Training method and device and prediction method and device of atmospheric pollution regulation and control model

Abstract

The invention provides a training method and device of an atmospheric pollution regulation and control model, an atmospheric pollution information prediction method and device, electronic equipment, a computer readable storage medium and a computer program product. The training method comprises the following steps: utilizing an air quality model, and according to meteorological condition information and a pollutant emission sample, calculating the atmospheric pollution regulation and control model according to the meteorological condition information and the pollutant emission sample; obtaining first pollutant prediction information; generating a training sample; obtaining second pollutant prediction information according to the training sample by using an atmospheric pollution regulation model; and determining a model loss function according to the first pollutant prediction information and the second pollutant prediction information, and performing adjustment processing on the atmospheric pollution regulation and control model. According to the method and the device, the pollutants such as PM2.5 can be predicted, pollutant discharge regulation and control can be performed, and on the basis of ensuring accurate prediction of the pollutants, the demand on computing resources can be remarkably reduced, the computing efficiency can be improved, the accuracy of pollutant prediction can be improved, and the scientificity and the precision of pollution treatment can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of atmospheric environmental science, and in particular, to a training method and device for an air pollution control model, an air pollution information prediction method and device, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] Atmospheric chemical transport models (CTMs) are used to simulate historical pollutant concentrations or provide operational air quality forecasts; air pollution control models simulate, analyze, and predict the temporal and spatial distribution changes of air pollutants under different pollution control strategies by establishing mathematical models. Models that use the principles of atmospheric chemical transport for air pollution control simulation include various air quality models such as CMAQ (Community Multiscale Air Quality). Currently, existing atmospheric chemical transport models are usually established based on partial differential (PDE) equations that describe atmospheric physical and chemical processes in detail. Due to the high coupling and complexity of the PDE equations, the solution process is very complex, the computational efficiency is low, a large amount of computing resources and a long computing time are required, and the applicability in real pollution control scenarios is limited. Summary of the Invention

[0003] The present disclosure provides a training method and device for an air pollution control model, an air pollution information prediction method and device, an electronic device, a computer-readable storage medium, and a computer program product.

[0004] According to a first aspect of the present disclosure, there is provided a training method for an air pollution control model, including: using an air quality model to obtain first pollutant prediction information according to meteorological condition information and pollutant emission samples; generating a training sample, where the training sample includes the meteorological condition information, the pollutant emission samples, and pollutant information in a historical state, and a label of the training sample is the first pollutant prediction information; using the air pollution control model to obtain second pollutant prediction information according to the training sample; determining a model loss function according to the first pollutant prediction information and the second pollutant prediction information; and adjusting and processing the air pollution control model according to the model loss function.

[0005] Optionally, obtaining the second pollutant prediction information by using the air pollution regulation model according to the training samples includes: generating meteorological variable input data according to the meteorological condition information; generating emission variable data according to the pollutant emission samples; generating historical state background data according to the pollutant information in the historical state; using the air pollution regulation model to obtain pollutant characteristic information according to the meteorological variable input data and the emission variable data, wherein the pollutant characteristic information includes pollutant concentration change information; using the air pollution regulation model to sum the pollutant concentration change information and the historical state background data to obtain pollutant concentration information; and obtaining the second pollutant prediction information, wherein the second pollutant prediction information includes the pollutant concentration information and the pollutant concentration change information.

[0006] Optionally, the air pollution regulation model includes a 3D convolutional layer, an activation function, a residual module, an encoder, a decoder, and a 2D convolutional layer; using the air pollution regulation model to obtain pollutant characteristic information according to the meteorological variable input data and the emission variable data includes: using the 3D convolutional layer to perform 3D spatial feature extraction on the meteorological variable input data and the emission input data; using the activation function to perform activation processing on the output data of the 3D convolutional layer, wherein a skip connection is established between the input end and the output end of the 3D convolutional layer through the residual module; using the encoder to perform downsampling and pooling processing on the output data of the activation function; using the decoder to perform upsampling processing on the output data of the encoder, wherein the encoder includes a 3D convolutional layer and a pooling layer, the decoder includes a plurality of transposed convolutional layers, and a skip connection is established between each layer of the encoder and the corresponding layer of the decoder; using the 2D convolutional layer to process the output data of the encoder to obtain the pollutant characteristic information, wherein the number of the 2D convolutional layers is multiple; using the air pollution regulation model to sum the pollutant concentration change information and the historical state background data to obtain pollutant concentration information includes: summing the pollutant concentration change information output by the 2D convolutional layer and the historical state background data to obtain the pollutant concentration information.

[0007] Optionally, both the first pollutant prediction information and the second pollutant prediction information include pollutant concentration information and pollutant concentration change information. The adaptive weights include a concentration adaptive weight and a concentration change adaptive weight. Determining the model loss function based on the first pollutant prediction information, the second pollutant prediction information, and the adaptive weights of the air pollution control model includes: determining a concentration loss function based on the pollutant concentration information in the first pollutant prediction information and the pollutant concentration information in the second pollutant prediction information; determining a concentration change loss function based on the pollutant concentration change information in the first pollutant prediction information and the pollutant concentration change information in the second pollutant prediction information; determining a concentration accuracy loss function based on the concentration loss function, the concentration adaptive weight, and a preset constant; determining a concentration change accuracy loss function based on the concentration change loss function, the concentration change adaptive weight, and the preset constant; and determining the model loss function based on the concentration accuracy loss function, the concentration change accuracy loss function, the concentration adaptive weight, and the concentration change adaptive weight.

[0008] Optionally, determining the model loss function based on the concentration accuracy loss function, the concentration change accuracy loss function, the concentration adaptive weight, and the concentration change adaptive weight includes: determining penalty information based on the concentration adaptive weight and the concentration change adaptive weight; and determining the model loss function based on the concentration accuracy loss function, the concentration change accuracy loss function, the concentration change adaptive weight, the concentration change adaptive weight, and the penalty information.

[0009] Optionally, the pollutant emission samples include a baseline scenario emission sample and a regulated scenario emission sample. The training method includes: sampling the anthropogenic source pollutant emission information in the atmospheric emission information to obtain a baseline scenario emission sample; obtaining atmospheric emission regulation information based on the anthropogenic source pollutant emission information in the atmospheric emission information; and sampling the atmospheric emission regulation information to obtain a regulated scenario emission sample.

[0010] Optionally, obtaining the atmospheric emission regulation information based on the anthropogenic source pollutant emission information in the atmospheric emission information includes: regulating the anthropogenic source pollutant emission information in the atmospheric emission information according to an emission regulation factor, as well as spatial and temporal dimensions, to obtain the atmospheric emission regulation information.

[0011] Optionally, sampling the atmospheric emission regulation information to obtain a regulated scenario emission sample includes: sampling the atmospheric emission regulation information according to the spatial dimension, the temporal dimension, and the sampling dimension to obtain the regulated scenario emission sample; sampling the anthropogenic source pollutant emission information in the atmospheric emission information to obtain a baseline scenario emission sample includes: sampling the anthropogenic source pollutant emission information in the atmospheric emission information according to the spatial dimension, the temporal dimension, and the sampling dimension to obtain the baseline scenario emission sample; wherein, the sampling dimension includes a pollution source dimension and a pollutant type dimension, the pollution source dimension includes at least one dimension of electricity, industry, and other industries; the pollutant type dimension includes at least one dimension of sulfur dioxide, nitrogen oxides, non-methane volatile organic compounds, ammonia, and particulate matter.

[0012] Optionally, using a meteorological model, the meteorological condition information is obtained according to meteorological reanalysis data, observation assimilation data, and terrain data.

