Atmospheric fine particulate matter PM 2.5 Impact factor analysis method and device

Abstract

The present invention provides an atmospheric fine particulate matter PM 2.5 The method and device for analyzing influencing factors belong to the field of data processing technology. The method comprises: obtaining multi-source heterogeneous data of a target area within a target time period, wherein the multi-source heterogeneous data includes atmospheric fine particulate matter PM 2.5 Spatial distribution data of concentration, natural environment data and social and economic activity data; pre-process multi-source heterogeneous data to obtain grid data with preset spatiotemporal resolution. The grid data includes PM 2.5 Gridded data of concentration and PM 2.5 The grid data of each influencing factor of PM 2.5 The grid data of concentration and grid data of each influencing factor are input into the pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model. The present invention can fully consider the nonlinear dynamic effects of multiple factors, quantitatively analyze each influencing factor, and improve PM 2.5 The accuracy of impact factor analysis.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, in particular to an atmospheric fine particulate matter PM 2.5 Impact factor analysis method and device. Background Art

[0002] Find out PM 2.5 The driving mechanism can quantitatively analyze the influencing factors on PM 2.5 The existing technology is in the analysis of the influencing factors on PM 2.5 In the process of the role of only considering a single factor and static driving mechanism, failed to fully consider the nonlinear-dynamic effects of multiple factors. Summary of the invention

[0003] The present invention provides an atmospheric fine particulate matter PM 2.5 The method and device for analyzing the impact factor are used to solve the problem of the existing technology in analyzing the impact factor on PM 2.5 In the process of the action of only considering a single factor and static driving mechanism, the defect of failing to fully consider the nonlinear-dynamic effects of multiple factors.

[0004] The present invention provides an atmospheric fine particulate matter PM 2.5 Impact factor analysis methods include:

[0005] Acquire multi-source heterogeneous data of the target area within the target time period, the multi-source heterogeneous data includes atmospheric fine particulate matter PM 2.5 Spatial distribution data of concentration, natural environment data and socio-economic activity data;

[0006] The multi-source heterogeneous data are preprocessed to obtain grid data with a preset spatiotemporal resolution, wherein the grid data includes PM 2.5 Gridded data of concentration and PM 2.5 Grid data of each influencing factor;

[0007] The PM 2.5 The grid data of the concentration and the grid data of each influencing factor are input into a pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model;

[0008] The impact factor analysis model is based on the PM of the sample area. 2.5 Sample grid data for concentration and PM 2.5 The sample grid data of each sample influencing factor and the spatial distribution pattern label of each sample influencing factor are trained.

[0009] In some embodiments, the impact factor analysis model includes a feature extraction layer, an impact analysis layer, a multi-scale geographically weighted regression module, and a clustering module;

[0010] Correspondingly, the PM 2.5 The grid data of the concentration and the grid data of each influencing factor are input into a pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model, including:

[0011] The PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the feature extraction layer to obtain the PM output by the feature extraction layer. 2.5 Characteristic information of concentration and characteristic information of each influencing factor;

[0012] The PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the impact analysis layer, and the influence of each influencing factor on the PM output by the impact analysis layer is obtained. 2.5 The extent and manner of the effect of concentration;

[0013] The above mentioned influencing factors are applied to the PM 2.5 The influence degree and influence mode of the concentration are input into the multi-scale geographical weighted regression module to obtain the spatial heterogeneity information of the influencing factors output by the multi-scale geographical weighted regression module, wherein the spatial heterogeneity information includes the nature of action, the degree of action and the spatial multi-scale property;

[0014] The spatial heterogeneity information of each influencing factor is input into the clustering module to obtain the spatial distribution pattern of each influencing factor output by the clustering module.

[0015] In some embodiments, the feature extraction layer includes a first feature extraction layer and a second feature extraction layer;

[0016] Correspondingly, the PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the feature extraction layer to obtain the PM output by the feature extraction layer. 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor include:

[0017] The PM 2.5 The grid data of the concentration is input into the first feature extraction layer, and the PM output by the first feature extraction layer is obtained. 2.5 Characteristic information of concentration;

[0018] The grid data of each influencing factor is input into the second feature extraction layer to obtain feature information of each influencing factor output by the second feature extraction layer.

[0019] In some embodiments, the impact analysis layer includes an impact degree analysis layer and an impact mode analysis layer;

[0020] Correspondingly, the PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the impact analysis layer, and the influence of each influencing factor on the PM output by the impact analysis layer is obtained. 2.5 The extent and manner of the impact of concentration include:

[0021] The PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the influence degree analysis layer, and the influence degree analysis layer outputs the influence factors on the PM 2.5 The degree of influence of concentration;

[0022] The PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the influencing mode analysis layer, and the influencing factors output by the influencing mode analysis layer on the PM are obtained. 2.5 How concentration affects the

[0023] In some embodiments, the impact degree analysis layer includes a time impact degree analysis layer and a space impact degree analysis layer;

[0024] Correspondingly, the PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the influence degree analysis layer, and the influence degree analysis layer outputs the influence factors on the PM 2.5 The degree of influence of concentration includes:

[0025] The PM 2.5 The time characteristics of the concentration and the time characteristics of each influencing factor are input into the time influence degree analysis layer, and the time influence degree analysis layer outputs the influence factors on the PM 2.5 The extent of the influence of concentration on the time dimension;

[0026] The PM 2.5 The spatial characteristics of the concentration and the spatial characteristics of each influencing factor are input into the spatial influence degree analysis layer, and the spatial influence degree analysis layer outputs the influence factors on the PM 2.5 The extent of the concentration effect in the spatial dimension.

