A city atmospheric pollution source positioning method based on intelligent tower

Abstract

This invention discloses a method for locating urban air pollution sources based on smart poles, belonging to the field of data analysis. Addressing the problem that existing sensors on smart poles can only detect air pollutant concentrations but lack accurate source location, this invention proposes a method for locating urban pollution sources using smart poles. Environmental pollution is a dynamic process; emissions and air transport patterns change over time. Using the LSTM algorithm, the collected concentration distribution map is extended in the time domain. Finally, data from different time periods are compared and analyzed to obtain the dynamic changes of environmental pollution sources. This allows for a better understanding of environmental pollution trends, enabling targeted measures for control and prevention.

Description

Technical Field

[0001] This invention relates to the field of gas diffusion, and more specifically to a method for locating urban air pollution sources based on smart poles. Background Technology

[0002] With rapid urban development, while enjoying the convenience brought by high technology, humanity also faces increasingly severe environmental problems. Urban pollution, especially air pollution, has become a major concern. Accurate and efficient monitoring and location of urban air pollution sources has become an urgent issue. Although various traditional monitoring methods (such as fixed monitoring stations and mobile vehicles equipped with monitoring equipment) can detect and locate pollution sources to some extent, these methods generally suffer from limited coverage, low real-time performance, and high costs, making it difficult to meet the current requirements for real-time, large-area, and low-cost environmental monitoring.

[0003] Against this backdrop, the emergence of smart pole technology offers new possibilities for addressing this situation. A smart pole is a tool integrating multiple functions such as LED lighting, video surveillance, advertising, and environmental monitoring. Relying on advanced communication technology, it enables data transmission and interaction between poles. The design of smart poles incorporates optimized layouts tailored to urban traffic characteristics. Not only are all poles evenly distributed, but they are also predominantly located along both sides of roads. In complex urban environments, wind direction at lower elevations is often aligned with the road due to building obstruction, improving positioning accuracy while reducing the complexity of model calculations. Furthermore, because smart poles inherently possess data interconnectivity, they can capture real-time data changes and interact with other pole nodes when locating air pollution sources, ensuring real-time and accurate positioning. In summary, smart pole technology is a new technology that has emerged alongside smart city construction. It makes large-scale, real-time monitoring and location of air pollution sources possible, playing a positive role in promoting environmental protection.

[0004] Leveraging the advantages of smart poles—large coverage area, high distribution density, reasonable installation locations, and fast data transmission—this invention proposes a method for locating urban pollution sources using smart poles, involving Gaussian plume models and Kriging interpolation. Gaussian plume models are mathematical models used to estimate the transport and diffusion of pollutants in the atmosphere. However, Gaussian plume models typically do not consider the influence of terrain and buildings on airflow and pollutant transport; therefore, their predictions may be inaccurate in complex terrain and urban environments. Kriging interpolation is a spatial interpolation method that uses observed values ​​at known points to infer values ​​at unknown points. This invention combines previous research, integrating Gaussian plume models and Kriging interpolation. The Gaussian plume model is used to roughly divide the distribution of pollution sources, and then Kriging interpolation is used to more accurately determine the distribution. The determination of interpolation points requires special processing based on the geographical characteristics of the urban environment. Finally, the time factor is incorporated into the model, and a neural network is used to extend the existing distribution map in the time domain to obtain the future trend of pollutant distribution. This can provide more accurate and realistic predictions of the spatiotemporal distribution of smoke concentration, enabling a better understanding and management of the smoke diffusion process. Summary of the Invention

[0005] The purpose of this invention is to address the problem that existing sensors on smart poles can only detect the concentration of air pollutants but lack accurate location of air pollution sources, and to propose a method for locating urban pollution sources on smart poles.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] A method for locating air pollution sources, comprising the following steps:

[0008] This facilitates data acquisition by sensors on smart poles;

[0009] A rough distribution map of pollutants was obtained using a Gaussian plume model;

[0010] Kriging interpolation was used to accurately locate pollutant distribution maps. Sensor data and interpolation points near high-rise buildings in the city were specially processed to obtain interpolation results that best match the concentration and wind direction.