[0013] Optionally, the pollutant information of the historical state, the first pollutant prediction information, and the second pollutant prediction information include: the concentrations of PM2.5 and oxides, the change amounts of the concentrations, and the components of PM2.5 and the types of oxides.

[0014] According to a second aspect of the present disclosure, there is provided a method for predicting air pollution information, including: generating model input data according to meteorological condition information, atmospheric emission information, and pollutant information of a historical state; using an air pollution regulation model to obtain pollutant prediction information according to the model input data; wherein, the air pollution regulation model is trained by the training method as described above.

[0015] According to a third aspect of the present disclosure, there is provided a training device for an air pollution regulation model, including: a first data obtaining module, configured to use an air quality model to obtain first pollutant prediction information according to meteorological condition information and a pollutant emission sample; a sample generating module, configured to generate a training sample, wherein the training sample includes the meteorological condition information, the pollutant emission sample, and the pollutant information of the historical state, and the label of the training sample is the first pollutant prediction information; a second data obtaining module, configured to use the air pollution regulation model to obtain second pollutant prediction information according to the training sample; a loss determining module, configured to determine a model loss function according to the first pollutant prediction information and the second pollutant prediction information; a model adjusting module, configured to perform an adjustment process on the air pollution regulation model according to the model loss function.

[0016] According to a fourth aspect of the present disclosure, there is provided an air pollution information prediction device, including: a data generation module configured to generate model input data according to meteorological condition information, air emission information, and pollutant information of a historical state; an information prediction module configured to generate pollutant prediction information according to the model input data by using an air pollution regulation model; wherein, the air pollution regulation model is obtained by training through the training method as described above.

[0017] According to a fifth aspect of the present disclosure, there is provided an electronic device, including: a memory; and a processor coupled to the memory, the processor being configured to execute the method as described above based on instructions stored in the memory.

[0018] According to a sixth aspect of the present disclosure, there is provided a computer-readable storage medium storing computer instructions, and when the instructions are executed by a processor, the steps of the method as described above are implemented.

[0019] According to a seventh aspect of the present disclosure, there is provided a computer program product including a computer program, and when the computer program is executed by a processor, the steps of the method as described above are implemented.

[0020] The training method, device, air pollution information prediction method, device, electronic device, computer-readable storage medium, and computer program product of the air pollution regulation model of the present disclosure use an air quality model to obtain first pollutant prediction information, generate training samples according to meteorological condition information, pollutant emission samples, and pollutant information of a historical state, and the label of the training sample is the first pollutant prediction information, so as to use the air pollution regulation model and obtain second pollutant prediction information according to the training samples; according to the first pollutant prediction information and the second pollutant prediction information, a model loss function is determined, and the air pollution regulation model is adjusted; through the trained air pollution regulation model, prediction of pollutants such as PM2.5, pollutant emission regulation, etc. can be performed, which can significantly reduce the demand for computing resources, improve computing efficiency, improve the accuracy of pollutant prediction, and enhance the scientificity and accuracy of pollution control. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By describing the embodiments of the present disclosure in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present disclosure will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present disclosure, and constitute a part of the specification, and are used to explain the present disclosure together with the embodiments of the present disclosure, and do not constitute a limitation to the present disclosure. The following describes the above and other objects and advantages of the present disclosure in more detail in conjunction with specific embodiments and with reference to the drawings. In the drawings, the same or corresponding technical features or components will be represented by the same or corresponding reference numerals.

[0022] Figure 1 Flow diagram of some embodiments of a method for training an air pollution control model according to the present disclosure;

[0023] Figure 2 Flow diagram of obtaining a pollutant emission sample in some embodiments of a method for training an air pollution control model according to the present disclosure;

[0024] Figure 3 Schematic diagram of a multi-dimensional cube structure for sampling;

[0025] Figure 4 Flow diagram of obtaining second pollutant prediction information in some embodiments of a method for training an air pollution control model according to the present disclosure;

[0026] Figure 5 Flow diagram of obtaining second pollutant prediction information in some embodiments of a method for training an air pollution control model according to the present disclosure;

[0027] Figure 6 Schematic diagram of the structure of an air pollution control model;

[0028] Figure 7 Flow diagram of determining a model loss function in some embodiments of a method for training an air pollution control model according to the present disclosure;

[0029] Figure 8 Scatter plot for validating a test set for validating an air pollution control model;

[0030] Figure 9 PM2.5 concentration response validation graph for validating an air pollution control model;

[0031] Figure 10 Flow diagram of some embodiments of a method for predicting air pollution information according to the present disclosure;

[0032] Figure 11 Module schematic diagram of some embodiments of a training device for an air pollution control model according to the present disclosure;

[0033] Figure 12 Module schematic diagram of some other embodiments of a training device for an air pollution control model according to the present disclosure;

[0034] Figure 13 Module schematic diagram of some embodiments of an air pollution information prediction device according to the present disclosure;

[0035] Figure 14 [[ID=5(4)]]Module schematic diagram of some embodiments of an electronic device according to the present disclosure. Detailed implementation manners

[0036] Exemplary embodiments of the present disclosure will be described below with reference to the accompanying drawings. For clarity and conciseness, not all features of the embodiments are described in the specification. However, it should be understood that many implementation-specific settings must be made during the implementation of the embodiments in order to achieve the specific goals of the developer, for example, to comply with those restrictions related to the device and the business, and these restrictions may vary with different implementations. In addition, it should also be understood that although the development work may be very complex and time-consuming, for those skilled in the art who benefit from the present disclosure, such development work is only a routine task.

[0037] It should be noted that: Unless otherwise specifically stated, the relative arrangements, numerical expressions, and numerical values of the components and steps set forth in these embodiments do not limit the scope of the present disclosure.

[0038] Those skilled in the art can understand that terms such as "first", "second", etc. in the embodiments of the present disclosure are only used to distinguish different steps, devices, or modules, etc., and neither represent any specific technical meaning nor indicate an inevitable logical order between them.

[0039] It should also be understood that in the embodiments of the present disclosure, "a plurality of" may refer to two or more, and "at least one" may refer to one, two, or more.

[0040] It should also be understood that for any component, data, or structure mentioned in the embodiments of the present disclosure, without clear limitation or contrary indication in the context, it can generally be understood as one or more.

[0041] In addition, the term "and / or" in the present disclosure is merely a description of the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in the present disclosure generally represents an "or" relationship between the associated objects before and after.

[0042] It should also be understood that the present disclosure emphasizes the differences between the various embodiments, and their similarities or similarities can be referred to each other. For the sake of brevity, they will not be described one by one.

[0043] At the same time, it should be understood that for the sake of description, the dimensions of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0044] The following description of at least one exemplary embodiment is actually only illustrative and in no way limits the present disclosure or its application or use.

[0045] Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, such technologies, methods, and devices should be regarded as part of the specification.

[0046] It should be noted that like reference numerals and letters refer to like items in the following figures, and thus, once an item is defined in one figure, further discussion thereof is not required in subsequent figures.

[0047] In addition, to avoid obscuring the present disclosure with unnecessary details, only the processing steps and / or device structures closely related to at least the solution according to the present disclosure are shown in the figures, while other details less relevant to the present disclosure are omitted. It should also be noted that like reference numerals and letters in the figures indicate like items, and thus once an item is defined in one figure, further discussion thereof is not required for subsequent figures.

[0048] Figure 1 It is a schematic flowchart of some embodiments of a method for training an air pollution control model according to the present disclosure. As Figure 1 shown, the method for training an air pollution control model includes steps S101 - S105.