[0027] In some embodiments, the impact pattern analysis layer includes a time impact pattern analysis layer and a space impact pattern analysis layer;

[0028] Correspondingly, the PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the influencing mode analysis layer, and the influencing factors output by the influencing mode analysis layer on the PM are obtained. 2.5 Concentrations are affected by:

[0029] The PM 2.5 The time characteristics of the concentration and the time characteristics of each influencing factor are input into the time influence mode analysis layer, and the time influence mode analysis layer outputs the influence factors on the PM 2.5 How concentration affects the time dimension;

[0030] The PM 2.5 The spatial characteristics of the concentration and the spatial characteristics of each influencing factor are input into the spatial influencing mode analysis layer, and the spatial influencing mode analysis layer outputs the influencing factors on the PM 2.5 How concentration affects spatial dimensions.

[0031] In some embodiments, the spatial heterogeneity information of each influencing factor is input into the clustering module to obtain the spatial distribution pattern of each influencing factor output by the clustering module, including:

[0032] According to the spatial heterogeneity information of each influencing factor, the influence of each influencing factor on PM in each grid of the target area is obtained. 2.5 The extent of the effect of concentration;

[0033] Based on the clustering module, the K-means algorithm is used to classify the PM according to the influencing factors of each grid. 2.5 The grids are clustered according to the degree of effect of concentration to obtain a plurality of grid sets, and the spatial distribution pattern of the influencing factors is determined according to the plurality of grid sets.

[0034] In some embodiments, the preprocessing of the multi-source heterogeneous data to obtain grid data of a preset spatiotemporal resolution includes:

[0035] Preprocessing the natural environment data to obtain natural environment impact factors;

[0036] Preprocessing the social and economic activity data to obtain social and economic activity impact factors;

[0037] For the PM 2.5The spatial distribution data of concentration, natural environment influencing factors and socio-economic activity influencing factors are preprocessed to obtain grid data with preset spatiotemporal resolution.

[0038] In some embodiments, the training process of the impact factor analysis model includes:

[0039] Acquire sample multi-source heterogeneous data of the sample area within the sample time period, wherein the sample multi-source heterogeneous data includes PM 2.5 Concentration sample spatial distribution data, sample natural environment data, and sample socio-economic activity data;

[0040] The sample multi-source heterogeneous data is preprocessed to obtain sample grid data with a preset spatiotemporal resolution, wherein the sample grid data includes PM 2.5 Sample grid data for concentration and PM 2.5 Sample grid data of each sample influencing factor;

[0041] Determine the spatial distribution pattern label of each sample impact factor;

[0042] According to the PM 2.5 The sample grid data of concentration and the sample grid data of each sample influencing factor are used as training samples, and the spatial distribution pattern labels of each sample influencing factor are used as sample labels to train the initial influencing factor analysis model. After the initial influencing factor analysis model is trained, the influencing factor analysis model is obtained.

[0043] The present invention also provides an atmospheric fine particulate matter PM 2.5 Impact factor analysis device, including:

[0044] An acquisition unit is used to acquire multi-source heterogeneous data of a target area in a target time period, wherein the multi-source heterogeneous data includes atmospheric fine particulate matter PM 2.5 Spatial distribution data of concentration, natural environment data and socio-economic activity data;

[0045] A preprocessing unit is used to preprocess the multi-source heterogeneous data to obtain grid data with a preset spatiotemporal resolution, wherein the grid data includes PM 2.5 Gridded data of concentration and PM 2.5 Grid data of each influencing factor;

[0046] A parsing unit for converting the PM 2.5 The grid data of the concentration and the grid data of each influencing factor are input into a pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model;

[0047] The impact factor analysis model is based on the PM of the sample area.2.5 Sample grid data for concentration and PM 2.5 The sample grid data of each sample influencing factor and the spatial distribution pattern label of each sample influencing factor are trained.

[0048] Atmospheric fine particulate matter PM provided by the present invention 2.5 The method and device for analyzing the impact factors obtains multi-source heterogeneous data of the target area within the target time period, pre-processes the multi-source heterogeneous data, and obtains grid data with preset spatiotemporal resolution. The grid data includes PM 2.5 Gridded data of concentration and PM 2.5 The grid data of each influencing factor of PM 2.5 The grid data of concentration and grid data of each influencing factor are input into the pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model, which fully considers the nonlinear-dynamic effects of multiple factors and can achieve PM 2.5 The fine-grained quantitative analysis of various influencing factors improves PM 2.5 The accuracy of impact factor analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 It is the atmospheric fine particulate matter PM provided by the embodiment of the present invention 2.5 One of the flowcharts of the impact factor analysis method;

[0051] Figure 2 It is the atmospheric fine particulate matter PM provided by the embodiment of the present invention 2.5 Flowchart 2 of the impact factor analysis method;

[0052] Figure 3 The influencing factors of the present invention provide the PM 2.5 Schematic diagram of the influence of concentration in spatial and temporal dimensions;

[0053] Figure 4 The population density and PM provided by the embodiment of the present invention are 2.5 Schematic diagram of the nonlinear relationship of concentration;

[0054] Figure 5 It is a flowchart of the training process of the impact factor analysis model provided by an embodiment of the present invention;

[0055] Figure 6 It is the atmospheric fine particulate matter PM provided by the embodiment of the present invention 2.5 Schematic diagram of the structure of the impact factor analysis device. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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 without creative work are within the scope of protection of the present invention.

[0057] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present invention can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more.

[0058] Figure 1 The atmospheric fine particulate matter PM provided by the embodiment of the present invention 2.5 One of the flow charts of the impact factor analysis method. Figure 1 As shown, a PM 2.5 The impact factor analysis method includes the following steps: step 110, step 120 and step 130. The steps of the method flow are only a possible implementation mode of the present invention.

