simplified model of OH radicals is established.

2. The characterization method according to claim 1, characterized in that, The air quality model includes a meteorological field model and a chemical transport model; And / or, the data for pollutants and meteorological factors include the full range of meteorological factor parameters, pollutant concentrations, and J-value parameters; And / or, the natural source emissions inventory uses MEGAN v3.0 output results; the anthropogenic source emissions inventory is obtained by gridding activity levels and emission factors; And / or, the meteorological field is a global reanalysis meteorological simulation field; And / or, the machine learning includes decision tree algorithms.

3. The characterization method according to claim 2, characterized in that, In the air quality model, the meteorological field model is WRFv3.9.1, the chemical transport model is CMAQ v5.3.2, and the meteorological factor data are obtained through simulation using WRFv3.9.1; the pollutant concentration, OH radical, and Jvalue data are obtained through simulation using CMAQ v5.3.2 based on the model input data. And / or, the air quality model is the WRF-CMAQ model; And / or, the decision tree algorithm includes random forest, gradient boosting, and extreme gradient boosting.

4. The characterization method according to claim 1, characterized in that, Before S3, there is also a step of standardizing ground observation data, which includes extracting ground observation data within the corresponding time period to retain only the data corresponding to the type of feature factor; And / or, the alignment readjusts the weights of the feature factors in the initial model in machine learning by minimizing the differences in feature factor distributions between the initial model and the ground observation data; And / or, the ground observation data is obtained through field measurements of the target area; And / or, the screening method described in S4 is a forward feature selection algorithm, which evaluates the importance and relative contribution of each feature factor to the optimization model. Based on the relative contribution of each feature factor, the features are accumulated from high to low until the correlation between multiple feature factors and the total feature factors reaches 0.85 or higher; then these multiple feature factors are key feature factors. The model corresponding to the key characteristic factor is the simplified model of OH free radical.

5. The characterization method according to claim 1, characterized in that, The characterization method also includes S5, which uses transfer learning to align key feature factors with ground observation data, further optimizing the simplified OH radical model.

6. The characterization method according to claim 1, characterized in that, The characteristic factors include temperature, relative humidity, wind speed, sea level pressure, solar radiation, nitric oxide, nitrogen dioxide, and kOH at 2 × 10⁻⁶. 2 ~ 5×10 2 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and kOH at 5 × 10⁻⁶. 2 ~ 2.5×10 3 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and kOH at 2.5 × 10⁻⁶. 3 ~ 5×10 3 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and kOH at 5 × 10⁻⁶. 3 ~ 1×10 4 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, and whose kOH concentration is greater than 1 × 10⁻⁶. 4 ppm -1 min -1 Alkanes and other non-aromatic compounds that react only with OH, acetylene, ethylene, propylene, isoprene, and those with kOH < 7 × 10⁻⁶. 4 ppm -1 min -1 Alkenes, kOH > 7×10 4 ppm -1 min -1 Alkenes, benzene, toluene, kOH < 2×10 4 ppm -1 min -1 Aromatic hydrocarbons, m-xylene, p-xylene, o-xylene, kOH > 2×10 4 ppm -1 min -1 Aromatic hydrocarbons, formaldehyde, NO2 photolysis rate, maximum 8-hour daily concentration of O3, peroxyacetyl nitrate, carbon monoxide, and inhalable particulate matter (PM2.5) 10 Fine particulate matter (PM) 2.5 One or more of them.

7. The characterization method according to claim 6, characterized in that, The kOH is at 2×10 2 ~ 5×10 2 ppm -1 min -1 Ethane is an alkane and other non-aromatic compound that reacts only with OH. And / or, the kOH is at 5 × 10 2 ~ 2.5×10 3 ppm -1 min -1 The only alkane and other non-aromatic compound that reacts with OH is propane; And / or, the kOH is at 2.5 × 10 3 ~ 5×10 3 ppm -1 min -1 The alkanes and other non-aromatic compounds that react only with OH are selected from one or more of 2,3,4-trimethylpentane, isobutane, and n-butane; And / or, the kOH is at 5 × 10 3 ~ 1×10 4 ppm -1 min -1 The alkanes and other non-aromatic compounds that react only with OH are selected from one or more of 2,2-dimethylbutane, 2,3-dimethylbutane, 2,4-dimethylpentane, 2-methylpentane, 3-methylpentane, cyclopentane, isopentane, methylcyclopentane, n-heptane, n-hexane, and n-pentane. And / or, the kOH is greater than 1 × 10⁻⁶ 4 ppm -1 min -1 The alkanes and other non-aromatic compounds that react only with OH are selected from one or more of the following: 2,4-trimethylpentane, 2-methylheptane, 2-methylhexane, 3-methylheptane, 3-methylhexane, cyclohexane, methylcyclohexane, n-decane, n-nonane, n-octane, and n-undecane. And / or, the kOH < 7 × 10 4 ppm -1 min -1 The olefin is selected from one or more of 1,3-butadiene, n-butene, and n-pentene; And / or, the kOH > 7 × 10 4 ppm -1 min -1 The olefin is selected from one or more of 2-methyl-1-pentene, cis-2-butene, cis-2-pentene, styrene, trans-2-butene, and trans-2-pentene; And / or, the kOH < 2 × 10 4 ppm -1 min -1 The aromatic hydrocarbons are selected from one or more of ethylbenzene, isopropylbenzene, and n-propylbenzene.

8. A parameterized characterization system for atmospheric OH radicals in a target region, characterized in that, include: The initial data acquisition module is used to acquire initial data for the target area based on an air quality model, using model input data that includes the target area range. The initial data includes OH radicals and characteristic factors, with the characteristic factors including data on pollutants and meteorological factors. The model input data includes natural source emission inventories, anthropogenic source emission inventories, and meteorological fields. Initial Model Building Module: Used to establish an initial model for the relationship between OH radicals and feature factors based on machine learning; The optimized model building module is used to align the feature factors with the ground observation data of the target area using the transfer learning method, and obtain an optimized model based on the initial model. Simplified model building module: used to filter feature factors to obtain key feature factors in the optimization model; A simplified model of OH radicals is established based on key characteristic factors.

9. A computer device for parameterized characterization of atmospheric OH radicals in a target region, characterized in that, The device includes: a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program stored in the memory to cause the device to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium for parameterized characterization of atmospheric OH radicals in a target region, wherein a computer program is stored thereon, characterized in that, When the program is executed by the processor, it implements the method described in any one of claims 1 to 7.

Images