[0049] Step S101, using an air quality model, obtain first pollutant prediction information according to meteorological condition information and pollutant emission samples.

[0050] The air quality model can be various models, such as the CMAQ model, etc. The meteorological condition information can be obtained by various methods. For example, the WRF (Weather Research and Forecasting Model) model can be used to obtain the meteorological condition information. Various methods can be used to obtain pollutant emission samples according to pollutant emission inventories, etc. Input the meteorological condition information, pollutant emission samples, etc. into the air quality model to obtain the first pollutant prediction information output by the air quality model.

[0051] Step S102, generate training samples, where the training samples include meteorological condition information, pollutant emission samples, and historical state pollutant information, and the label of the training samples is the first pollutant prediction information.

[0052] The pollutant information of the historical state is data and characteristics related to pollutants within a specific past time period. The data includes pollutant concentrations, etc. For example, the pollutant information of the historical state includes the pollutant concentration information within a specific past time period. The pollutant information of the historical state can be used for model learning, understanding pollutant behavior patterns, variation laws, and interactions, etc. Multiple training samples can be generated based on meteorological condition information, pollutant emission samples, and the pollutant information of the historical state to form a training sample set. The first pollutant prediction information obtained using an air quality model and corresponding to the meteorological condition information and pollutant emission samples can be used as the label information for the training samples.

[0053] Step S103: Using the air pollution control model, obtain the second pollutant prediction information according to the training samples.

[0054] The air pollution control model can be models in various forms. For example, the air pollution control model is a convolutional neural network model, a generative adversarial network model, etc. Input the training samples into the air pollution control model to obtain the second pollutant prediction information output by the air pollution control model.

[0055] The pollutants in the pollutant information of the historical state, the first pollutant prediction information, and the second pollutant prediction information include PM2.5, oxides, etc. The pollutant information of the historical state, the first pollutant prediction information, and the second pollutant prediction information can include the concentrations and concentration change amounts of PM2.5 and oxides, as well as information such as the components of PM2.5 and the types of oxides. PM2.5 refers to particulate matter in the atmosphere with a diameter less than or equal to 2.5 micrometers. Oxides can be various oxides. The components of PM2.5 can include sulfates, nitrates, ammonium salts, organic particulate matter, secondary organic particulate matter, black carbon, and other particulate matter, etc. The types of oxides include ozone, nitrogen oxides, and atmospheric free radicals, etc.

[0056] Step S104: Determine the model loss function according to the first pollutant prediction information and the second pollutant prediction information.

[0057] Step S105: Perform adjustment processing on the air pollution control model according to the model loss function.

[0058] Multiple training methods can be used to train the air pollution control model. For example, determine the model loss function according to the first pollutant prediction information obtained using the air quality model and the second pollutant prediction information obtained using the air pollution control model. Perform iterative training on the air pollution control model according to the model loss function, adjust the parameters of the air pollution control model according to the function value of the model loss function, and gradually reduce the function value of the model loss function until the function value of the model loss function is less than a preset threshold to obtain a trained air pollution control model.

[0059] The training method of the air pollution control model of the present disclosure uses an air quality model to obtain first pollutant prediction information, generates training samples based on meteorological condition information, pollutant emission samples, and pollutant information of historical states, and the label of the training sample is the first pollutant prediction information, so as to use the air pollution control model and obtain second pollutant prediction information according to the training samples; according to the first pollutant prediction information and the second pollutant prediction information, a model loss function is determined for adjusting and processing the air pollution control model; through the trained air pollution control model, predictions of pollutants such as PM2.5 and pollutant emission control can be carried out, which can significantly reduce the demand for computing resources, improve computing efficiency, provide a basic model tool and scientific decision-making support for pollution control, improve the accuracy of pollutant prediction, and can be applied to fields such as environmental management, policy formulation, and pollution prevention and control, improving the scientificity and accuracy of pollution control.

[0060] A meteorological model can be used to obtain meteorological condition information based on meteorological reanalysis data, observation assimilation data, and terrain data; the meteorological model can be a WRF model or the like.

[0061] In some embodiments, simulation nesting covering regions such as China can be set, and range parameters such as the projection method, grid resolution, and vertical layer number of the simulation region can be determined. The setting parameters of the simulation nesting are shown in Table 1 below:

[0062] Projection method Lambert conformal conic projection Horizontal resolution 36 km × 36 km Number of grids 127×172 Number of vertical levels 23 (meteorology); 14 (chemistry) Number of nestings 1

[0063] Table 1 - Table of setting parameters for simulation nesting

[0064] Meteorological conditions can be simulated within the simulation nesting determined based on the parameters in Table 1 using meteorological analysis data such as global meteorological reanalysis data NCEP-FNL (as initial and boundary conditions), terrain data such as USGS data, etc. The WRF model sets a variety of parameterization schemes for different physical processes, such as cloud microphysics modules, boundary layer modules, long and short wave radiation modules, etc. The simulation accuracy of specific meteorological conditions varies greatly among different combinations of physical schemes. Based on synoptic principles, the parameterization scheme combination of the WRF model is optimized, and the parameterization scheme combination of this simulation region is shown in Table 2 below:

[0065] Parameterization scheme Scheme name Shortwave radiation Goddard shortwave scheme Longwave radiation RRTM scheme Surface layer Pleim-Xiu surface layer Boundary layer ACM2 (Pleim) PBL Cumulus convection Kain-Fritsch (new Eta) scheme Microphysics

[0066] Table 2 - Table of parameterization scheme combinations of the WRF model

[0067] When the WRF model simulates the meteorological field, there are usually systematic errors. For example, it is difficult to consider the weakening effect of surface topography on wind speed. To improve the simulation accuracy of the WRF model, a series of assimilation modules are added, and the simulation process of the WRF model is constrained by using observation assimilation data such as NCEP - OBS data. The assimilation information settings of the WRF model are shown in Table 3:

[0068] WSM 6-class graupel scheme Assimilation module Assimilated variables Sea surface temperature update Sea surface temperature Analysis assimilation Temperature, specific humidity, wind field (wind field only in the boundary layer) Observation assimilation Integrated layer temperature, humidity, wind field Soil assimilation

[0069] Table 3 - Configuration Table of Assimilation Information of WRF Model

[0070] After configuring the WRF model, using the WRF model and based on meteorological analysis data such as global meteorological re - analysis data NCEP - FNL, observation assimilation data such as NCEP - OBS data, and terrain data such as USGS data, meteorological condition information is obtained. The meteorological condition information includes temperature, pressure, humidity, wind field, cloud characteristics, precipitation, radiation, etc. for the target area.

[0071] Soil temperature, soil moisture, surface wind field It is a schematic flow chart for obtaining pollutant emission samples in some embodiments of the training method of the air pollution control model according to the present disclosure. The pollutant emission samples include a baseline scenario emission sample and a regulated scenario emission sample, as Figure 2 shown:

[0072] Step S201, sample the anthropogenic source pollutant emission information in the atmospheric emission information to obtain a baseline scenario emission sample.

[0073] Step S202, obtain atmospheric emission regulation information according to the anthropogenic source pollutant emission information in the atmospheric emission information.

[0074] Step S203, sample the atmospheric emission regulation information to obtain a regulated scenario emission sample;

[0075] Step S204, after sampling and regulation, merge the natural source pollutant emission inventory in the atmospheric emission information.