[0011] By using neural networks to extend the results in the time domain, we can obtain predictions of the spatiotemporal distribution of pollutants.

[0012] Step 1: Deploy multiple sensors on the corresponding smart tower nodes according to preset requirements to collect real-time monitoring data on multiple pollutants, wind speed, and wind direction. The sensor deployment takes into account the distribution of pollution sources and environmental characteristics within the area to ensure the accuracy and comprehensiveness of the data. In cities, sensors are deployed more densely on smart towers in traffic-intensive areas, industrial zones, and densely populated areas.

[0013] Step 2: Use the improved Gaussian plume model to roughly divide the distribution map of pollution sources. The Gaussian plume model is a commonly used air pollution transport model. Based on fluid dynamics principles and the corresponding mass conservation equation, it can estimate the transport and diffusion of pollutants in the atmosphere. The improved Gaussian model can more accurately predict the physical model of chimney emissions and is more consistent with the actual factory emissions. The formula is shown in equation (1).

[0014]

[0015] In the formula, C (x,y,0) Let (x, y, 0) be the pollutant concentration at any point (x, y, 0) below the pollution source, in mg / m³. 3 Q represents the source strength. P represents the wind speed at a specified altitude, a function of atmospheric stability, which can be found in the environmental impact assessment guidelines. U0 represents the wind speed at an altitude of Z0 meters, in m / s. σ y and σ z These are the lateral diffusion parameter and the vertical diffusion parameter, respectively; σ x Here, A is the longitudinal diffusion parameter; A is the vertical variation parameter of wind speed, expressed as follows:

[0016] A = (P+1) × 10 P (2)

[0017] H e Let the effective height of the chimney be m, and the formula is:

[0018]

[0019] Where H s Where is the geometric height of the chimney, in meters; W is the lift parameter, which is related to the dynamic and thermal conditions of the flue gas.

[0020] Step 3: Perform appropriate Kriging interpolation on some blank points on the map. The Kriging interpolation formula is as follows:

[0021]

[0022] Where λ i The core idea of ​​the Kriging algorithm is that the difference between two attributes is positively correlated with the distance between them within a certain range, where the weighting coefficients are to be determined. In the above equation, Z(x0) is required to be an unbiased estimate that minimizes the variance of the estimate under the unbiased condition, i.e., it must satisfy the following equation.

[0023]

[0024] Where μ is a Lagrange multiplier.

[0025] Then, the weight coefficients are solved, and the covariance function is replaced by the variogram function. The system of equations is as follows:

[0026]

[0027] The Gaussian variogram with a sill value is chosen as the variogram because it better describes multi-scale diffusion processes and captures the gradual weakening of the gas relative to the source. The Gaussian variogram has a smooth shape and is suitable for diffusion in the near-source region and at medium distances. The formula is as follows:

[0028]

[0029] Step 4: Special processing of interpolation points. Due to the complexity of the urban environment, the selection of interpolation points is particularly important. Because the height of sensors in urban areas is limited, gas blockage can easily occur near tall buildings. Geographic correlation is closely related to the orientation of buildings near sensors close to tall buildings. Therefore, in addition to the existing rectangular interpolation points, interpolation points inside buildings are removed, and points within the building area are considered to have a concentration of 0. Furthermore, to ensure the effectiveness of interpolation points around buildings, special processing is performed during interpolation.

[0030] First, determine whether the real node data supports Kriging interpolation. First, use the values ​​of neighboring nodes without building obstruction to predict the node to be interpolated. For the values ​​between neighboring nodes and the node to be interpolated, the concentration is predicted according to traditional Kriging interpolation.

[0031] Then, for nodes with building obstruction, it is not possible to directly predict the node to be interpolated. Therefore, we find other points that can be jointly predicted. The jointly predicted point is not obstructed by buildings from the node to be interpolated. The concentration of the jointly predicted point is calculated by selecting two nodes, one of which is obstructed by buildings from the node to be interpolated.