[0059] Step 110: Acquire multi-source heterogeneous data of the target area within the target time period, where the multi-source heterogeneous data includes atmospheric fine particulate matter PM 2.5 Spatial distribution data of concentration, natural environment data and socio-economic activity data.

[0060] Among them, the natural environment data at least includes: remote sensing image data, temperature data, relative humidity data, wind speed data, and altitude data.

[0061] Among them, data on social and economic activities include at least: population density data, night light data, land use data, road network data, and statistical yearbook data; statistical yearbook data include annual statistical data on various aspects such as economy, society, and culture.

[0062] Optionally, the time interval can be set to 1 year or multiple years, such as 2000, 2010, 2019, etc.

[0063] Optionally, multi-source heterogeneous data are obtained through remote sensing satellites, geographic information systems, environmental monitoring stations, meteorological monitoring stations, etc.

[0064] It can be understood that by obtaining multi-source heterogeneous data of the target area within the target time period, the subsequent PM 2.5 Impact factor analysis provides a rich and comprehensive data foundation, which helps to improve the accuracy of impact factor analysis.

[0065] Step 120: pre-process the multi-source heterogeneous data to obtain grid data with a preset spatiotemporal resolution. The grid data includes PM 2.5 Gridded data of concentration and PM 2.5 Grid data of each influencing factor.

[0066] Optionally, preprocessing such as data cleaning, alignment, and interpolation is performed on multi-source heterogeneous data.

[0067] Among them, the influencing factors include at least natural environment influencing factors and social and economic activity influencing factors.

[0068] In some embodiments, preprocessing multi-source heterogeneous data to obtain grid data of preset spatiotemporal resolution includes:

[0069] Preprocess the natural environment data to obtain the natural environment impact factors;

[0070] Preprocess the data of social and economic activities to obtain the impact factors of social and economic activities;

[0071] About PM 2.5 The spatial distribution data of concentration, natural environment influencing factors and socio-economic activity influencing factors are preprocessed to obtain grid data with preset spatiotemporal resolution.

[0072] Among them, natural environment influencing factors include at least: vegetation coverage, temperature, relative humidity, wind speed and surface undulation.

[0073] Optionally, the normalized vegetation index is calculated based on remote sensing image data of vegetation, and the vegetation coverage is obtained by inverting the normalized vegetation index and a pixel binary model; and the surface undulation is calculated based on altitude data.

[0074] Among them, the influencing factors of social and economic activities include at least: population density, night light index, building density, road network density and gross domestic product (GDP).

[0075] Optionally, the building density is calculated based on the land use data; and the road network density is calculated based on the road network data and the land use data.

[0076] Optionally, PM 2.5 The spatial distribution data of concentration, natural environment influencing factors and socio-economic activity influencing factors are preprocessed by reprojection, resampling, clipping, interpolation and gridding.

[0077] Optionally, the missing values ​​of the vacant grid are filled based on the values ​​of the nearest neighbors of the vacant grid.

[0078] Optionally, the time resolution in the preset time and space resolution may be 1 day, 10 days, 30 days, etc., and the spatial resolution may be 1km×1km, 2km×2km, 3km×3km, etc.

[0079] Step 130: PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model.

[0080] Among them, the impact factor analysis model is based on the PM of the sample area. 2.5 Sample grid data for concentration and PM 2.5 The sample grid data of each sample influencing factor and the spatial distribution pattern label of each sample influencing factor are used for training.

[0081] Optionally, the influencing factor analysis model is used to analyze the nonlinear relationship, threshold effect and spatial multi-scale effect between each influencing factor and PM2.5.

[0082] Figure 2 The atmospheric fine particulate matter PM provided by the embodiment of the present invention 2.5 The second flowchart of the impact factor analysis method. Figure 2 As shown, a PM 2.5 The impact factor analysis method includes the following steps:

[0083] Acquire multi-source heterogeneous data of the target area within the target time period, including atmospheric fine particulate matter PM 2.5 Spatial distribution data of concentration, natural environment data, and economic and social data;

[0084] Preprocess the natural environment data to obtain the natural environment impact factors, and preprocess the economic and social data to obtain the economic and social impact factors;

[0085] About PM 2.5 The spatial distribution data of concentration, natural environmental impact factors and economic and social impact factors are preprocessed (such as reprojection, resampling, clipping, etc.) to obtain grid data with preset spatiotemporal resolution (such as benchmark data with a resolution of 1 km);

[0086] Determine the time information entropy of the grid data, and based on the time information entropy, obtain the time series characteristics of the grid data (including the time series change intensity and time series change trend);

[0087] The time series characteristics of grid data are processed based on the random forest algorithm to obtain the influence of each factor on PM 2.5 The influence of each influencing factor on PM concentration is obtained by processing the time series characteristics of the grid data based on the partial dependence graph. 2.5 How concentration affects

[0088] Using K-means algorithm, PM 2.5 The influence degree and influence mode of concentration are determined, and the grids in the target area are clustered to obtain the spatial distribution pattern of each influencing factor.

[0089] In the embodiment of the present invention, multi-source heterogeneous data of the target area within the target time period are obtained, and the multi-source heterogeneous data are pre-processed to obtain grid data of preset spatiotemporal resolution, the grid data including PM 2.5 Gridded data of concentration and PM 2.5 The grid data of each influencing factor of PM 2.5 The grid data of concentration and grid data of each influencing factor are input into the pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model, which fully considers the nonlinear-dynamic effects of multiple factors and can achieve PM 2.5 The fine-grained quantitative analysis of various influencing factors improves PM 2.5 The accuracy of impact factor analysis.

