PM2.5 and O3 combined simulation method, storage medium and equipment

Abstract

The invention relates to the technical field of urban air quality space-time fine mapping, in particular to a PM2.5 and O3 combined simulation method, a storage medium and equipment. The method comprises the following steps: determining an ENVI-met input index system; initial ENVI-met simulation data is obtained; acquiring a corrected data set of spatial and temporal distribution of air pollution in the target area; and constructing a PM2.5 and O3 combined simulation model based on the improved U-Net network, and importing the data set into the combined simulation model to predict and obtain a PM2.5 and O3 combined mapping result. Computer program instructions are stored on the storage medium, and the method is implemented when the instructions are executed by the processor. The apparatus includes at least one processor, at least one memory, and computer program instructions stored in the memory that when executed by the processor implement the method. According to the method, the PM2.5 and O3 combined drawing result with high resolution, high interpretation and high accuracy can be obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of urban air quality spatiotemporal fine mapping, and in particular to a PM 2.5 and O3 joint simulation method, storage medium and device. Background Art

[0002] At present, the analysis of air pollution characteristics at the building unit scale still mainly relies on fixed-site monitoring. This monitoring method has the advantages of high temporal resolution and high accuracy, but insufficient spatial resolution. In recent years, mobile air quality navigation technology based on vehicles or drones has gradually been promoted. This technology has the advantages of high spatial resolution and high accuracy, but it is difficult to cover a wide range and long time series. With the rapid development of computational fluid dynamics research and computer science, a variety of indirect simulation technologies such as numerical models and empirical statistics have gradually matured. Among them, the numerical simulation method that solves the partial differential equation of air pollutant diffusion and transmission has strong interpretability and can provide high spatiotemporal resolution data over a wide range and long time series; however, its high dependence on data such as emission inventories leads to limited accuracy, high computing power requirements, and fixed physical models. Relatively speaking, deep learning based on big data analysis has great advantages in computing speed and accuracy, and its network framework can be freely set, and it can couple multi-source data to carry out high spatiotemporal resolution air quality mapping research. However, its accuracy is affected by the number of samples, and due to the "black box" attribute of machine learning, the mechanism interpretability is often questioned.

[0003] In summary, it is necessary to develop a PM that combines the advantages of high temporal resolution, high spatial resolution, high interpretability and high accuracy. 2.5 and O3 joint simulation method, storage medium and device. Summary of the invention

[0004] The present invention aims to provide a PM 2.5 The specific technical solutions of the joint simulation method, storage medium and equipment with O3 are as follows:

[0005] In a first aspect, the present invention provides a PM 2.5 and O3 joint simulation methods, including:

[0006] Step S1, using encrypted fixed sites in the target area to obtain site encrypted monitoring data at actual measurement time points within a one-year period; combining the site encrypted monitoring data and the urban morphological data and geographical environment elements of the target area within a one-year period to form an ENVI-met input indicator system;

[0007] Step S2, importing the ENVI-met input index system into the ENVI-met simulation software to form a simulation file; the built-in solver in the ENVI-met simulation software performs ENVI-met iterative calculation on the simulation file, and outputs the solution result to obtain initial ENVI-met simulation data;

[0008] Step S3, using mobile cruise monitoring in the target area to obtain mobile cruise monitoring data at measured time points within a year, and preprocessing the mobile cruise monitoring data; using spatiotemporal matching to construct a modeling sample that includes the initial ENVI-met simulation data, the site encrypted monitoring data, the preprocessed mobile cruise monitoring data, and the geographical environment elements in a one-to-one correspondence; substituting the modeling sample into the linear mixed effects model LME to predict and obtain a data set of the corrected spatiotemporal distribution of air pollution in the target area; the data set includes PM at spatiotemporal points without measured data 2.5 Concentration values ​​and O3 concentration values ​​and PM at time and space points with measured data 2.5 concentration value and O3 concentration value;

[0009] Step S4: Use the U-Net network with multi-scale channel attention mechanism MS-CAM to construct PM 2.5 and O3 joint simulation model; importing the data set into the joint simulation model to obtain PM 2.5 Joint mapping results with O3.

[0010] Optionally, in step S1, the urban morphology data includes building height data and road traffic emission data; wherein the road traffic emission data is in the form of PM 2.5 Emissions and NO2 emission characterization;

[0011] The geographical environment factors include PM 2.5 Initial concentration data of, initial concentration data of O3, meteorological data, digital elevation model DEM data and time data.

