Atmospheric pollution tracing method based on multi-scale ecological environment big data integration

Abstract

The invention discloses an air pollution tracing method based on multi-scale ecological environment big data integration, and belongs to the technical field of environmental protection. According to the method, multi-source information such as satellite remote sensing, ground monitoring, meteorological data and emission data is integrated, and algorithms such as data fusion, machine learning classification, clustering analysis, regression analysis and trajectory tracking are combined, so that accurate identification, dynamic tracking and visual display of the atmospheric pollution source are realized. According to the method, the problems of limited coverage, low traceability precision, poor timeliness and the like of a traditional monitoring method are solved, and scientific and effective decision support can be provided for an environment management department.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of environmental protection, and particularly relates to an atmospheric pollution tracing method based on multi-scale ecological environment big data integration. BACKGROUND

[0002] Atmospheric pollution is a major global environmental problem, posing a serious threat to human health and the ecological environment. Traditional monitoring methods mainly rely on limited ground monitoring station data, which has the problems of limited coverage, poor timeliness, and low tracing accuracy. According to statistics, about 7 million people die from air pollution-related diseases every year, especially in developing countries.

[0003] In the prior art, CN114386735A discloses an atmospheric pollution tracing method based on big data analysis, which realizes pollution tracing by setting factor over-limit threshold and applying label algorithm. However, this method still has deficiencies in data fusion accuracy, multi-scale analysis capability, and real-time dynamic optimization, especially in handling complex weather conditions and multi-source data collaborative analysis.

[0004] With the development of Internet of Things technology and big data analysis technology, how to fully utilize multi-source monitoring data to achieve accurate atmospheric pollution tracing and real-time early warning has become the focus of current technological development. Therefore, there is an urgent need in the field for an atmospheric pollution tracing method that can fully utilize multi-scale ecological environment big data and combine advanced machine learning algorithms to achieve high-precision, real-time dynamic atmospheric pollution tracing. SUMMARY

[0005] To achieve the above purpose, the present application provides the following technical solutions: A method for tracing atmospheric pollution based on multi-scale ecological environment big data integration, comprising the following steps: S1: Real-time collection of multi-source ecological environment data including meteorological data, ground monitoring data, satellite remote sensing data, and industrial emission data through an Internet of Things monitoring device network; S2: Preprocessing of collected data: cleaning, outlier detection, and standardization of collected raw data, and sorting and classifying collected multi-scale data for subsequent data processing and analysis; S3: Fusion analysis of preprocessed data: using a geographic information system and adopting a weighted interpolation method to realize multi-source data fusion calculation, and constructing a unified atmospheric pollution big data set from the preprocessed data, the calculation formula of the interpolation method being as follows: I(x,y,t)=∑I i (x,y,t)∗w i (x,y,t) wherein I(x, y, t) represents the fused data value, I i (x, y, t) represents the data value of the i-th data source, w i (x, y, t) represents the weight value of the i-th data source at the current position, x represents the horizontal coordinate, y represents the vertical coordinate, and t represents the time; Support vector machine is used to classify and identify the type of pollution source: the feature vector is classified by constructing a classification hyperplane, and the optimization objective of support vector machine classification is to minimize the following loss function: min‖w‖ ²+∑ α i wherein w represents the weight vector of the support vector machine, α i represents the Lagrange multiplier corresponding to the i-th support vector; i The Euclidean distance matrix D is calculated by clustering analysis and modeling to determine the number and spatial aggregation characteristics of the pollution sources: D=sqrt(∑(x i −x j )²+(y i −y j )²) wherein (x i ,y i ) and (x j ,y j ) represent the coordinates of two pollution sources, respectively; S4: The particle filter algorithm is used to simulate and calculate the diffusion path of the pollutant particles to obtain the pollutant diffusion model, and the displacement formula of the pollutant particles is: Δx=v×Δt×cosθ Δy=v×Δt×sinθ wherein v represents the wind speed, Δt represents the diffusion time, and θ represents the wind direction; S5: Through the pollutant concentration monitoring data and the pollutant diffusion model, the source intensity of the pollution source is inversed by using regression analysis, and the optimization objective of linear regression analysis is to minimize the sum of squared residuals: min∑ (y i −β0+β1∗x i )² wherein y i represents the pollution concentration observation value, x i represents the pollution source emission amount, β0 and β1 represent the regression coefficients, respectively.​ S6: Visualizing the results of the traceability.

[0006] Further, the ecological environment data in step S1 includes temperature, humidity, wind speed, wind direction meteorological data, SO2, NOx, PM2.5 ground monitoring data, vegetation index, land surface temperature remote sensing data, and emission, emission height pollution source data.

[0007] Further, in step S3, the weight value of each data source is determined according to the spatial correlation, temporal correlation and data quality of the monitoring data. The stronger the correlation, the greater the weight value. The parameters in the fusion calculation are dynamically adjusted according to the real-time monitoring data.

[0008] Further, the data identified by the support vector machine classification in step S3 includes historical monitoring data and real-time monitoring data.

[0009] Further, the visualization in step S6 includes generating a pollution source distribution map, drawing a pollution source movement trajectory, and making an emission intensity heat map.