[0076] In some embodiments, the atmospheric emission information can be information obtained from multiple sources. For example, the atmospheric emission information includes an anthropogenic source emission inventory, etc. The anthropogenic source emission inventory is a summary of various atmospheric pollutant emission information generated by human activities. The natural source emission inventory is the collation and statistics of atmospheric pollutant emission information generated by natural processes. After sampling and regulation, the natural source pollutant emission inventory is merged into the atmospheric emission information to generate new atmospheric emission information.

[0077] The MEIC-TOOL written in JAVA language and the MEIC2CTM tool written in PYTHON language can be used to allocate the total emission inventory of China calculated by the MEIC (Multi-resolution Emission Inventory for China) and MEIC-HR (Multiresolution Emission Inventory for China-High Resolution) models to hourly grid data according to specific spatio-temporal parameters, and couple with the MIX anthropogenic emission inventory in East Asia. The key parameters of emission inventories such as the total emission inventory of China and the MIX anthropogenic emission inventory are shown in Table 4 below:

[0078]

[0079]

[0080] Table 4 - Key Parameter Table of Emission Inventory

[0081] The MCIP (Meteorology-Chemistry Interface Processor) module can be used to process the meteorological field simulation results output by the WRF model; calculate the natural source emission factors using MODIS-LAI leaf area index, PFT vegetation type data, etc., input the meteorological conditions output by the MCIP module into the MEGAN (Model of Emissions of Gases and Aerosols) model, calculate the natural source emissions corresponding to the hourly meteorological conditions, and obtain the natural source emission inventory. The anthropogenic emission inventory and the natural source emission inventory can be combined to obtain atmospheric emission information such as the regional total emission inventory.

[0082] The initial condition ICON model and the boundary condition BCON module of the CMAQ model adopt the default data of the model. The atmospheric chemistry transport module CCTM of the CMAQ model uses the optimized chemical mechanism and online emissions. The output results of CCTM include various important pollutants, free radicals, and atmospheric chemical intermediates. The post-processing module is used to output the PM10 concentration, PM2.5 and its component concentrations, the daily maximum eight-hour ozone concentration, the daily maximum one-hour concentration, the NO2 concentration, the SO2 concentration, etc. The configuration table of the CMAQ model is shown in Table 5 below:

[0083] Figure 2 Configuration name Specific settings Gas phase chemical mechanism CB05 Liquid phase chemical mechanism RADM Aerosol module AERO6 Aerosol thermodynamics ISORROPIA Cloud module ACM On-line emissions

[0084] Table 5 - Configuration Table of CMAQ Model

[0085] Based on the above-set simulation range, meteorological condition information, atmospheric emission information, etc., pollutant concentrations in the baseline scenario can be obtained using the CMAQ model. A meteorology-emission-air quality model system can be built on a high-performance computing platform to support parallel simulations of emission control scenarios.

[0086] In some embodiments, the anthropogenic source pollutant emission information in the atmospheric emission information can be regulated according to emission control factors, as well as spatial and temporal dimensions, to obtain atmospheric emission regulation information.

[0087] For example, multiple regulation scenarios for model training can be set. For a regulation scenario, a set of emission control factors (Intervention Factors, IFs) and their applicable spatio-temporal scales (spatial and temporal dimensions) can be set. The control factor ranges from 0 to 1. The control factor is defined to characterize the regulation degree of emissions of various pollutants such as sulfur dioxide, nitrogen oxides, non-methane volatile organic compounds, etc. The control factor is set between 0 (completely shutting down emissions) and 1 (no regulation, remaining consistent with the baseline state).

[0088] IFs are used to regulate the anthropogenic source pollutant emission information in the atmospheric emission information under the baseline state of the model, thereby generating different atmospheric emission regulation information. The spatial dimension includes at least one of the dimensions of country, province, and geographical grid. For the country dimension, the same regulation factor is used for the national grid; for the provincial dimension, the same regulation factor is used for the provincial grid; for the geographical grid dimension, heterogeneous regulation factors are used for all grids within China. The temporal dimension includes at least one of the dimensions of year, season, month, and day. In the spatial dimension, a three-level scale of country-province-geographical grid can be used for regulation.

[0089] In some embodiments, the atmospheric emission regulation information is sampled according to the spatial dimension, temporal dimension, and sampling dimension to obtain regulation scenario emission samples; the atmospheric emission information is sampled according to the spatial dimension, temporal dimension, and sampling dimension to obtain baseline scenario emission samples. The sampling dimension includes at least one of the dimensions of pollution source and pollutant type. The pollution source dimension includes at least one of the dimensions of electricity, industry, and other industries. The pollutant type dimension includes at least one of the dimensions of sulfur dioxide, nitrogen oxides, non-methane volatile organic compounds, ammonia, and particulate matter. The regulation scenario emission samples and the baseline scenario emission samples can include pollutant emission information with temporal and spatial characteristics, and the pollutants can include one or more of sulfur dioxide, nitrogen oxides, non-methane volatile organic compounds, ammonia, particulate matter, etc.

[0090] For example, a variety of algorithms such as the Sobol algorithm can be used to calculate the emission regulation tensor according to the spatial dimension, temporal dimension, and sampling dimension, and act on the atmospheric emission regulation information (adopting space) for random sampling to obtain a large-scale regulated scenario emission sample. The multi-dimensional cube structure for sampling is as Dust emissions and sea salt aerosols are turned on, lightning emissions are turned off shown.

[0091] The pollution source dimension can be three dimensions such as electricity, industry, and other industries, and the pollutant type dimension includes five dimensions such as sulfur dioxide, nitrogen oxides, non-methane volatile organic compounds, ammonia, and particulate matter as pollutant types. According to the pollution source dimension and the pollutant type dimension, a 15-dimensional independent sampling space can be constructed to generate regulated scenario emission samples.

[0092] Special regulated scenario samples can be obtained. By setting the number and types of dimensions included in the pollutant type dimension, it is possible to regulate the emissions of nitrogen oxides and non-methane volatile organic compounds with stronger non-linearity, and obtain regulated scenario emission samples for regulation scenarios such as single emission source emission regulation.

[0093] It is possible to combine special cases in policy evaluation to formulate targeted regulated scenario emission samples; increase the regulated scenarios for the gradual reduction of NO x and NMVOCs to improve the learning ability of the non-linear response of PM2.5 and its components concentration, and increase the regulated scenarios for reducing emissions only for a certain pollution source to improve the adaptability to actual policy applications.

[0094] According to the spatial dimension, temporal dimension, and sampling dimension, the quasi-random sampling technique of Sobol high-dimensional sampling can be used to perform automated sampling in the 15-dimensional sampling space in the atmospheric emission information, efficiently obtain the emission perturbation tensor, and act on the baseline scenario emission sample.

[0095] Input the obtained baseline scenario emission sample and regulated scenario emission sample, as well as the meteorological condition information obtained through the WRF model, into the CMAQ model to simulate the baseline scenario and a large number of regulated scenarios in parallel, and obtain the first pollutant prediction information output by the CMAQ model. Environmental monitoring data can be used to test the pollutant simulation accuracy corresponding to the baseline scenario emission sample. For the WRF model - CMAQ model, before each annual / monthly simulation, a warm-up time of a preset number of days such as 17 days is added to eliminate the influence of the initial conditions on the simulation results (the first pollutant prediction information).

[0096] In some embodiments, according to the heights of different pollution sources, the MEIC sub-sector lists (industry, power, agriculture, civil, transportation) can be combined into industrial sources, power sources, and other sources to determine the pollution source dimension. The pollutant dimension of the sampling space can be determined in combination with the controlled species of the emission control policy. For example, according to different emission policies, etc., the sampled pollutants are determined to be sulfur dioxide, nitrogen oxides, ammonia, non-methane volatile organic compounds, particulate matter, etc. Since PM 2.5 related pollutants usually use the same end-treatment equipment, so for primary PM 2.5 , black carbon (BC), organic carbon (OC), PM 10 maintain the same dimensional perturbation.