[0090] In some embodiments, the impact factor analysis model includes a feature extraction layer, an impact analysis layer, a multi-scale geographically weighted regression module, and a clustering module;

[0091] Correspondingly, in step 130, PM 2.5 The grid data of concentration and grid data of each influencing factor are input into the pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model, including:

[0092] Step 131: PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the feature extraction layer to obtain the PM output by the feature extraction layer. 2.5 Characteristic information of concentration and characteristic information of each influencing factor;

[0093] Step 132: PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the impact analysis layer, and the output of the impact analysis layer is the influence of each influencing factor on PM2.5 The extent and manner of the effect of concentration;

[0094] Step 133: The influence factors on PM 2.5 The influence degree and influence mode of concentration are input into the multi-scale geographical weighted regression module to obtain the spatial heterogeneity information of each influencing factor output by the multi-scale geographical weighted regression module. The spatial heterogeneity information includes the nature of action, the degree of action and the spatial multi-scale property.

[0095] Step 134: input the spatial heterogeneity information of each influencing factor into the clustering module to obtain the spatial distribution pattern of each influencing factor output by the clustering module.

[0096] Optionally, PM 2.5 The characteristic information of concentration includes PM 2.5 The intensity and trend of concentration changes. The characteristic information of each influencing factor includes the intensity and trend of concentration changes.

[0097] In some embodiments, the feature extraction layer includes a first feature extraction layer and a second feature extraction layer;

[0098] Correspondingly, in step 131, PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the feature extraction layer to obtain the PM output by the feature extraction layer. 2.5 Characteristic information of concentration and each influencing factor, including:

[0099] PM 2.5 The grid data of the concentration is input into the first feature extraction layer, and the PM output by the first feature extraction layer is obtained. 2.5 Characteristic information of concentration;

[0100] The grid data of each influencing factor is input into the second feature extraction layer to obtain the feature information of each influencing factor output by the second feature extraction layer.

[0101] Optionally, PM 2.5 The characteristic information of concentration includes PM 2.5 The temporal and spatial characteristics of concentration, the characteristic information of each influencing factor includes the temporal and spatial characteristics of each influencing factor.

[0102] Optionally, according to PM 2.5 The first time information entropy is calculated based on the grid data of the concentration, and the PM 2.5 Temporal characteristics of concentration.

[0103] Optionally, a second time information entropy is calculated according to the grid data of each influencing factor, and based on the second time information entropy, a time feature of each influencing factor is calculated.

[0104] Optionally, the calculation formula of the first time information entropy is as follows:

[0105] ;

[0106] ;

[0107] ;

[0108] in, represents the value of information entropy at the first time, is the first time frequency factor, indicating PM 2.5 The gridded data time scale of the concentration, is the total number of years, Indicates Year, is the scaling factor, Indicates PM 2.5 The time characteristics of the concentration, sgn represents the sign function, For the The complexity parameter of the year is used to quantify the PM 2.5 The complexity or uncertainty of the temporal characteristics of concentration over the time series.

[0109] It should be noted that the calculation formula of the second time information entropy is similar to the calculation formula of the first time information entropy, and the principles of the two are the same.

[0110] In some embodiments, the impact analysis layer includes an impact degree analysis layer and an impact mode analysis layer;

[0111] Correspondingly, PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the impact analysis layer, and the output of the impact analysis layer is the influence of each influencing factor on PM 2.5 The extent and manner of the impact of concentration include:

[0112] PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the influence degree analysis layer, and the influence degree analysis layer outputs the influence factors on PM 2.5 The degree of influence of concentration;

[0113] PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the influence mode analysis layer, and the influence mode analysis layer outputs the influence factors on PM 2.5 How concentration affects the

[0114] Optionally, the impact analysis layer uses a random forest algorithm to 2.5The characteristic information of concentration and the characteristic information of each influencing factor are analyzed to construct PM 2.5 The correlation between concentration and each influencing factor is obtained to obtain the influence of each influencing factor on PM 2.5 The influence of concentration.

[0115] Optionally, the influence mode analysis layer uses a partial dependency graph to analyze PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are analyzed to construct PM 2.5 The nonlinear relationship between concentration and each influencing factor is obtained, and the influence of each influencing factor on PM 2.5 How concentration affects the

[0116] In some embodiments, the impact level analysis layer includes a temporal impact level analysis layer and a spatial impact level analysis layer;

[0117] Correspondingly, PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the influence degree analysis layer, and the influence degree analysis layer outputs the influence factors on PM 2.5 The degree of influence of concentration includes:

[0118] PM 2.5 The time characteristics of concentration and the time characteristics of each influencing factor are input into the time impact degree analysis layer, and the output of the time impact degree analysis layer is the influence of each influencing factor on PM 2.5 The extent of the influence of concentration on the time dimension;

[0119] PM 2.5 The spatial characteristics of concentration and the spatial characteristics of each influencing factor are input into the spatial impact degree analysis layer, and the output of the spatial impact degree analysis layer is the influence of each influencing factor on PM 2.5 The extent of the concentration effect in the spatial dimension.

[0120] Figure 3 The influencing factors provided by the embodiment of the present invention have an impact on PM 2.5 Schematic diagram of the influence of concentration on spatial and temporal dimensions. Figure 3As shown in the figure, the influencing factors include: Population density (POP), Nighttime Light (NL), Building Use Index (BUI), Road Network Density (RND), Gross Domestic Product (GDP), Fractional Vegetation Cover (FVC), Temperature (TEM), Relative Humidity (RH), Wind Speed ​​(WS), Relief Degree of Land Surface (RDLS). The Contribution indicates the contribution percentage of each influencing factor to the PM2.5 concentration. Figure 3 (a) indicates the influence of various factors on PM 2.5 Contribution of GDP, TEM and WS to PM 2.5 The concentration has a greater impact on the spatial dimension. RDLS has a greater impact on PM 2.5 Concentration has the smallest effect in the spatial dimension; Figure 3 (b) shows the influence of various factors on PM 2.5 The contribution of GDP to PM 2.5 Concentration has the greatest impact on time, far exceeding other influencing factors. POP and FVC have the greatest impact on PM 2.5 The effect of concentration on the time dimension is smaller.

