A method for separating exogenous and local atmospheric pollution based on edge computing

Abstract

In the embodiment of the present invention, a method for separating exogenous and local air pollution based on edge computing is provided, which belongs to the field of computing technology, and specifically includes: monitoring concentration data of air pollutants and meteorological data and using them to train a random forest model; performing multiple random sampling of meteorological characteristics of meteorological data; inputting the sampled meteorological characteristics into the trained random forest model to obtain air pollutant data that eliminates the influence of meteorological factors; matching central cities with surrounding cities to form different city pairs between all cities; establishing a pollution diffusion function estimation model between cities; estimating the pollution diffusion function estimation model to obtain an existing pollution diffusion function; constructing a local air pollution concentration measurement model based on the existing pollution diffusion function and solving it, obtaining the local air pollution concentration of all cities and displaying it in a digital form. Through the scheme of the present invention, the recognition accuracy, adaptability and real-time performance are improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of computing technology, and in particular to a method for separating exogenous and local atmospheric pollution based on edge computing. Background Art

[0002] At present, atmospheric pollutants are affected by the horizontal movement of the atmosphere, which will form different degrees of cross-border transmission. For example, wind will blow pollution from upwind to downwind. The wind direction determines the direction of movement of pollutants, and the wind speed determines the speed of migration; at the same time, the different temperature, rainfall, air pressure, humidity and other factors in each region also have an impact on the monitoring of pollutant concentration data. Therefore, the pollutant concentration observed at the air pollution monitoring station not only includes the pollution generated locally, but also includes the pollution from other places transported by the wind, and the interference of factors such as temperature, rainfall, air pressure, humidity, etc. on the collected data cannot be ruled out.

[0003] The data errors collected by the monitoring stations and the phenomenon of cross-border transmission of pollution will cause trouble in identifying the emission reduction effects of air pollution prevention and control policies, making it challenging to effectively evaluate the completion of emission reduction work. Eliminating the interference of local climate on monitoring data, achieving lightweight stripping of external pollution, and identifying local pollution are the keys to accurately identifying the completion of emission reduction work. At the same time, the data of each monitoring station is complicated, and the monitoring data fails to eliminate the interference of local climate on the monitoring results. Centralized data processing will bring huge computing pressure, and the errors in monitoring data will directly interfere with the subsequent separation process.

[0004] It can be seen that there is an urgent need for a method for separating exogenous and local atmospheric pollution based on edge computing with high recognition accuracy, adaptability and real-time performance. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a method for separating exogenous and local atmospheric pollution based on edge computing, which at least partially solves the problems of poor recognition accuracy, adaptability and real-time performance in the prior art.

[0006] The embodiment of the present invention provides a method for separating exogenous and local air pollution based on edge computing, including:

[0007] Step 1, obtaining atmospheric pollutant monitoring concentration data and meteorological data of different cities;

[0008] Step 2: Use monitoring data and meteorological data to train the random forest model;

[0009] Step 3, under the condition that the time dimension remains unchanged, the meteorological characteristics of the meteorological data are randomly sampled multiple times;

[0010] Step 4: Input the sampled meteorological features into the trained random forest model to obtain the air pollutant data without the influence of meteorological factors and send it to the data center;

[0011] Step 5: Match the central city and the surrounding cities according to the geographical distance and terrain factors between the cities to form different city pairs between all the cities;

[0012] Step 6, establishing a pollution diffusion function estimation model between cities based on the wind direction data and wind speed data of the city pairs and each city;

[0013] Step 7, using air pollutant data to estimate the pollution diffusion function estimation model to obtain the existing pollution diffusion function;

[0014] Step 8: construct a local air pollution concentration calculation model based on the existing pollution diffusion function;

[0015] Step 9, solving the local air pollution concentration calculation model to obtain the local air pollution concentration of all cities;

[0016] Step 10: Display the local air pollution concentration in a digital way.

[0017] According to a specific implementation of an embodiment of the present invention, the monitoring concentration data includes daily monitoring concentrations of PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone;

[0018] The meteorological data include maximum temperature, minimum temperature, average temperature, precipitation, sunshine, surface temperature, air pressure and humidity.