[0012] Optionally, in step S2, the simulation file includes an ENVI-met.inx file and an ENVI-met.simx file;

[0013] The data sources forming the ENVI-met.inx file include the DEM data in the geographical environment elements and the urban morphology data;

[0014] The data sources forming the ENVI-met.simx file include the road traffic emission data in the urban morphology data, and the meteorological data and time data in the geographical environment elements.

[0015] Optionally, in step S3, the preprocessing of the mobile cruise monitoring data includes clustering the mobile cruise monitoring data into data with a resolution of 10-60 m per hour by using spatiotemporal clustering.

[0016] Optionally, in step S3, the linear mixed effects model LME is expressed by the following formula:

[0017] Y = Xβ + Zγ + ε;

[0018] Wherein, Y represents the dependent variable, which is the encrypted monitoring data of the site and the preprocessed mobile cruise monitoring data; X is the fixed effect design matrix composed of all independent variables, which are the ENVI-met simulation data and the digital elevation model DEM data and time data in the geographical environment elements; Z is the random effect design matrix composed of time information, which is the time data in the geographical environment elements; β is the coefficient of the fixed effect; γ is the coefficient of the random effect; ε is the random error term.

[0019] Optionally, the joint simulation model includes a coding stage, a PM 2.5 The MS-CAM is set at the end of the encoding stage; the MS-CAM includes two output ends, one of which is connected to the PM 2.5 The other output terminal is connected with the O3 decoding stage.

[0020] Optionally, the encoding stage includes four encoding blocks arranged in sequence; wherein, the first three encoding blocks each include 2 3*3 convolution operations and 1 2*2 maximum pooling operation, and the last encoding block includes 2 3*3 convolution operations and 1 MS-CAM operation.

[0021] Optionally, the PM 2.5 The decoding stage and the O3 decoding stage are symmetrically arranged at the two output ends of the MS-CAM, and both include four decoding blocks arranged in sequence; the four decoding blocks are arranged corresponding to the four encoding blocks in the encoding stage; wherein the first three decoding blocks each include one 2*2 transposed convolution and two 3*3 convolution operations, and the last decoding block includes one 1*1 convolution operation.

[0022] In a second aspect, the present invention provides a storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the PM 2.5 and O3 joint simulation method.

[0023] In a third aspect, the present invention provides a device comprising: at least one processor, at least one memory and computer program instructions stored in the memory, which when executed by the processor implement the PM 2.5 and O3 joint simulation method.

[0024] The application of the technical solution of the present invention has at least the following beneficial effects:

[0025] The present invention provides a PM 2.5 The O3 joint simulation method, storage medium and equipment couple multiple air pollution monitoring simulation methods in steps S1-S4, including encrypted fixed sites, mobile cruise monitoring, ENVI-met simulation and U-Net network deep learning with multi-scale channel attention mechanism MS-CAM, to obtain PM with high temporal resolution, high spatial resolution, high interpretability and high accuracy. 2.5 and O3 joint mapping results. Specifically, the present invention adopts a combination of steps S1-S3, that is, through long-term encrypted fixed sites and mobile cruise monitoring, ENVI-met simulation data is post-corrected to obtain a high-precision data set of spatiotemporal distribution of air pollution; the multi-scale channel attention mechanism MS-CAM is introduced into the joint simulation model of step S4 to enhance the learning and integration capabilities of features at different scales; the U-Net network, as a type of deep convolutional neural network CNN, can first merge images and then use convolution and transposed convolution to extract feature information, which can not only provide more detailed results than other methods, but also has the advantages of fast speed and high accuracy; the joint simulation model is effective for urban PM 2.5 The temporal and spatial differences of O3 concentrations were explained by 80% and 86% respectively, which can be used to analyze PM2.5 and PM3 at the building unit scale. 2.5 and O3 concentrations; importing the dataset into the joint simulation model to obtain PM 2.5 Joint mapping results with O3.