[0121] In some embodiments, the impact pattern analysis layer includes a temporal impact pattern analysis layer and a spatial impact pattern analysis layer;

[0122] Correspondingly, PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the influence mode analysis layer, and the influence mode analysis layer outputs the influence factors on PM 2.5 Concentrations are affected by:

[0123] PM 2.5 The time characteristics of concentration and the time characteristics of each influencing factor are input into the time impact mode analysis layer, and the output of the time impact mode analysis layer is the influence of each influencing factor on PM 2.5 How concentration affects the time dimension;

[0124] PM 2.5The spatial characteristics of concentration and the spatial characteristics of each influencing factor are input into the spatial influencing mode analysis layer, and the output of the spatial influencing mode analysis layer is used to obtain the influence of each influencing factor on PM 2.5 How concentration affects spatial dimensions.

[0125] It should be noted that the various influencing factors and PM 2.5 The concentration relationship is nonlinear.

[0126] Figure 4 Population density and PM provided by the embodiment of the present invention 2.5 Schematic diagram of the nonlinear relationship between concentration. Figure 4 As shown, population density and PM 2.5 The concentration shows a nonlinear relationship. With the increase of population density, PM 2.5 The concentration first rises rapidly, then decreases slightly, and then remains constant; the time series information entropy value of population density (POP entropy) and PM 2.5 The information entropy of the time series of concentration (PM 2.5 entropy is nonlinear, with the time series information entropy value of population density (PM 2.5 entropy) increase, PM 2.5 The information entropy of the time series of concentration (PM 2.5 entropy) first decreases and then increases.

[0127] In some embodiments, PM 2.5 The multi-scale geographically weighted regression model was used to capture PM concentration as the dependent variable and various influencing factors as independent variables. 2.5 The spatial heterogeneity of the driving mechanism can be used to obtain the influence of each influencing factor on PM 2.5 The nature, degree and spatial multi-scale of its action.

[0128] Optionally, the calculation formula of the regression coefficient of each influencing factor is as follows:

[0129] ;

[0130] in, Indicates PM of each grid 2.5 Concentration data, Indicates The grid Impact factors, The bandwidth is bw. The local regression coefficient of each influencing factor is It is The centroid coordinates of the grid, is the regression residual.

[0131] In some embodiments, the spatial heterogeneity information of each influencing factor is input into a clustering module to obtain the spatial distribution pattern of each influencing factor output by the clustering module, including:

[0132] According to the spatial heterogeneity information of each influencing factor, the influence of each influencing factor on PM in each grid of the target area is obtained. 2.5 The extent of the effect of concentration;

[0133] Based on the clustering module, the K-means algorithm is used to classify PM according to the influencing factors of each grid. 2.5 The grids are clustered according to the degree of concentration effect to obtain multiple grid sets, and the spatial distribution pattern of each influencing factor is determined based on the multiple grid sets.

[0134] Optionally, for each grid set, a new center is continuously calculated, and then other grid sets are reallocated according to the new center, and the iteration ends until the centers of all grids no longer change.

[0135] Optionally, the Euclidean distances of different grids are calculated as follows:

[0136] ;

[0137] ;

[0138] ;

[0139] in, Representation Network and The Euclidean distance of For the Grid The center of is the total cluster center, For the cluster centers, n is the total number of grids in the grid set, is the objective function, and k is the number of clusters.

[0140] Figure 5 The following is a flow chart of the training process of the impact factor analysis model provided by the embodiment of the present invention. Figure 5 As shown, in some embodiments, the training process of the impact factor analysis model includes:

[0141] Step 510: Obtain sample multi-source heterogeneous data of the sample area within the sample time period, where the sample multi-source heterogeneous data includes PM 2.5 Concentration sample spatial distribution data, sample natural environment data, and sample socio-economic activity data;

[0142] Step 520: pre-process the sample multi-source heterogeneous data to obtain sample grid data of preset spatiotemporal resolution. The sample grid data includes PM 2.5 Sample grid data for concentration and PM 2.5 Sample grid data of each sample influencing factor;

[0143] Step 530, determining the spatial distribution pattern label of each sample impact factor;

[0144] Step 540: PM 2.5 The sample grid data of concentration and the sample grid data of each sample influencing factor are used as training samples, and the spatial distribution pattern label of each sample influencing factor is used as the sample label to train the initial influencing factor analysis model. After the initial influencing factor analysis model is trained, the influencing factor analysis model is obtained.

[0145] Optionally, the initial impact factor analysis model includes an initial feature extraction layer, an initial impact analysis layer, an initial multi-scale geographically weighted regression module and an initial clustering module;

[0146] Correspondingly, PM 2.5 The grid data of concentration and grid data of each influencing factor are input into the pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model, including:

[0147] PM 2.5 The sample grid data of concentration and the sample grid data of each sample influencing factor are input into the initial feature extraction layer to obtain the PM output by the initial feature extraction layer. 2.5 Sample characteristic information of concentration and sample characteristic information of each sample influencing factor;

[0148] PM 2.5 The sample characteristic information of concentration and the sample characteristic information of each sample influencing factor are input into the initial impact analysis layer, and the output of the initial impact analysis layer is the influence of each sample influencing factor on PM 2.5 The extent and manner of the effect of concentration;

[0149] The influence factors of each sample on PM 2.5 The influence degree and influence mode of concentration are input into the initial multi-scale geographical weighted regression module to obtain the sample spatial heterogeneity information of each sample influencing factor output by the initial multi-scale geographical weighted regression module. The sample spatial heterogeneity information includes the nature of action, degree of action and spatial multi-scale.