[0019] According to a specific implementation of the embodiment of the present invention, step 2 specifically includes:

[0020] The monitoring concentration data at the daily level is used as the target variable, and the time variable and meteorological variable are used as the characteristic variables, among which the meteorological variable is the meteorological data, and the time variable includes the specific time characteristics of a specific day;

[0021] Based on the set target variables and feature variables, the data is divided into training set, validation set and test set according to the preset ratio, and the random forest model is pre-trained. The model with the best comprehensive effect of training, validation and testing is taken as the optimal model, and the model parameters are saved.

[0022] According to a specific implementation of the embodiment of the present invention, step 5 specifically includes:

[0023] Taking each city as the central city, the geographical distance between it and other surrounding cities is calculated. At the same time, the terrain factors between cities are taken into consideration, and the central city, surrounding cities, geographical distance and terrain factors are regarded as fixed city pairs.

[0024] According to a specific implementation of an embodiment of the present invention, the expression of the pollution diffusion function estimation model between cities is:

[0025]

[0026]

[0027] in, represents the monitoring concentration of air pollutants in the central city f on day t of m month of year y, It represents the monitoring concentration of air pollutants in the central city n of the surrounding cities on day t of month m of year y. represents the set of all cities, d is the distance band, represents a positive integer, is a dummy variable. If the distance between nearby city n and central city f is in distance band d, The value is 1. represents the intercept of each distance band d, represents the cross-border transmission coefficient of atmospheric pollutants, It represents the angle between the straight line direction from the central city to the surrounding cities and the wind direction of the central city on day t of year m month y. The value range is 0-180 degrees. represents the city fixed effect, are year-month-day fixed effects, represents standard error.

[0028] According to a specific implementation of the embodiment of the present invention, step 7 specifically includes:

[0029] The pollution diffusion function estimation model is estimated using air pollutant data that eliminates the influence of meteorological factors to obtain the cross-border transmission coefficient of atmospheric pollutants. , different distance bands d have different corresponding values, and finally a piecewise linear function that changes with the distance band is obtained as the existing pollution diffusion function.

[0030] According to a specific implementation of an embodiment of the present invention, the expression of the local air pollution concentration calculation model is:

[0031]

[0032]

[0033] in, is the air pollution concentration at the central city monitoring station on day t of year y, is the coefficient corresponding to the existing air pollution diffusion function, It is the angle between the wind direction of the surrounding cities and the straight line direction from the surrounding cities to the central city. and They represent the air pollution concentrations generated locally in the central city and surrounding cities respectively.

[0034] According to a specific implementation of the embodiment of the present invention, step 9 specifically includes:

[0035] The air pollutant data without the influence of meteorological factors are input into the local air pollution concentration calculation model, and the concentration of air pollutants generated locally in all cities is obtained through model solution.

[0036] The edge computing-based exogenous and local air pollution separation scheme in the embodiment of the present invention includes: step 1, obtaining air pollutant monitoring concentration data and meteorological data of different cities; step 2, training a random forest model using monitoring data and meteorological data; step 3, performing multiple random sampling of meteorological characteristics of meteorological data under the condition that the time dimension remains unchanged; step 4, inputting the sampled meteorological characteristics into the trained random forest model, obtaining air pollutant data that eliminates the influence of meteorological factors and sending it to the data middle station; step 5, matching central cities and surrounding cities according to the geographical distance and terrain factors between cities, forming different city pairs between all cities; step 6, establishing a pollution diffusion function estimation model between cities based on the city pairs and the wind direction data and wind speed data of each city; step 7, using air pollutant data to estimate the pollution diffusion function estimation model to obtain an existing pollution diffusion function; step 8, constructing a local air pollution concentration measurement model based on the existing pollution diffusion function; step 9, solving the local air pollution concentration measurement model to obtain the local air pollution concentration of all cities; step 10, displaying the local air pollution concentration in a digital way.

[0037] The beneficial effects of the embodiments of the present invention are as follows: through the scheme of the present invention, based on high-frequency atmospheric pollutant monitoring concentration data, by constructing a local meteorological factor impact elimination mechanism based on random forests and a pollution diffusion function model and a local atmospheric pollution concentration measurement model, it is possible to accurately separate the pollution from foreign cities and measure the concentration of atmospheric pollutants generated locally in the city. At the same time, the implementation of this technology based on the edge computing framework further reduces the requirements of this technology for hardware equipment and the use environment, making the atmospheric pollution concentrations of different cities at different stages comparable, which is conducive to timely measuring the effects of atmospheric environmental protection measures, accurately evaluating the implementation effects of air pollution prevention and control policies, and providing scientific support for regional atmospheric coordinated governance and precise decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only 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.