[0032] Step 5: Combine the data obtained from S1 and S2 using a hybrid model, based on the specific environment, to obtain the final result. The principle of the hybrid model is as follows:

[0033] 1) Data normalization: Normalize the data obtained from both sources so that they can be analyzed in the same environment.

[0034] 2) Data merging: When merging data, the closer the value is to the building or the actual sensor node, the greater the prediction weight of Kriging interpolation. The weight is selected using the Gaussian kernel function.

[0035] Step 6: Based on the approximate division obtained from the Gaussian plume model and the detailed distribution obtained from Kriging interpolation, and combined with the functions of the geographic information system, generate the final pollution source distribution map.

[0036] Step 7: Extend the data obtained in S5 in the time domain to obtain a dynamic distribution map. Environmental pollution is a dynamic process; emissions and airborne transport of pollutants change over time. Using the LSTM algorithm, the collected concentration distribution map is extended in the time domain. Finally, data from different time periods are compared and analyzed to obtain the dynamic changes of environmental pollution sources. This allows for a better understanding of environmental pollution trends and enables targeted measures for control and prevention. Attached Figure Description

[0037] Figure 1 This invention provides a diagram illustrating the location scheme for atmospheric pollution sources.

[0038] Figure 2 A flowchart illustrating the specific solution provided by this invention;

[0039] Figure 3 This invention provides a schematic diagram of the virtual node interpolation principle next to obstacles. Detailed Implementation

[0040] Step 1: Deploy multiple sensors on the corresponding smart tower nodes according to preset requirements to collect real-time monitoring data on multiple pollutants, wind speed, and wind direction. The sensor deployment takes into account the distribution of pollution sources and environmental characteristics within the area to ensure the accuracy and comprehensiveness of the data. In cities, sensors are deployed more densely on smart towers in traffic-intensive areas, industrial zones, and densely populated areas.

[0041] Step 2: Use the improved Gaussian plume model to roughly divide the distribution map of pollution sources. The Gaussian plume model is a commonly used air pollution transport model. Based on fluid dynamics principles and corresponding mass conservation equations, it can estimate the transport and diffusion of pollutants in the atmosphere. The improved Gaussian model can more accurately predict the physical model of chimney emissions, and is more consistent with actual factory emissions. (Formula...) [1] as follows

[0042]

[0043] In the formula, C (x,y,0) Let (x, y, 0) be the pollutant concentration at any point (x, y, 0) below the pollution source, in mg / m³. 3Q represents the source strength. P represents the wind speed at a specified altitude, a function of atmospheric stability, which can be found in the environmental impact assessment guidelines. U0 represents the wind speed at an altitude of Z0 meters, in m / s. σ y and σ z These are the lateral diffusion parameter and the vertical diffusion parameter, respectively; σ x Here, A is the longitudinal diffusion parameter; A is the vertical variation parameter of wind speed, expressed as follows:

[0044] A = (P+1) × 10 P (9)

[0045] H e Let the effective height of the chimney be m, and the formula is:

[0046]

[0047] Where H s Where is the geometric height of the chimney, in meters; W is the lift parameter, which is related to the dynamic and thermal conditions of the flue gas.

[0048] Step 3: Perform appropriate Kriging interpolation on some blank points on the map. The Kriging interpolation formula is as follows:

[0049]

[0050] Where λ i The core idea of ​​the Kriging algorithm is that the difference between two attributes is positively correlated with the distance between them within a certain range, where the weighting coefficients are to be determined. In the above equation, Z(x0) is required to be an unbiased estimate that minimizes the variance of the estimate under the unbiased condition, i.e., it must satisfy the following equation.

[0051]

[0052] Where μ is a Lagrange multiplier.