[0150] The sample spatial heterogeneity information of each sample influencing factor is input into the initial clustering module to obtain the spatial distribution pattern of each sample influencing factor output by the initial clustering module.

[0151] The atmospheric fine particulate matter PM provided in the embodiment of the present invention is as follows: 2.5 The influencing factor analysis device is described below. The atmospheric fine particulate matter PM 2.5 The impact factor analysis device and the atmospheric fine particulate matter PM described above 2.5 The impact factor analysis methods can be referenced to each other.

[0152] Figure 6 The atmospheric fine particulate matter PM provided by the embodiment of the present invention 2.5 The schematic diagram of the impact factor analysis device is as follows: Figure 6 As shown, the atmospheric fine particulate matter PM 2.5 The impact factor analysis device 600 includes:

[0153] The acquisition unit 610 is used to acquire multi-source heterogeneous data of the target area in the target time period, and the multi-source heterogeneous data includes atmospheric fine particulate matter PM 2.5 Spatial distribution data of concentration, natural environment data and socio-economic activity data;

[0154] The preprocessing unit 620 is used to preprocess the multi-source heterogeneous data to obtain grid data with a preset spatiotemporal resolution. The grid data includes PM 2.5 Gridded data of concentration and PM 2.5 Grid data of each influencing factor;

[0155] The analysis unit 630 is used to convert PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model;

[0156] Among them, the impact factor analysis model is based on the PM of the sample area. 2.5 Sample grid data for concentration and PM 2.5 The sample grid data of each sample influencing factor and the spatial distribution pattern label of each sample influencing factor are used for training.

[0157] Optionally, the impact factor analysis model includes a feature extraction layer, an impact analysis layer, a multi-scale geographically weighted regression module, and a clustering module;

[0158] Correspondingly, PM 2.5 The grid data of concentration and grid data of each influencing factor are input into the pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model, including:

[0159] PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the feature extraction layer to obtain the PM output by the feature extraction layer.2.5 Characteristic information of concentration and characteristic information of each influencing factor;

[0160] PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the impact analysis layer, and the output of the impact analysis layer is the influence of each influencing factor on PM 2.5 The extent and manner of the effect of concentration;

[0161] The influence factors on PM 2.5 The influence degree and influence mode of concentration are input into the multi-scale geographical weighted regression module to obtain the spatial heterogeneity information of each influencing factor output by the multi-scale geographical weighted regression module. The spatial heterogeneity information includes the nature of action, the degree of action and the spatial multi-scale property.

[0162] The spatial heterogeneity information of each influencing factor is input into the clustering module to obtain the spatial distribution pattern of each influencing factor output by the clustering module.

[0163] Optionally, the feature extraction layer includes a first feature extraction layer and a second feature extraction layer;

[0164] Correspondingly, PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the feature extraction layer to obtain the PM output by the feature extraction layer. 2.5 Characteristic information of concentration and each influencing factor, including:

[0165] PM 2.5 The grid data of the concentration is input into the first feature extraction layer, and the PM output by the first feature extraction layer is obtained. 2.5 Characteristic information of concentration;

[0166] The grid data of each influencing factor is input into the second feature extraction layer to obtain the feature information of each influencing factor output by the second feature extraction layer.

[0167] Optionally, the impact analysis module includes an impact degree analysis layer and an impact mode analysis layer;

[0168] Correspondingly, PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the impact analysis layer, and the output of the impact analysis layer is the influence of each influencing factor on PM 2.5 The extent and manner of the impact of concentration include:

[0169] PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the influence degree analysis layer, and the influence degree analysis layer outputs the influence factors on PM 2.5 The degree of influence of concentration;

[0170] PM2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the influence mode analysis layer, and the influence mode analysis layer outputs the influence factors on PM 2.5 How concentration affects the

[0171] Optionally, the impact degree analysis layer includes a time impact degree analysis layer and a space impact degree analysis layer;

[0172] Correspondingly, PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the influence degree analysis layer, and the influence degree analysis layer outputs the influence factors on PM 2.5 The degree of influence of concentration includes:

[0173] PM 2.5 The time characteristics of concentration and the time characteristics of each influencing factor are input into the time impact degree analysis layer, and the output of the time impact degree analysis layer is the influence of each influencing factor on PM 2.5 The extent of the influence of concentration on the time dimension;

[0174] PM 2.5 The spatial characteristics of concentration and the spatial characteristics of each influencing factor are input into the spatial impact degree analysis layer, and the output of the spatial impact degree analysis layer is the influence of each influencing factor on PM 2.5 The extent of the concentration effect in the spatial dimension.

[0175] Optionally, the impact mode analysis layer includes a time impact mode analysis layer and a space impact mode analysis layer;

[0176] Correspondingly, PM 2.5 The characteristic information of concentration and the characteristic information of each influencing factor are input into the influence mode analysis layer, and the influence mode analysis layer outputs the influence factors on PM 2.5 Concentrations are affected by:

[0177] PM 2.5 The time characteristics of concentration and the time characteristics of each influencing factor are input into the time impact mode analysis layer, and the output of the time impact mode analysis layer is the influence of each influencing factor on PM 2.5 How concentration affects the time dimension;

[0178] PM 2.5 The spatial characteristics of concentration and the spatial characteristics of each influencing factor are input into the spatial influencing mode analysis layer, and the output of the spatial influencing mode analysis layer is used to obtain the influence of each influencing factor on PM 2.5 How concentration affects spatial dimensions.