[0026] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0028] Figure 1 is a schematic diagram of the structure of a joint simulation model in an embodiment of the present invention;

[0029] Figure 2 It's PM 2.5Actual observed values ​​and PM 2.5 Verify the comparison results of the fitted values;

[0030] Figure 3 It is the comparison result between the actual observed value of O3 and the verified fitted value of O3;

[0031] Figure 4 The initial ENVI-met simulated PM at 16:00 on August 13 2.5 The result after correction in step S3;

[0032] Figure 5 The PM obtained in step S4 at 16:00 on August 13th 2.5 Mapping results;

[0033] Figure 6 It is the result of the initial ENVI-met simulated O3 at 16:00 on August 13 after correction in step S3;

[0034] Figure 7 It is the O3 mapping result obtained in step S4 at 16:00 on August 13;

[0035] Among them, Figure 1 In the figure, arrow ① indicates conv 3*3, ReLU; arrow ② indicates maxpool 2*2; arrow ③ indicates MS-CAM; arrow ④ indicates copy and crop; arrow ⑤ indicates trans-conv 2*2; arrow ⑥ indicates conv 1*1; conv indicates convolution; ReLU indicates ReLU activation function; max pool indicates maximum pooling; copy and crop indicates copy and crop; trans-conv indicates transposed convolution;

[0036] exist Figure 2 In the above table, Reference fine indicates the baseline, corresponding to PM 2.5 Observed value; Fitting line represents the fitted line, corresponding to PM 2.5 Verify the fitted values;

[0037] exist Figure 3 In the figure, Reference fine refers to the baseline, corresponding to the O3 observation value; Fitting line refers to the fitting line, corresponding to the O3 verification fitting value;

[0038] exist Figure 2-Figure 3 In, Density represents the density value;

[0039] exist Figure 4-Figure 7 In the table, Value represents the pollutant concentration value. DETAILED DESCRIPTION

[0040] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0041] Example:

[0042] A PM 2.5 and O3 joint simulation methods, including:

[0043] Step S1, using encrypted fixed sites in the target area (specifically, an actual block in a certain place) to obtain site encrypted monitoring data at actual measurement time points (8:00, 9:00, 10:00, 14:00, 15:00, 16:00, 20:00, 21:00 and 22:00 every day) within a year; combining the site encrypted monitoring data and the urban morphological data and geographical environment elements of the target area within a year to form an ENVI-met input indicator system;

[0044] Step S2, importing the ENVI-met input index system into ENVI-met simulation software (specifically ENVI-met5.6.1 software) to form a simulation file; the built-in solver in the ENVI-met simulation software performs ENVI-met iterative calculation on the simulation file, and outputs the solution result to obtain initial ENVI-met simulation data;

[0045] Step S3, using mobile cruise monitoring in the target area to obtain mobile cruise monitoring data at measured time points (8:00, 9:00, 10:00, 14:00, 15:00, 16:00, 20:00, 21:00 and 22:00 every day) within a year, and preprocessing the mobile cruise monitoring data; using spatiotemporal matching to construct a modeling sample that includes the initial ENVI-met simulation data, the site encrypted monitoring data, the preprocessed mobile cruise monitoring data and the geographical environment elements in a one-to-one correspondence; substituting the modeling sample into the linear mixed effects model LME to predict and obtain a data set of the corrected spatiotemporal distribution of air pollution in the target area; the data set includes PM at spatiotemporal points without measured data 2.5 Concentration values ​​and O3 concentration values ​​and PM at time and space points with measured data 2.5 concentration value and O3 concentration value;

[0046] Step S4: Use the U-Net network (a type of deep convolutional neural network CNN) with multi-scale channel attention mechanism MS-CAM (see the literature DAI Y, GIESEKE F, OEHMCKE S, et al. Attentional Feature Fusion [Z]. 2021 IEEE Winter Conference on Applications of Computer Vision (WACV). Waikoloa, HI, USA. 2021: 3559-68) to construct PM 2.5 and O3 joint simulation model; importing the data set into the joint simulation model to obtain PM 2.5 Joint mapping results with O3.

[0047] In step S1, the urban morphology data includes building height data and road traffic emission data; wherein the road traffic emission data is in the form of PM 2.5 Emissions and NO2 emission characterization;

[0048] The geographical environment factors include PM 2.5 The initial concentration data of O, the initial concentration data of O3, meteorological data, digital elevation model (DEM) data and time data were used. Among them, the meteorological data used the real-time product data of the Land Data Assimilation System (CLDAS-V2.0) of the China Meteorological Administration, and the data was obtained from the China Meteorological Data Network (http: / / data.cma.cn / ) through application. The digital elevation model (DEM) data is used to characterize the surface topography and geomorphology, which will affect the formation of local microclimate and the migration and diffusion of air pollutants. The digital elevation model (DEM) data set with a resolution of 1 km was downloaded from the Loess Plateau Branch of the National Earth System Science Data Center (http: / / loess.geodata.cn). The data is 90m resolution SRTM data obtained through projection transformation and resampling.