[0053] Then, the weight coefficients are solved, and the covariance function is replaced by the variogram function. The system of equations is as follows:

[0054]

[0055] The Gaussian variogram with a sill value is chosen as the variogram because it better describes multi-scale diffusion processes and captures the gradual weakening of the gas relative to the source. The Gaussian variogram has a smooth shape and is suitable for diffusion in the near-source region and at medium distances. The formula is as follows:

[0056]

[0057] Step 4: Special processing of interpolation points. Due to the complexity of the urban environment, the selection of interpolation points is particularly important. Because the height of sensors in urban areas is limited, gas blockage can easily occur near tall buildings. Geographic correlation is closely related to the orientation of buildings near sensors close to tall buildings. Therefore, in addition to the original rectangular interpolation points, interpolation points inside buildings are removed, and points within the building area are considered to have a concentration of 0. Furthermore, to ensure the effectiveness of interpolation points around buildings, the following processing is performed during interpolation:

[0058] 1) such as Figure 3 As shown, for the points in the red area of ​​the figure, we first determine whether the real node data supports Kriging interpolation. We first use the value of node 1 to predict the concentration, and then use traditional Kriging interpolation to predict the concentration of the values ​​on this side.

[0059] 2) Due to the obstruction of buildings, the red point cannot be directly predicted for nodes 2, 3, and 4. Therefore, we need to find other points that can be predicted together, such as green points. We use points 1 and 2 to predict the value of the green point, and then use point 1 and the green point to jointly predict the value of the red point. In order to ensure the effectiveness of the red prediction point, at least 5 green points are needed to predict each red point.

[0060] Step 5: Combine the data obtained from S1 and S2 using a mixture model to obtain the final result. The principle of the mixture model is as follows:

[0061] 1) Data normalization: The data obtained from both methods are normalized to ensure they can be analyzed under the same conditions. The normalization model uses min-max normalization, as shown in the following formula:

[0062]

[0063] 2) Data merging: When merging data, the closer the value is to the building or the actual sensor node, the greater the prediction weight of the Kriging interpolation. The weight is selected using the Gaussian kernel function, as shown in the following formula.

[0064]

[0065] x′ is Kernel function The center, i.e., the building or the actual sensor node; ||xx′|| 2 Let σ be the Euclidean distance (L2 norm) between vectors x and x′. As the distance between the two vectors increases, the Gaussian kernel function monotonically decreases. σ controls the range of influence of the Gaussian kernel function; the larger its value, the greater the local influence range of the Gaussian kernel function.

[0066] Step 6: Based on the approximate division obtained from the Gaussian plume model and the detailed distribution obtained from Kriging interpolation, and combining the functions of a Geographic Information System (GIS), generate the final pollution source distribution map. The specific steps are as follows:

[0067] 1) Use GIS software (such as ArcGIS, QGIS, etc.) to create a new project and import the collected data into the project. Ensure that the geographic location information is in the correct data format and is correlated with the pollution source concentration data, and then create layers based on this.

[0068] 2) Based on the approximate division data obtained from the Gaussian plume model, draw corresponding polygons or regions in GIS software to represent the location and extent of the pollution source. Drawing tools or spatial analysis tools can be used for this purpose.

[0069] 3) Load the interpolated pollution source concentration data and the pollution source delineation layer into the GIS software, and set the corresponding rendering style, such as using color gradients or contour lines to display different concentration levels. Other geographic information, such as road networks, rivers, and residential areas, can be added as needed to provide more comprehensive background information.

[0070] 4) Use the printing, exporting, or layer saving functions in GIS software to generate the final pollution source distribution map. You can choose to output as an image, PDF file, or interactive web map.

[0071] Step 7: Extend the data obtained in Step 6 in the time domain to obtain a dynamic distribution map. Environmental pollution is a dynamic process; emissions from pollution sources and airborne transport change over time. Using the LSTM algorithm, the collected concentration distribution map is extended and predicted in the time domain. Finally, data from different time periods are compared and analyzed to obtain the dynamic changes of environmental pollution sources.