[0179] Optionally, the spatial heterogeneity information of each influencing factor is input into the clustering module to obtain the spatial distribution pattern of each influencing factor output by the clustering module, including:

[0180] According to the spatial heterogeneity information of each influencing factor, the influence of each influencing factor on PM in each grid of the target area is obtained. 2.5 The extent of the effect of concentration;

[0181] Based on the clustering module, the K-means algorithm is used to classify PM according to the influencing factors of each grid. 2.5 The grids are clustered according to the degree of concentration effect to obtain multiple grid sets, and the spatial distribution pattern of each influencing factor is determined based on the multiple grid sets.

[0182] Optionally, preprocessing the multi-source heterogeneous data to obtain grid data of preset spatiotemporal resolution includes:

[0183] Preprocess the natural environment data to obtain the natural environment impact factors;

[0184] Preprocess the data of social and economic activities to obtain the impact factors of social and economic activities;

[0185] About PM 2.5 The spatial distribution data of concentration, natural environment influencing factors and socio-economic activity influencing factors are preprocessed to obtain grid data with preset spatiotemporal resolution.

[0186] Optionally, the training process of the impact factor analysis model includes:

[0187] Obtain sample multi-source heterogeneous data in the sample area within the sample time period. The sample multi-source heterogeneous data includes PM 2.5 Concentration sample spatial distribution data, sample natural environment data, and sample socio-economic activity data;

[0188] Preprocess the sample multi-source heterogeneous data to obtain sample grid data with preset temporal and spatial resolution. The sample grid data includes PM 2.5 Sample grid data for concentration and PM 2.5 Sample grid data of each sample influencing factor;

[0189] Determine the spatial distribution pattern label of each sample's influencing factor;

[0190] PM 2.5 The sample grid data of concentration and the sample grid data of each sample influencing factor are used as training samples, and the spatial distribution pattern label of each sample influencing factor is used as the sample label to train the initial influencing factor analysis model. After the initial influencing factor analysis model is trained, the influencing factor analysis model is obtained.

[0191] It should be noted that the atmospheric fine particulate matter PM provided in the embodiment of the present invention 2.5The impact factor analysis device can achieve the above-mentioned atmospheric fine particulate matter PM 2.5 All the method steps implemented in the embodiment of the influencing factor analysis method can achieve the same technical effect, and the parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0192] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0193] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0194] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. PM, a type of atmospheric fine particulate matter 2.5 The impact factor analysis method is characterized by: include: Acquire multi-source heterogeneous data of the target area within the target time period, the multi-source heterogeneous data includes atmospheric fine particulate matter PM 2.5 Spatial distribution data of concentration, natural environment data and socio-economic activity data; The multi-source heterogeneous data is preprocessed to obtain grid data with a preset spatiotemporal resolution, wherein the grid data includes PM 2.5 Gridded data of concentration and PM 2.5 Grid data of each influencing factor; The PM 2.5 The grid data of the concentration and the grid data of each influencing factor are input into a pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model; The impact factor analysis model is based on the PM of the sample area. 2.5 Sample grid data for concentration and PM 2.5 The sample grid data of each sample influencing factor and the spatial distribution pattern label of each sample influencing factor are trained; The impact factor analysis model includes a feature extraction layer, an impact analysis layer, a multi-scale geographically weighted regression module and a clustering module; Correspondingly, the PM 2.5 The grid data of the concentration and the grid data of each influencing factor are input into a pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model, including: The PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the feature extraction layer to obtain the PM output by the feature extraction layer. 2.5 Characteristic information of concentration and characteristic information of each influencing factor; The PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the impact analysis layer, and the influence of each influencing factor on the PM output by the impact analysis layer is obtained. 2.5 The extent and manner of the effect of concentration; The above mentioned influencing factors are applied to the PM 2.5 The influence degree and influence mode of the concentration are input into the multi-scale geographical weighted regression module to obtain the spatial heterogeneity information of the influencing factors output by the multi-scale geographical weighted regression module, wherein the spatial heterogeneity information includes the nature of action, the degree of action and the spatial multi-scale property; The spatial heterogeneity information of each influencing factor is input into the clustering module to obtain the spatial distribution pattern of each influencing factor output by the clustering module.

2. The atmospheric fine particulate matter PM according to claim 1 2.5 The impact factor analysis method is characterized by: The feature extraction layer includes a first feature extraction layer and a second feature extraction layer; Correspondingly, the PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the feature extraction layer to obtain the PM output by the feature extraction layer. 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor include: The PM 2.5 The grid data of the concentration is input into the first feature extraction layer, and the PM output by the first feature extraction layer is obtained. 2.5 Characteristic information of concentration; The grid data of each influencing factor is input into the second feature extraction layer to obtain feature information of each influencing factor output by the second feature extraction layer.

3. The atmospheric fine particulate matter PM according to claim 1 2.5 The impact factor analysis method is characterized by: The impact analysis layer includes an impact degree analysis layer and an impact mode analysis layer; Correspondingly, the PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the impact analysis layer, and the influence of each influencing factor on the PM output by the impact analysis layer is obtained. 2.5 The extent and manner of the impact of concentration include: The PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the influence degree analysis layer, and the influence degree analysis layer outputs the influence factors on the PM 2.5 The degree of influence of concentration; The PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the influencing mode analysis layer, and the influencing factors output by the influencing mode analysis layer on the PM are obtained. 2.5 How concentration affects the 4. The atmospheric fine particulate matter PM according to claim 3 2.5 The impact factor analysis method is characterized by: The impact degree analysis layer includes a time impact degree analysis layer and a space impact degree analysis layer; Correspondingly, the PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the influence degree analysis layer, and the influence degree analysis layer outputs the influence factors on the PM 2.5 The degree of influence of concentration includes: The PM 2.5 The time characteristics of the concentration and the time characteristics of each influencing factor are input into the time influence degree analysis layer, and the time influence degree analysis layer outputs the influence factors on the PM 2.5 The extent of the influence of concentration on the time dimension; The PM 2.5 The spatial characteristics of the concentration and the spatial characteristics of each influencing factor are input into the spatial influence degree analysis layer, and the spatial influence degree analysis layer outputs the influence factors on the PM 2.5 The extent of the concentration effect in the spatial dimension.