[0010] Compared with the prior art, the beneficial effects of the present application mainly include: 1. Improve the accuracy of traceability: multi-source data fusion and machine learning algorithm combination, break through the limitation of traditional monitoring point, accurate positioning of pollution source type, location and intensity.

[0011] 2. Enhance timeliness: real-time data access and model dynamic optimization, realize minute-level tracking of pollution diffusion, support early warning and control.

[0012] 3. Support dynamic decision-making: visualize pollution source distribution and diffusion path, provide intuitive basis for environmental protection departments, and help "precise pollution control".

[0013] 4. Universality and expansibility: modular design of algorithm, can adapt to different regions (city, region, global) and scenes (industrial source, traffic source, biomass source), wide application range. DETAILED DESCRIPTION

[0014] Example 1 S1: In a city area, we collected PM2.5 ground monitoring data of different monitoring points within a week and corresponding satellite remote sensing data, as well as meteorological data including wind speed and wind direction.

[0015] S2: select five ground monitoring points, at a time t, the coordinates of each monitoring point are (x1, y1), (x2, y2), (x3, y3), (x4, y4), (x5, y5), and the corresponding PM2.5 monitoring values are Ig1(x1, y1, t) = 35, Ig2(x2, y2, t) = 42, Ig3(x3, y3, t) = 38, Ig4(x4, y4, t) = 45, Ig5(x5, y5, t) = 40, where g represents ground monitoring data, and the unit is μg / m 3 .

[0016] The overall PM2.5 distribution data of the region is obtained by satellite remote sensing, and after preprocessing, the remote sensing data values corresponding to the above five monitoring point positions are Ir1(x1, y1, t) = 32, Ir2(x2, y2, t) = 40, Ir3(x3, y3, t) = 36, Ir4(x4, y4, t) = 43, Ir5(x5, y5, t) = 39, where r represents remote sensing data, and the unit is μg / m 3 .

[0017] S3: using geographic information system, according to the representative factors of monitoring points and surrounding areas, set the weight of each data source, assuming that the weight of ground monitoring data wg1=0.6, wg2=0.6, wg3=0.6, wg4=0.6, wg5=0.6, and the weight of remote sensing data wr1=0.4, wr2=0.4, wr3=0.4, wr4=0.4, wr5=0.4.

[0018] According to the data fusion formula I(x, y, t) = ∑iI i (x, y, t) × w i (x, y, t), calculate the fused PM2.5 data value: For monitoring point 1: I(x1, y1, t) = I1(x1, y1, t) × wg1 + Ir1(x1, y1, t) × wr1 = 35 × 0.6 + 32 × 0.4 = 21 + 12.8 = 33.8 (μg / m3) Similarly, calculate the fused data value of other monitoring points: Monitoring point 2: I(x2, y2, t) = 42 × 0.6 + 40 × 0.4 = 25.2 + 16 = 41.2 (μg / m3) Monitoring point 3: I(x3, y3, t) = 38 × 0.6 + 36 × 0.4 = 22.8 + 14.4 = 37.2 (μg / m3) Monitoring point 4: I(x4, y4, t) = 45 × 0.6 + 43 × 0.4 = 27 + 17.2 = 44.2 (μg / m3) Monitoring point 5: I(x5, y5, t) = 40 x 0.6 + 39 x 0.4 = 24 + 15.6 = 39.6 (μg / m3); Then the support vector machine is used to classify and identify the pollution source type: the feature vector is classified by constructing a classification hyperplane to determine the type of pollution source, and the optimization objective of support vector machine classification is to minimize the following loss function: min ‖w‖ ²+∑ α i Wherein, w represents the weight vector of the support vector machine, α i represents the Lagrange multiplier corresponding to the i th support vector; Then the Euclidean distance matrix D is calculated by clustering analysis and modeling to determine the number and spatial aggregation characteristics of the pollution sources: D=sqrt(∑(x i −x j )²+(y i −y j )²) Wherein, (x i ,y i ) and (x j ,y j ) represent the coordinates of two pollution sources respectively; S4: Assuming that at the initial time t0, a pollution aggregation is found at a point (x0, y0) on the edge of the region, a plurality of particles are released from the point as the starting point to simulate the trajectory tracking of the pollution particles. It is known that the average wind speed v = 2 m / s and the wind direction is northeast, that is, the angle θ = 45 with the east direction.

[0019] After a time interval Δt = 1 s, according to the wind speed and wind direction, the components of the particle displacement in the x direction and the y direction are respectively: Δx=v×Δt×cosθ=2×1×cos45∘=2≈1.414m Δy=v×Δt×sinθ=2×1×sin45∘=2≈1.414m Assuming that the initial particle position is (x0, y0) = (100 m, 100 m), then after Δt = 1 s, the particle position is updated to (x1, y1) = (100 + 1.414, 100 + 1.414) = (101.414 m, 101.414 m).

[0020] With the passage of time, the particle position is constantly updated according to new wind speed, wind direction and other meteorological data, and at the same time, the particle trajectory is constantly adjusted in combination with the subsequent time point of each monitoring point of the pollutant concentration change data, so as to speculate the possible moving path of the pollution source.