Claims

1. A method for locating urban atmospheric pollution sources based on a smart tower, characterized in that, Includes the following steps: Step 1: Deploy multiple sensors on the corresponding smart tower nodes according to the preset requirements to collect multiple pollution factors, wind speed and wind direction as monitoring data in real time; Step 2: Divide the distribution map of pollution sources using the following formula; ; In the formula, Any point below the pollution source Pollutant concentration at the location, in units ; For the source of strength; The altitude at which wind speed is assigned is a function of atmospheric stability, which can be found in the Environmental Impact Assessment Guidelines; For height is Wind speed at a distance of meters, unit , This represents the aerodynamic coefficient, obtained from experimental measurements or numerical simulations. and These are the lateral diffusion parameters and the vertical diffusion parameters, respectively. express of Power of; Here, A is the longitudinal diffusion parameter; A is the vertical variation parameter of wind speed, and the formula is: ； The effective height of the chimney, in units of The formula is: ； in The geometric height of the chimney, in units of ; For pick-up parameters; Step 3: Perform Kriging interpolation on some blank points on the map, using a Gaussian variogram with sill values ​​as the variogram function. Step 4: Near sensors close to high-rise buildings, geographical correlation is closely related to building orientation. Based on this, on the original rectangular interpolation points, interpolation points inside the building are removed, and points within the building area are considered to have a concentration of 0. The specific interpolation method is as follows: Above the existing rectangular interpolation points, interpolation points inside the building are removed, and points within the building area are considered to have a concentration of 0. Meanwhile, to ensure the validity of interpolation points around the building, the following processing is performed during interpolation: First, determine whether the real node data supports Kriging interpolation. First, use the values ​​of neighboring nodes without building obstruction to predict the node to be interpolated. For the values ​​between neighboring nodes and the node to be interpolated, the concentration is predicted according to traditional Kriging interpolation. Then, for nodes that are obstructed by buildings, it is not possible to directly predict the node to be interpolated. Therefore, we search for points that can be jointly predicted at other locations. These points are not obstructed by buildings from the node to be interpolated. The method for calculating the concentration of the common prediction point is as follows: select two nodes, namely node 1 and node 2, where node 2 is obstructed by a building from the node to be interpolated. Use the two selected nodes to calculate the concentration of the common prediction point; then calculate the concentration of the node to be interpolated using node 1 and the common prediction point. Step 5: Combine the data obtained in Steps 1 and 2 using a hybrid model based on the specific environment to obtain the final result; generate the final pollution source distribution map by combining the approximate division obtained from the Gaussian plume model and the detailed distribution obtained from Kriging interpolation with the functions of the geographic information system; and extend the obtained data in the time domain to obtain a dynamic distribution map.

2. The method for locating urban air pollution sources based on smart poles as described in claim 1, characterized in that: The smart pole includes a wind speed sensor, a wind direction sensor, and a sensor for the concentration of specified pollutants.

3. In the method for locating urban air pollution sources based on smart poles as described in claim 1, step 5 combines the data obtained in steps 2, 3, and 4 using a hybrid model to obtain the final result; the hybrid model is as follows: 1) Data normalization: The data obtained from both methods are normalized to ensure they can be analyzed under the same conditions. The normalization model uses min-max normalization, and the formula is as follows: ； 2) Data merging: When merging data, the closer the value is to the building or the actual sensor node, the greater the prediction weight of Kriging interpolation. The weight is selected using the Gaussian kernel function, as shown in the following formula. ； The kernel function center is either a building or a real sensor node; For vectors sum vector The Euclidean distance between two vectors decreases monotonically as the distance between them increases. The larger the value of the Gaussian kernel function, the greater its local influence range. Based on the distribution identified in Step 1, and utilizing the functions of the Geographic Information System (GIS), the final pollution source distribution map is generated; the specific steps are as follows: 1) Use GIS software to create a new project and import the collected data into the project; ensure that the data format of the geographic location information is correct and correlate it with the pollution source concentration data, and then create layers based on this; 2) Based on the division results in step 1, draw the corresponding polygons or regions in GIS software to represent the location and extent of the pollution source; 3) Load the interpolated pollution source concentration data and the pollution source classification layer into the GIS software and set the corresponding rendering style; add road network, river, and residential area information as needed; 4) Use the printing, exporting, or layer saving functions in GIS software to generate the final pollution source distribution map.

Images