5. The atmospheric fine particulate matter PM according to claim 3 2.5 The impact factor analysis method is characterized by: The impact mode analysis layer includes a time impact mode analysis layer and a space impact mode analysis layer; Correspondingly, the PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the influencing mode analysis layer, and the influencing factors output by the influencing mode analysis layer on the PM are obtained. 2.5 Concentrations are affected by: The PM 2.5 The time characteristics of the concentration and the time characteristics of each influencing factor are input into the time influence mode analysis layer, and the time influence mode analysis layer outputs the influence factors on the PM 2.5 How concentration affects the time dimension; The PM 2.5 The spatial characteristics of the concentration and the spatial characteristics of each influencing factor are input into the spatial influencing mode analysis layer, and the spatial influencing mode analysis layer outputs the influencing factors on the PM 2.5 How concentration affects spatial dimensions.

6. The atmospheric fine particulate matter PM according to claim 1 2.5 The impact factor analysis method is characterized by: Inputting the spatial heterogeneity information of each influencing factor into the clustering module to obtain the spatial distribution pattern of each influencing factor output by the clustering module includes: According to the spatial heterogeneity information of each influencing factor, the influence of each influencing factor on PM in each grid of the target area is obtained. 2.5 The extent of the effect of concentration; Based on the clustering module, the K-means algorithm is used to classify the PM according to the influencing factors of each grid. 2.5 The grids are clustered according to the degree of effect of concentration to obtain a plurality of grid sets, and the spatial distribution pattern of the influencing factors is determined according to the plurality of grid sets.

7. The atmospheric fine particulate matter PM according to any one of claims 1 to 6 2.5 The impact factor analysis method is characterized by: Preprocessing the multi-source heterogeneous data to obtain grid data of preset spatiotemporal resolution includes: Preprocessing the natural environment data to obtain natural environment impact factors; Preprocessing the social and economic activity data to obtain social and economic activity impact factors; For the PM 2.5 The spatial distribution data of concentration, natural environment influencing factors and socio-economic activity influencing factors are preprocessed to obtain grid data with preset spatiotemporal resolution.

8. The atmospheric fine particulate matter PM according to claim 1 2.5 The impact factor analysis method is characterized by: The training process of the impact factor analysis model includes: Acquire sample multi-source heterogeneous data of the sample area within the sample time period, wherein the sample multi-source heterogeneous data includes PM 2.5 Concentration sample spatial distribution data, sample natural environment data, and sample socio-economic activity data; The sample multi-source heterogeneous data is preprocessed to obtain sample grid data with a preset spatiotemporal resolution, wherein the sample grid data includes PM 2.5 Sample grid data for concentration and PM 2.5 Sample grid data of each sample influencing factor; Determine the spatial distribution pattern label of each sample impact factor; According to the PM 2.5 The sample grid data of concentration and the sample grid data of each sample influencing factor are used as training samples, and the spatial distribution pattern labels of each sample influencing factor are used as sample labels to train the initial influencing factor analysis model. After the initial influencing factor analysis model is trained, the influencing factor analysis model is obtained.

9. PM, a type of atmospheric fine particulate matter 2.5 The impact factor analysis device is characterized by: include: An acquisition unit is used to acquire multi-source heterogeneous data of a target area in a target time period, wherein the multi-source heterogeneous data includes atmospheric fine particulate matter PM 2.5 Spatial distribution data of concentration, natural environment data and socio-economic activity data; A preprocessing unit is used to preprocess the multi-source heterogeneous data to obtain grid data with a preset spatiotemporal resolution, wherein the grid data includes PM 2.5 Gridded data of concentration and PM 2.5 Grid data of each influencing factor; A parsing unit for converting the PM 2.5 The grid data of the concentration and the grid data of each influencing factor are input into a pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model; The impact factor analysis model is based on the PM of the sample area. 2.5 Sample grid data for concentration and PM 2.5 The sample grid data of each sample influencing factor and the spatial distribution pattern label of each sample influencing factor are trained; The impact factor analysis model includes a feature extraction layer, an impact analysis layer, a multi-scale geographically weighted regression module and a clustering module; Correspondingly, the PM 2.5 The grid data of the concentration and the grid data of each influencing factor are input into a pre-trained influencing factor analysis model to obtain the spatial distribution pattern of each influencing factor output by the influencing factor analysis model, including: The PM 2.5 The grid data of concentration and the grid data of each influencing factor are input into the feature extraction layer to obtain the PM output by the feature extraction layer. 2.5 Characteristic information of concentration and characteristic information of each influencing factor; The PM 2.5 The characteristic information of the concentration and the characteristic information of each influencing factor are input into the impact analysis layer, and the influence of each influencing factor on the PM output by the impact analysis layer is obtained. 2.5 The extent and manner of the effect of concentration; The above mentioned influencing factors are applied to the PM 2.5 The influence degree and influence mode of the concentration are input into the multi-scale geographical weighted regression module to obtain the spatial heterogeneity information of the influencing factors output by the multi-scale geographical weighted regression module, wherein the spatial heterogeneity information includes the nature of action, the degree of action and the spatial multi-scale property; The spatial heterogeneity information of each influencing factor is input into the clustering module to obtain the spatial distribution pattern of each influencing factor output by the clustering module.

Images