[0097] Considering that the responses of PM 2.5 and its components to emission regulation are different in different seasons, the four months of January, April, July, and October with different emission amounts and meteorological conditions are used to represent different seasons (time dimension). The three-level spatial scales of the national scale, provincial scale, and grid scale (spatial dimension) are used for regulation to characterize the spatial heterogeneity of emission regulation measures.

[0098] Figure 3 It is a schematic flow chart of obtaining the second pollutant prediction information in some embodiments of the training method of the air pollution regulation model according to the present disclosure, as Figure 4 shown:

[0099] Step S501, generate meteorological variable input data according to the meteorological condition information.

[0100] Step S502, generate emission variable data according to the pollutant emission samples.

[0101] Step S503, generate historical state background data according to the pollutant information of the historical state.

[0102] Step S504, use the air pollution regulation model to obtain pollutant characteristic information according to the meteorological variable input data and the emission variable data. The pollutant characteristic information includes pollutant concentration change information.

[0103] Step S505, add the pollutant concentration change information and the historical state background data to obtain the pollutant concentration information.

[0104] Step S506, obtain the second pollutant prediction information, where the second pollutant prediction information includes pollutant concentration information and pollutant concentration change information; the second pollutant prediction information may also include the components or types of pollutants.

[0105] In some embodiments, for the air pollution control model, the intervention state of the model is defined, including the emission baseline scenario and the emission control scenario; the emission baseline scenario is the actual emission scenario, and the emission control scenario is the scenario regulated based on the baseline scenario. The emission input without regulation is the baseline scenario, and the obtained multi-dimensional emission perturbation tensor is used to regulate the baseline scenario, which is the control scenario.

[0106] The pollutant information included in the baseline scenario emission samples and the control scenario emission samples is the main anthropogenic emissions pollutants, including PM2.5, black carbon (BC), organic carbon (OC), particulate matter with an aerodynamic diameter between 2.5 μm and 10 μm (PMcoarse), NMVOC, NH3, NO x and SO2, etc.; the pollutants in the baseline scenario emission samples and the control scenario emission samples cover power, industry and other sources. The pollutant information of the baseline scenario emission samples and the control scenario emission samples can adopt a vertical aggregation scheme to obtain the daily emissions and unify the spatial two-dimensional format. Extract the variables output by the WRF model, which mainly describe temperature, air pressure, humidity, wind field, cloud characteristics, precipitation, radiation, heat flux and underlying surface characteristics, etc.; the vertical layers of the 3D variables of the CMAQ model can be flattened to the feature dimension to obtain unified two-dimensional meteorological variables (meteorological condition information).

[0107] The input data of the air pollution control model is set to include meteorological condition information, pollutant emission samples (baseline scenario emission samples, control scenario emission samples), pollutant information of the historical state (pollutant concentration and oxidant of the historical state), etc.; the pollutant information of the historical state can be the pollutant information of years such as 2017, which is obtained by the WRF-CMAQ model based on the meteorological conditions of the intervention scenario and the fixed anthropogenic emissions in 2017. The pollutant information of the historical state is only related to the meteorological data of the intervention scenario within a limited year range, so they have been prepared in advance and integrated into the model. When using the model, the user only needs to ensure that the input defining the baseline state of the model matches the meteorological data of the selected intervention scenario.

[0108] The min-max normalization method is used to normalize all input variables (test samples and test samples, etc.) of the air pollution control model to between 0 and 1; a data generator is constructed to batch generate the input features of the model.

[0109] The generated training samples and test samples include meteorological condition information, pollutant emission samples, and pollutant information of historical states, etc.; the labels of the training samples and test samples are the corresponding first pollutant prediction information. The ratio of the number of test samples to the number of training samples is 4:1. The stratified random splitting method can be used to ensure that the emission samples of different types of regulation scenarios are evenly distributed in both the test sample set and the training sample set.

[0110] Figure 4 FIG. is a schematic flow chart of obtaining the second pollutant prediction information in some embodiments of the training method of the air pollution regulation model according to the present disclosure. The air pollution regulation model includes a 3D convolutional layer, an activation function, a residual module, an encoder, a decoder, and a 2D convolutional layer, etc., as Figure 5 shown:

[0111] Step S601, use the 3D convolutional layer to perform 3D spatial feature extraction on the meteorological variable input data and the emission input data.

[0112] Step S602, use the activation function to perform activation processing on the output data of the 3D convolutional layer. Among them, a skip connection is established between the input end and the output end of the 3D convolutional layer through the residual module.

[0113] Step S603, use the encoder to perform downsampling and pooling processing on the output data of the activation function.

[0114] Step S604, use the decoder to perform upsampling processing on the output data of the encoder. The encoder includes a 3D convolutional layer and a pooling layer, and the decoder includes multiple transposed convolutional layers. A skip connection is established between each layer of the encoder and the corresponding layer of the decoder.

[0115] Step S605, use the 2D convolutional layer to process the output data of the encoder to obtain pollutant feature information. The number of 2D convolutional layers is multiple, and the pollutant feature information includes pollutant concentration change information.

[0116] Step S606, add the pollutant concentration change information output by the 2D convolutional layer to the historical state background data to obtain pollutant concentration information.

[0117] Step S607, obtain the second pollutant prediction information. The second pollutant prediction information includes the pollutant concentration information and the pollutant concentration change information output by the 2D convolutional layer. The second pollutant prediction information may also include the components or types of pollutants.

[0118] The air pollution regulation model can be various models, such as a three-dimensional residual UNet network model, etc. As Figure 5As shown, the air pollution control model includes a 3D convolutional layer, an activation function, a residual module, an encoder, a decoder, and a 2D convolutional layer, etc.; multi-variable spatio-temporal features such as meteorology, emissions, pollutants, and oxidants are input into the air pollution control model, and the concentrations of pollutants and oxidants adopt the background information of historical states.

[0119] Use the initial 3D convolutional layer to extract 3D spatial features from the input data, and use the activation function to increase the non-linear expression ability; build a residual connection, add a skip connection between the input and output of the 3D convolutional layer to reduce the problem of gradient disappearance, improve the training stability of the model, and allow the network to learn the incremental changes between the input and output, making the model easier to optimize.

[0120] Use the encoder to perform downsampling, reduce the spatial size of the feature map through a series of 3D convolutional layers, extract high-level features, and further reduce the data dimension in combination with pooling. Use the decoder to perform upsampling, adopt the transposed convolution method to restore the feature map to the original size. Combine the skip connection to fuse the high-resolution and low-resolution feature information.

[0121] Use a series of 2D convolutional layers to convert the features into 2D variables and output the pollutant concentration changes, use the summation connection for feature aggregation, superimpose the regulated pollutant concentration changes on the historical pollutant concentration to obtain the regulated pollutant concentration information. By adjusting the hyperparameters of each module of the air pollution control model, the optimal model structure configuration can be obtained.

[0122] Figure 6 It is a schematic flow chart for determining the model loss function in some embodiments of the training method of the air pollution control model according to the present disclosure. Both the first pollutant prediction information and the second pollutant prediction information include pollutant concentration information and pollutant concentration change information. The adaptive weights include concentration adaptive weights and concentration change adaptive weights, as Figure 7 shown:

[0123] Step S801, determine the concentration loss function according to the pollutant concentration information in the first pollutant prediction information and the pollutant concentration information in the second pollutant prediction information.