[0049] In step S2, when using ENVI-met simulation software to process the index system, a cubic modeling area of ​​1940m*2490m*90m is first determined, and the minimum unit grid size is selected as 10m*10m*3m; an equidistant vertical grid structure is used, combined with a vertical expansion ratio of 30%, to ensure that the total height of the model is not less than 2 times the height of the highest building. In addition, the buildings within the 8 grids outside the boundary are eliminated to eliminate the edge effects that may be caused by the peripheral buildings acting on the boundary conditions.

[0050] In step S2, the simulation files include ENVI-met.inx file and ENVI-met.simx file;

[0051] The data sources forming the ENVI-met.inx file include the DEM data in the geographical environment elements and the urban morphology data;

[0052] The data sources forming the ENVI-met.simx file include the road traffic emission data in the urban morphology data, and the meteorological data and time data in the geographical environment elements.

[0053] In step S3, the preprocessing of the mobile cruise monitoring data includes clustering the mobile cruise monitoring data into data with a resolution of 30 m per hour by using spatiotemporal clustering.

[0054] In step S3, the linear mixed effect model LME is expressed by the following formula:

[0055] Y = Xβ + Zγ + ε;

[0056] Wherein, Y represents the dependent variable, which is the encrypted monitoring data of the site and the preprocessed mobile cruise monitoring data; X is the fixed effect design matrix composed of all independent variables, which are the ENVI-met simulation data and the digital elevation model DEM data and time data in the geographical environment elements; Z is the random effect design matrix composed of time information, which is the time data in the geographical environment elements; β is the coefficient of the fixed effect; γ is the coefficient of the random effect; ε is the random error term.

[0057] The joint simulation model includes the coding stage, PM 2.5 The MS-CAM is set at the end of the encoding stage; the MS-CAM includes two output ends, one of which is connected to the PM 2.5 The other output terminal is connected with the O3 decoding stage.

[0058] See also Figure 1 , the encoding stage includes four encoding blocks arranged in sequence; wherein the first three encoding blocks each include two 3*3 convolution operations and one 2*2 maximum pooling operation, and the last encoding block includes two 3*3 convolution operations and one MS-CAM operation. The PM 2.5The decoding stage and the O3 decoding stage are symmetrically arranged at the two output ends of the MS-CAM, and both include four decoding blocks arranged in sequence; the four decoding blocks are arranged correspondingly to the four encoding blocks in the encoding stage; wherein the first three decoding blocks each include one 2*2 transposed convolution and two 3*3 convolution operations, and the last decoding block includes one 1*1 convolution operation. 2.5 The O3 decoding stage obtains PM 2.5 In addition, in order to reduce the spatial information loss caused by the downsampling process, a skip-layer connection method (i.e., copy and crop) is introduced so that the feature map restored by the transposed convolution contains more semantic information.

[0059] See also Figure 2-Figure 3 , relative to the real PM 2.5 and O3, this embodiment uses the 10-fold cross validation method to construct the PM in step S4 on the corrected data set obtained in step S3. 2.5 PM obtained after verification of the joint simulation model with O3 2.5 The results of the combination with O3, specifically, PM 2.5 Verification of R 2 (R 2 represents goodness of fit) and RMSE (RMSE represents root mean square error) are 0.8 and 16.22 μg / m 3 , O3 verification R 2 and RMSE were 0.86 and 16.53 μg / m 3 This indicates that relative to the real PM 2.5 and O3 observation values, the PM obtained in this embodiment 2.5 The joint mapping results with O3 have higher simulation accuracy.

[0060] See also Figure 4-Figure 7 By comparing the corrected data set obtained in step S3 with the PM obtained in step S4 2.5 Combined with O3 mapping results, the latter can better reflect PM 2.5 The spatial differentiation trend of O3 and O3 shows only slight underestimation of high values ​​and overestimation of low values. Overall, the joint simulation model has good effect and can reflect the pollution generation and diffusion process, which can be applied to subsequent hourly pollutant concentration mapping throughout the year.