[0021] S5: A suspected pollution source area is preliminarily determined through trajectory analysis, and three pollution sources with known emission characteristics, such as industrial pollution sources, are selected in the area, and their emission amounts x1=10kg / h, x2=15kg / h, x3=20kg / h are recorded respectively, and the PM2.5 concentration observation values y1=50μg / m 3 , y2=60μg / m 3 , y3=70μg / m 3 near these pollution sources at the corresponding time are obtained through ground monitoring. We assume that there is a linear relationship between PM2.5 concentration y and pollution source emission amount x y=β0+β1*x, According to the regression analysis formula min∑i(y i −(β0+β1×x i )) 2 , We can solve β0 and β1 by least square method. Calculate the relevant data: ∑i=βx i =10+15+20=45 ∑i=βy i =50+60+70=180 ∑i=βx i 2 =10 2 +15 2 +20 2 =100+225+400=725 ∑i=βx i y i =10×50+15×60+20×70=500+900+1400=2800 n=3 According to the formula: β1=n∑i=lnx i 2−(∑i=1nx i )2n∑i=1nx i y i −∑i=1nx i ∑i=1ny i β1=3×725−4523×2800−45×180=2175−20258400−8100=150300=2 β0=n∑i=1ny i −β1∑ i =1nx i=3180−2×45=3180−90=30 Therefore, the obtained linear regression equation is y=30+2x. If a new suspected pollution source is subsequently discovered in this area, and its emissions are estimated using other methods to be x4=12kg / h, then the PM2.5 concentration that may be generated near the pollution source can be predicted as y4=30+2×12=54μg / m3. By comparing with actual monitoring data, the model can be further verified and adjusted, and a more accurate pollution source emission intensity can be calculated in reverse.

[0022] S6: Visual display of the traceability results.

[0023] The method provided by this invention fully utilizes multi-scale ecological and environmental big data, improving the accuracy and reliability of air pollution source tracing and contributing to refined management and precise pollution control. Furthermore, this invention employs advanced data mining and machine learning techniques, possessing strong universality and scalability, and can be applied to air pollution source tracing analysis in different regions and scenarios.

Claims

1. A method for atmospheric pollution source tracing based on multi-scale ecological environment big data integration, characterized in that, Comprise the following steps: S1: through the Internet of Things monitoring device network, real-time collection of multi-source ecological environment data including meteorological data, ground monitoring data, satellite remote sensing data, industrial emission data; S2: preprocessing of the collected data: cleaning, outlier detection and standardization processing of the collected raw data, sorting and classification of the collected data at all levels; S3: fusion analysis of the preprocessed data: using geographic information system, combined with weighted interpolation method to realize the fusion calculation of multi-source data, the preprocessed data is constructed unified atmospheric pollution big data set, the calculation formula of interpolation method is as follows: wherein I(x, y, t) represents the fused data value, I i (x, y, t) represents the data value of the i-th data source, w i (x, y, t) represents the weight value of the i-th data source at the current position, x represents the horizontal coordinate, y represents the vertical coordinate, and t represents the time. Then use support vector machine to classify and identify pollution source type: realize feature vector classification by constructing classification hyperplane: wherein w denotes a weight vector of the support vector machine, a i denotes the Lagrange multiplier corresponding to the i-th support vector; Then calculate the Euclidean distance matrix D through clustering analysis and modeling to determine the number and spatial aggregation characteristics of pollution sources: ； where (x i ,y i ) and (x j ,y j ) represent the coordinates of the two pollution sources, respectively; S4: using particle filtering algorithm to simulate and calculate the diffusion path of pollutant particles, get the pollutant diffusion model, the displacement formula of pollutant particles is: ; S5: Through the pollutant concentration monitoring data and the pollutant diffusion model, the emission intensity of the pollution source is inversely calculated by regression analysis: where y i represents the pollution concentration observation value, x i represents the pollution source emission, and β0 and β1 represent the regression coefficients, respectively; S6: visual display of the realization of traceability results. 2.The atmospheric pollution source tracing method based on multi-scale ecological environment big data integration according to claim 1, characterized in that, The ecological environment data in step S1 includes meteorological data such as temperature, humidity, wind speed and wind direction, ground monitoring data such as SO2, NOx and PM2.5, remote sensing data such as vegetation index and land surface temperature, and pollution source data such as emission amount and emission height. 3.The atmospheric pollution source tracing method based on multi-scale ecological environment big data integration according to claim 1, characterized in that, In step S3, the weight value of each data source is determined according to the spatial correlation, time correlation and data quality of the monitoring data, and the parameters in the fusion calculation are dynamically adjusted according to the real-time monitoring data. 4.The atmospheric pollution source tracing method based on multi-scale ecological environment big data integration of claim 1, characterized in that, The data for support vector machine classification and identification in step S4 includes historical monitoring data and real-time monitoring data.

5. The method according to claim 1, wherein, The visual display in step S8 includes generating pollution source distribution map, drawing pollution source moving track and making emission intensity heat map.

Images