[0124] Step S802, determine the concentration change loss function according to the pollutant concentration change information in the first pollutant prediction information and the pollutant concentration change information in the second pollutant prediction information.

[0125] Step S803: Determine the concentration accuracy loss function according to the concentration loss function, the concentration adaptive weight, and a preset constant.

[0126] Step S804: Determine the concentration change accuracy loss function according to the concentration change loss function, the concentration change adaptive weight, and a preset constant.

[0127] Step S805: Determine the model loss function according to the concentration accuracy loss function, the concentration change accuracy loss function, the concentration adaptive weight, and the concentration change adaptive weight.

[0128] Multiple methods can be used to determine the model loss function according to the concentration accuracy loss function, the concentration change accuracy loss function, the concentration adaptive weight, and the concentration change adaptive weight. For example, determine the penalty information according to the concentration adaptive weight and the concentration change adaptive weight; determine the model loss function according to the concentration accuracy loss function, the concentration change accuracy loss function, the concentration change adaptive weight, the concentration change adaptive weight, and the penalty information.

[0129] In some embodiments, the first pollutant prediction information generated by the CMAQ model (WRF-CMAQ model) includes pollutant concentration information and pollutant concentration change information, and the second pollutant prediction information generated by the air pollution control model of the present disclosure includes pollutant concentration information and pollutant concentration change information; the pollutants include PM2.5 and oxides, etc. The components of PM2.5 include seven types such as sulfate, nitrate, ammonium salt, organic particulate matter, secondary organic particulate matter, black carbon, and other particulate matters, and the oxides include three types, that is, the total number of components or types of pollutants is 10.

[0130] Determine the concentration loss function as shown in Formula 1-1 according to the pollutant concentration information in the first pollutant prediction information of the pollutant and the pollutant concentration information in the second pollutant prediction information:

[0131]

[0132] Determine the concentration change loss function as shown in Formula 1-2 according to the pollutant concentration change information in the first pollutant prediction information of the pollutant and the pollutant concentration change information in the second pollutant prediction information:

[0133]

[0134] where, V m,n is used to calculate the weight value of each geographical grid, as shown in Formula 1-3:

[0135]

[0136] Among them, a geographical grid divides the Earth's surface into grid cells of equal size and regular shape according to certain rules and scales. The grid cells can be of different shapes such as squares, rectangles, hexagons, etc., and each grid has a unique identifier. For example, the grid cell (horizontal resolution) set according to the simulation nesting parameters is 36 km × 36 km.

[0137] By setting the geographical grid, a geographical grid matrix can be formed. The grid simulation domain is characterized by the geographical grid matrix. The row and column indices of the grids in the grid simulation domain are the row and column indices (row numbers and column numbers) of the geographical grids in the geographical grid matrix, and the grids in the grid simulation domain correspond to the grid cells in the geographical grid matrix.

[0138] In Formulas 1-1 to 1-3, C represents concentration, and ΔC represents the change in concentration; m and n respectively represent the row and column indices of the grids in the grid simulation domain (geographical grid matrix), and M and N represent the total number of rows and columns in the grid simulation domain (geographical grid matrix); AI represents that the data source is the air pollution control model, and WRF-CMAQ represents that the data source is the CMAQ model; represents the pollutant concentration information in the second pollutant prediction information of the pollutants in the grids in the grid simulation domain, represents the pollutant concentration information in the first pollutant prediction information of the pollutants in the grids in the grid simulation domain; represents the pollutant concentration change information in the second pollutant prediction information of the pollutants in the grids in the grid simulation domain, represents the pollutant concentration change information in the first pollutant prediction information of the pollutants in the grids in the grid simulation domain; i is the component number of PM2.5 or the type number of oxides, and the value of i is 1, 2... 10. The components of PM2.5 include seven types such as sulfate, nitrate, ammonium salt, organic particulate matter, secondary organic particulate matter, black carbon, and other particulate matters, and the types of oxides include three types such as ozone.

[0139] According to the concentration loss function, the concentration adaptive weight, and a preset constant, a concentration accuracy loss function is determined to evaluate the accuracy of the concentration prediction task. The concentration accuracy loss function is shown in Formula 1-4 as follows:

[0140]

[0141] Among them, i is the component number of PM2.5 or the type number of oxides, and σ 1i is the concentration adaptive weight of the i-th component of PM2.5 or the type of oxides; the concentration adaptive weight is the weight used in the air pollution control model to automatically adjust the weight distribution according to different concentration-related factors; ε is a preset constant.

[0142] Determine the concentration change accuracy loss function based on the concentration change loss function, the concentration change adaptive weight, and a preset constant, which is used to evaluate the accuracy of the concentration change prediction task. The concentration change accuracy loss function is shown in the following formulas 1-5:

[0143]

[0144] σ 2i is the concentration change adaptive weight for the i-th type of PM2.5 component or oxide; the concentration change adaptive weight is a weight used in the air pollution control model to automatically adjust the weight distribution according to different factors related to concentration changes.

[0145] Design the total loss function Ltotal consisting of three terms, including: the loss function Lconc for evaluating the accuracy of the concentration prediction task, the loss function LΔconc for evaluating the accuracy of the concentration change prediction task, and the logarithmic penalty term; use the total loss function Ltotal as the model loss function, as shown in formula 1-6:

[0146] as shown in formula 1-6:

[0147]

[0148] where the subscripts "conc" and "Δconc" represent the terms of concentration and concentration change respectively; σ is the adaptive weight automatically learned by the air pollution control model, where σ conc is the concentration adaptive weight, is the concentration change adaptive weight; ε is a small constant, such as 0.01, and ε is used as a constraint to prevent the denominator from approaching zero too closely.

[0149] Use the total loss function (model loss function) as the objective to carry out supervised learning, dynamically adjust various parameters of the air pollution control model, and obtain the optimal air pollution control model.

[0150] In some embodiments, multiple methods can be used to evaluate the performance of the air pollution control model. For example, obtain test samples, where the test samples include meteorological condition information, pollutant emission samples, and pollutant information of historical states, and the label of the test samples is the first pollutant prediction information; verify the trained air pollution control model according to the test samples. The verification results using the training sample set are as shown Figure 7 where the AI model is the air pollution control model of the present disclosure.

[0151] The air pollution control model can be deduced according to the test samples, and the model prediction results can be comprehensively compared with the simulated true values of the WRF-CMAQ model in various control scenarios, and verified in different regions, different time periods, different particulate components, and different concentration levels.

[0152] Several application scenarios can be constructed according to the actual environmental decision-making requirements to evaluate whether the air pollution control model can support control applications; design three different intensity emission control scenarios at the urban scale during heavy pollution periods, and use the WRF-CMAQ model and the trained air pollution control model to predict the pollutant prediction information corresponding to the three control scenarios respectively, evaluate whether the air pollution control model can support practical applications in environmental fields such as pollution emergency management and control, and the verification results are as Figure 8 shown. The AI model is the air pollution control model of the present disclosure.

[0153] The training method of the air pollution control model in the above embodiments can be used to predict pollutants such as PM2.5 and control pollutant emissions through the trained air pollution control model, which can significantly reduce the demand for computing resources, improve computing efficiency, provide a basic model tool and scientific decision-making support for pollution control, improve the accuracy of pollutant prediction, and can be applied to environmental management, policy formulation, pollution prevention and control and other fields, improving the scientificity and accuracy of pollution control.