[0061] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A PM 2.5 and O3 joint simulation method, characterized in that, include: Step S1, using encrypted fixed sites in the target area to obtain encrypted site monitoring data at actual measurement time points within a one-year period; The encrypted monitoring data of the site and the urban morphology data and geographical environment elements of the target area within a one-year period are combined to form an ENVI-met input indicator system; Step S2, importing the ENVI-met input index system into the ENVI-met simulation software to form a simulation file; the built-in solver in the ENVI-met simulation software performs ENVI-met iterative calculation on the simulation file, and outputs the solution result to obtain initial ENVI-met simulation data; Step S3, using mobile cruise monitoring in the target area to obtain mobile cruise monitoring data at actual time points within a year, and preprocessing the mobile cruise monitoring data; using spatiotemporal matching to construct a modeling sample that includes the initial ENVI-met simulation data, the site encrypted monitoring data, the preprocessed mobile cruise monitoring data, and the geographical environment elements in a one-to-one correspondence; substituting the modeling sample into the linear mixed effects model LME to predict and obtain a data set of the corrected spatiotemporal distribution of air pollution in the target area; Step S4: Use the U-Net network with multi-scale channel attention mechanism MS-CAM to construct PM 2.5 and O3 joint simulation model; importing the data set into the joint simulation model to obtain PM 2.5 Joint mapping results with O3.

2. The PM according to claim 1 2.5 and O3 joint simulation method, characterized in that, In step S1, the urban morphology data includes building height data and road traffic emission data; wherein the road traffic emission data is in the form of PM 2.5 Emissions and NO2 emission characterization; The geographical environment factors include PM 2.5 Initial concentration data of, initial concentration data of O3, meteorological data, digital elevation model DEM data and time data.

3. The PM according to claim 2 2.5 and O3 joint simulation method, characterized in that, In step S2, the simulation files include ENVI-met.inx file and ENVI-met.simx file; The data sources forming the ENVI-met.inx file include the DEM data in the geographical environment elements and the urban morphology data; The data sources forming the ENVI-met.simx file include the road traffic emission data in the urban morphology data, and the meteorological data and time data in the geographical environment elements.

4. The PM according to claim 1 2.5 and O3 joint simulation method, characterized in that, In step S3, the preprocessing of the mobile cruise monitoring data includes clustering the mobile cruise monitoring data into data with a resolution of 10-60 m per hour by using spatiotemporal clustering.

5. The PM according to claim 1 2.5 and O3 joint simulation method, characterized in that, In step S3, the linear mixed effect model LME is expressed by the following formula: Y = Xβ + Zγ + ε; Wherein, Y represents the dependent variable, which is the encrypted monitoring data of the site and the preprocessed mobile cruise monitoring data; X is the fixed effect design matrix composed of all independent variables, which are the ENVI-met simulation data and the digital elevation model DEM data and time data in the geographical environment elements; Z is the random effect design matrix composed of time information, which is the time data in the geographical environment elements; β is the coefficient of the fixed effect; γ is the coefficient of the random effect; ε is the random error term.

6. The PM according to claim 1 2.5 and O3 joint simulation method, characterized in that, The joint simulation model includes the coding stage, PM 2.5 The MS-CAM is set at the end of the encoding stage; the MS-CAM includes two output ends, one of which is connected to the PM 2.5 The other output terminal is connected with the O3 decoding stage.

7. The PM according to claim 6 2.5 and O3 joint simulation method, characterized in that, The encoding stage includes four encoding blocks arranged in sequence; wherein the first three encoding blocks each include 2 3*3 convolution operations and 1 2*2 maximum pooling operation, and the last encoding block includes 2 3*3 convolution operations and 1 MS-CAM operation.

8. The PM according to claim 7 2.5 and O3 joint simulation method, characterized in that, The PM 2.5 The decoding stage and the O3 decoding stage are symmetrically arranged at the two output ends of the MS-CAM, and both include four decoding blocks arranged in sequence; the four decoding blocks are arranged corresponding to the four encoding blocks in the encoding stage; wherein the first three decoding blocks each include one 2*2 transposed convolution and two 3*3 convolution operations, and the last decoding block includes one 1*1 convolution operation.

9. A storage medium, characterized in that: Computer program instructions are stored thereon, and when the computer program instructions are executed by a processor, the PM according to any one of claims 1 to 8 is implemented. 2.5 and O3 joint simulation method.

10. A device, characterized in that: include: at least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the PM according to any one of claims 1 to 8. 2.5 and O3 joint simulation method.

Images