[0154] Figure 9 It is a schematic flowchart of some embodiments of the air pollution information prediction method according to the present disclosure, as Figure 10 shown:

[0155] Step S1101, generate model input data according to meteorological condition information, atmospheric emission information and pollutant information of historical states.

[0156] A variety of methods can be used to generate meteorological condition information. For example, meteorological models such as the WRF model can be used to obtain meteorological condition information according to meteorological analysis data such as meteorological reanalysis data, observation assimilation data and terrain data.

[0157] A variety of methods can be used to obtain atmospheric emission information, and the atmospheric emission information can include control scenario emission samples, etc. For example, merge the anthropogenic source pollutant emission inventory and the natural source pollutant emission inventory to obtain atmospheric emission information; sample the anthropogenic source pollutant emission information in the atmospheric emission information to obtain the baseline scenario emission sample; obtain the atmospheric emission control information according to the anthropogenic source pollutant emission information in the atmospheric emission information; sample the atmospheric emission control information to obtain the control scenario emission sample.

[0158] Step S1102, use the air pollution control model to obtain pollutant prediction information according to the model input data.

[0159] The air pollution control model is trained by the training method of any of the above embodiments. The pollutant prediction information output by the air pollution control model includes the concentrations and concentration changes of PM2.5 and oxides, as well as the components of PM2.5 and the types of oxides, etc.

[0160] The air pollution information prediction method of the present disclosure can increase the simulation speed of the air pollution control model to the minute level, support the simulation optimization of the continuous improvement path of PM2.5 air quality refined to the geographical grid at the ten-thousand-group level and the optimization of the heavy pollution emergency control plan, and can be widely applied to fields such as environmental management, policy formulation, and pollution prevention and control. It can serve research institutions to formulate refined pollution control strategies and optimize the air quality improvement path; government ecological environment protection agencies can use the air pollution control model to conduct PM2.5 pollution control scenario simulation and policy evaluation, which helps to improve the scientificity and accuracy of PM2.5 pollution control in China and promotes the development of intelligent pollution control.

[0161] In some embodiments, as Figure 10 shown, the present disclosure provides a training device 1200 for an air pollution control model, including a first data acquisition module 1201, a sample generation module 1202, a second data acquisition module 1203, a loss determination module 1204, and a model adjustment module 1205.

[0162] The first data acquisition module 1201 uses an air quality model to obtain first pollutant prediction information according to meteorological condition information and pollutant emission samples. The sample generation module 1202 generates training samples, and the training samples include meteorological condition information, pollutant emission samples, and pollutant information of historical states. The label of the training samples is the first pollutant prediction information.

[0163] The second data acquisition module 1203 uses the air pollution control model to obtain second pollutant prediction information according to the training samples. The loss determination module 1204 determines a model loss function according to the first pollutant prediction information and the second pollutant prediction information. The model adjustment module 1205 adjusts the air pollution control model according to the model loss function until the function value of the model loss function is less than a preset threshold.

[0164] As Figure 11 shown, the present disclosure provides a training device 1200' for an air pollution control model. In addition to including all the modules of the training device 1200 for an air pollution control model, it further includes a scenario information generation module 1206 and a meteorological condition generation module 1207.

[0165] The pollutant emission samples include the baseline scenario emission samples and the regulated scenario emission samples. The scenario information generation module 1206 merges the anthropogenic source pollutant emission inventory and the natural source pollutant emission inventory to obtain the atmospheric emission information; the scenario information generation module 1206 samples the anthropogenic source pollutant emission information in the atmospheric emission information to obtain the baseline scenario emission samples; the scenario information generation module 1206 obtains the atmospheric emission regulation information according to the anthropogenic source pollutant emission information in the atmospheric emission information; the scenario information generation module 1206 samples the atmospheric emission regulation information to obtain the regulated scenario emission samples.

[0166] For example, the scenario information generation module 1206 adjusts and processes the anthropogenic source pollutant emission information in the atmospheric emission information according to the emission regulation factors, as well as the spatial dimension and the time dimension, to obtain the atmospheric emission regulation information.

[0167] The scenario information generation module 1206 samples the atmospheric emission regulation information according to the spatial dimension, the time dimension, and the sampling dimension to obtain the regulated scenario emission samples; the scenario information generation module 1206 samples the anthropogenic source pollutant emission information in the atmospheric emission information according to the spatial dimension, the time dimension, and the sampling dimension to obtain the baseline scenario emission samples. The sampling dimension includes the pollution source dimension and the pollutant type dimension. The pollution source dimension includes at least one dimension among the power, industrial, and other industries; the pollutant type dimension includes at least one dimension among sulfur dioxide, nitrogen oxides, non-methane volatile organic compounds, ammonia, and particulate matter.

[0168] The meteorological condition generation module 1207 is used to obtain the meteorological condition information by using a meteorological model according to the meteorological reanalysis data, the observation assimilation data, and the terrain data.

[0169] In some embodiments, as Figure 12 shown, the present disclosure provides an air pollution information prediction device 1400, including a data generation module 1401 and an information prediction module 1402.

[0170] The data generation module 1401 generates model input data according to the meteorological condition information, the atmospheric emission information, and the pollutant information of the historical state. The information prediction module 1402 uses an air pollution regulation model to generate pollutant prediction information according to the model input data.

[0171] As Figure 13 Figure 14 shown, the electronic device may include a memory 1501, a processor 1502, a communication interface 1503, and a bus 1504. The memory 1501 is used to store instructions. The processor 1502 is coupled to the memory 1501. The processor 1502 is configured to execute the training method of the above-mentioned air pollution regulation model or the air pollution information prediction method based on the instructions stored in the memory 1501.

[0172] The memory 1501 can be a high-speed RAM memory, a non-volatile memory, etc. The memory 1501 can also be a memory array. The memory 1501 may also be partitioned, and the partitions can be combined into virtual volumes according to certain rules. The processor 1502 can be a central processing unit CPU, or an application specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the training method of the air pollution regulation model or the air pollution information prediction method of the present disclosure.

[0173] In some embodiments, the present disclosure provides a computer-readable storage medium storing computer instructions, which when executed by a processor, implement the training method of the air pollution regulation model in any of the above embodiments.

[0174] The computer-readable storage medium can adopt any combination of one or more readable media. The readable media can be a readable signal medium or a readable storage medium. The readable storage medium can, for example, include but is not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the readable storage medium can include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0175] An embodiment of the present disclosure can also be a computer program product, which includes computer program instructions that, when run by a processor, cause the processor to execute the steps in the methods according to various embodiments of the present disclosure described in the "Exemplary Methods" section above of this specification.

[0176] The basic principles of the present disclosure have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present disclosure are only examples and not limitations. It cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present disclosure. In addition, the above-disclosed specific details are only for the purpose of illustration and facilitating understanding, and are not limitations. The above details do not limit the present disclosure to necessarily adopt the above specific details for implementation.

[0177] In the present specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments, and the same or similar parts among the various embodiments can be referred to each other. For system embodiments, since they basically correspond to method embodiments, the description is relatively simple, and the relevant parts can be referred to the corresponding descriptions in the method embodiments.

[0178] The block diagrams of the devices, apparatuses, equipment, and systems involved in the present disclosure are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open-ended words, meaning "including but not limited to", and can be used interchangeably with each other. The word "or" and "and" used herein refer to the word "and / or", and can be used interchangeably with each other, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to", and can be used interchangeably with each other.

[0179] It should also be noted that in the devices, equipment, and methods of the present disclosure, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent solutions of the present disclosure.

[0180] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0181] The above description has been given for purposes of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art should understand that the above embodiments are merely illustrative and do not limit the scope of the present disclosure. Those skilled in the art should understand that the above embodiments can be combined, modified, or replaced without departing from the scope and essence of the present disclosure.

Claims

1. A training method for an air pollution control model, comprising: Using an air quality model, obtaining first pollutant prediction information according to meteorological condition information and pollutant emission samples; Generating training samples, wherein the training samples include the meteorological condition information, the pollutant emission samples and pollutant information of historical states, and the label of the training samples is the first pollutant prediction information; Using an air pollution control model, obtaining second pollutant prediction information according to the training samples; Determining a model loss function according to the first pollutant prediction information and the second pollutant prediction information; Performing an adjustment process on the air pollution control model according to the model loss function.

2. The training method according to claim 1, wherein The step of using an air pollution control model to obtain second pollutant prediction information according to the training samples includes: Generating meteorological variable input data according to the meteorological condition information; Generating emission variable data according to the pollutant emission samples; Generating historical state background data according to the pollutant information of historical states; Using the air pollution control model to obtain pollutant feature information according to the meteorological variable input data and the emission variable data, wherein the pollutant feature information includes pollutant concentration change information; Using the air pollution control model to sum the pollutant concentration change information and the historical state background data to obtain pollutant concentration information; Obtaining the second pollutant prediction information, wherein the second pollutant prediction information includes the pollutant concentration information and the pollutant concentration change information.

3. The training method according to claim 2, wherein, The air pollution control model includes a 3D convolutional layer, an activation function, a residual module, an encoder, a decoder and a 2D convolutional layer; the step of using the air pollution control model to obtain pollutant feature information according to the meteorological variable input data and the emission variable data includes: Using the 3D convolutional layer to perform 3D spatial feature extraction on the meteorological variable input data and the emission input data; Using the activation function to perform activation processing on the output data of the 3D convolutional layer, wherein a skip connection is established between the input end and the output end of the 3D convolutional layer through the residual module; Using the encoder to perform downsampling and pooling processing on the output data of the activation function; Using the decoder to perform upsampling processing on the output data of the encoder, wherein the encoder includes a 3D convolutional layer and a pooling layer, the decoder includes a plurality of transposed convolutional layers, and a skip connection is established between each layer of the encoder and the corresponding layer of the decoder; Using the 2D convolutional layer to process the output data of the encoder to obtain the pollutant feature information, wherein the number of the 2D convolutional layers is multiple; The step of using the air pollution control model to sum the pollutant concentration change information and the historical state background data to obtain the pollutant concentration information includes: Summing the pollutant concentration change information output by the 2D convolutional layer and the historical state background data to obtain the pollutant concentration information.

4. The training method according to claim 1, wherein Both the first pollutant prediction information and the second pollutant prediction information include pollutant concentration information and pollutant concentration change information. The adaptive weights include a concentration adaptive weight and a concentration change adaptive weight. Determining the model loss function according to the first pollutant prediction information, the second pollutant prediction information, and the adaptive weights of the air pollution control model includes: Determining a concentration loss function according to the pollutant concentration information in the first pollutant prediction information and the pollutant concentration information in the second pollutant prediction information; Determining a concentration change loss function according to the pollutant concentration change information in the first pollutant prediction information and the pollutant concentration change information in the second pollutant prediction information; Determining a concentration accuracy loss function according to the concentration loss function, the concentration adaptive weight, and a preset constant; Determining a concentration change accuracy loss function according to the concentration change loss function, the concentration change adaptive weight, and the preset constant; Determining the model loss function according to the concentration accuracy loss function, the concentration change accuracy loss function, the concentration adaptive weight, and the concentration change adaptive weight.

5. The training method according to claim 4, wherein, The determining the model loss function according to the concentration accuracy loss function, the concentration change accuracy loss function, the concentration adaptive weight, and the concentration change adaptive weight includes: Determining penalty information according to the concentration adaptive weight and the concentration change adaptive weight; Determining the model loss function according to the concentration accuracy loss function, the concentration change accuracy loss function, the concentration change adaptive weight, the concentration change adaptive weight, and the penalty information.

6. The training method according to claim 1, wherein, The pollutant emission samples include a baseline scenario emission sample and a regulated scenario emission sample. The training method includes: Sampling the anthropogenic source pollutant emission information in the air emission information to obtain a baseline scenario emission sample; Obtaining air emission regulation information according to the anthropogenic source pollutant emission information in the air emission information; Sampling the air emission regulation information to obtain a regulated scenario emission sample.

7. The training method according to claim 6, wherein, The obtaining air emission regulation information according to the anthropogenic source pollutant emission information in the air emission information includes: Performing regulation processing on the anthropogenic source pollutant emission information in the air emission information according to an emission regulation factor, as well as spatial and temporal dimensions, to obtain the air emission regulation information.

8. The training method according to claim 7, wherein, The sampling the anthropogenic source pollutant emission information in the air emission regulation information to obtain a regulated scenario emission sample includes: Sampling the anthropogenic source pollutant emission information in the air emission regulation information according to the spatial dimension, the temporal dimension, and a sampling dimension to obtain the regulated scenario emission sample; The sampling the anthropogenic source pollutant emission information in the air emission information to obtain a baseline scenario emission sample includes: Sampling the anthropogenic source pollutant emission information in the air emission information according to the spatial dimension, the temporal dimension, and the sampling dimension to obtain the baseline scenario emission sample; Among them, the sampling dimensions include a pollution source dimension and a pollutant type dimension. The pollution source dimension includes at least one dimension among the dimensions of electricity, industry, and other industries; the pollutant type dimension includes at least one dimension among sulfur dioxide, nitrogen oxides, non-methane volatile organic compounds, ammonia, and particulate matter.

9. The training method according to claim 1, comprising: Using a meteorological model, obtaining the meteorological condition information according to meteorological reanalysis data, observation assimilation data, and terrain data.

10. The training method according to any one of claims 1 to 9, wherein The pollutant information of the historical state, the first pollutant prediction information, and the second pollutant prediction information include: the concentrations of PM2.5 and oxides, the concentration change amounts, and the PM2.5 components and oxide types.

11. An air pollution information prediction method, comprising: Generating model input data according to meteorological condition information, air emission information, and pollutant information of the historical state; Using an air pollution control model, obtaining pollutant prediction information according to the model input data; Among them, the air pollution control model is trained by the training method according to any one of claims 1 to 10.

12. An air pollution control model training device, comprising: A first data obtaining module, configured to use an air quality model to obtain first pollutant prediction information according to meteorological condition information and pollutant emission samples; A sample generation module, configured to generate training samples, wherein the training samples include the meteorological condition information, pollutant emission samples, and pollutant information of the historical state, and the label of the training samples is the first pollutant prediction information; A second data obtaining module, configured to use an air pollution control model to obtain second pollutant prediction information according to the training samples; A loss determination module, configured to determine a model loss function according to the first pollutant prediction information and the second pollutant prediction information; A model adjustment module, configured to perform adjustment processing on the air pollution control model according to the model loss function.

13. An air pollution information prediction device, comprising: A data generation module, configured to generate model input data according to meteorological condition information, air emission information, and pollutant information of the historical state; An information prediction module, configured to use an air pollution control model to generate pollutant prediction information according to the model input data; Among them, the air pollution control model is trained by the training method according to any one of claims 1 to 10.

14. An electronic device, comprising: A memory; And a processor coupled to the memory, the processor being configured to execute the method according to any one of claims 1-11 based on instructions stored in the memory.

15. A computer-readable storage medium, the computer-readable storage medium stores computer instructions, and when the instructions are executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

16. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images