[0039] Figure 1 A flow chart of a method for separating exogenous and local air pollution based on edge computing provided in an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of the overall architecture of a method for separating exogenous and local air pollution based on edge computing provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0042] The following describes the embodiments of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0043] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present invention, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0044] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention. The illustrations only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the form, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0045] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0046] An embodiment of the present invention provides a method for separating exogenous and local air pollution based on edge computing, which can be applied to the process of air pollution prevention and control in environmental protection scenarios.

[0047] See also Figure 1 , is a flow chart of a method for separating exogenous and local air pollution based on edge computing provided by an embodiment of the present invention. Figure 1 and Figure 2 As shown, the method mainly comprises the following steps:

[0048] Step 1, obtaining atmospheric pollutant monitoring concentration data and meteorological data of different cities;

[0049] Optionally, the monitoring concentration data includes daily monitoring concentrations of PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone;

[0050] The meteorological data include maximum temperature, minimum temperature, average temperature, precipitation, sunshine, surface temperature, air pressure and humidity.

[0051] For example, you can obtain atmospheric pollutant monitoring concentration data and meteorological data for different cities. Atmospheric pollution monitoring concentration data include PM 2.5 、PM 10 The daily monitoring concentrations of SO2, NO2, CO, and O3 are monitored in real time by each pollutant monitoring station. The monitoring station collects air quality data every hour. Meteorological data includes meteorological elements such as maximum temperature, minimum temperature, average temperature, precipitation, and humidity, which are monitored in real time by meteorological stations in each city. Climate stations are also updated regularly. After obtaining pollution and meteorological data, the raw data of each hour is calculated through inverse distance weighting and arithmetic averaging to obtain the daily pollution monitoring concentration data of the city and stored.

[0052] Step 2: Use monitoring data and meteorological data to train the random forest model;

[0053] Based on the above embodiment, step 2 specifically includes:

[0054] The monitoring concentration data at the daily level is used as the target variable, and the time variable and meteorological variable are used as the characteristic variables, among which the meteorological variable is the meteorological data, and the time variable includes the specific time characteristics of a specific day;

[0055] Based on the set target variables and feature variables, the data is divided into training set, validation set and test set according to the preset ratio, and the random forest model is pre-trained. The model with the best comprehensive effect of training, validation and testing is taken as the optimal model, and the model parameters are saved.

[0056] In specific implementation, the random forest model is an integrated model composed of hundreds of individual decision tree models. The main feature of random forest is "randomness", which is reflected in two aspects: first, the sub-dataset is obtained by random sampling of the original data set, and second, the features and split features of each decision tree in the random forest are also generated by random selection. In the random forest prediction model, each tree can be used for prediction, and the final prediction result is the average value of the predictions of each tree. In the random forest model constructed by this method, the target variable is the daily level of atmospheric pollutant monitoring concentration, including PM2.5 concentration, PM10 concentration, sulfur dioxide concentration, nitrogen dioxide concentration, carbon monoxide concentration and ozone concentration. Feature variables are divided into time variables and meteorological variables. Meteorological variables are a series of meteorological elements, including maximum temperature, minimum temperature, average temperature, precipitation, sunshine, surface temperature, air pressure and humidity. Time variables are used to measure daily emission levels, which are indicated by the specific time characteristics of a specific day, such as the timestamp information of the day, the day of the week, and the week of the year. In the method of the present invention, the time variables include day_unix, day_julian, weekday and week.

[0057] Based on the set target variables and feature variables, the random forest model is pre-trained in combination with the collected models. 70% of the data is used as the training set, 10% of the data is used as the validation set, and 20% of the data is used as the test set. The model with the best comprehensive effect among the three situations of training, validation, and testing is selected as the optimal model, and the model parameters are saved.

[0058] Taking place A as an example, after collecting historical data within the scope of place A, the random forest model is used to fit the pollutant monitoring concentration in place A. The optimal solution is the initial random forest model corresponding to place A. Other cities can also obtain their corresponding random forest models based on this logic.

[0059] Step 3, under the condition that the time dimension remains unchanged, the meteorological characteristics of the meteorological data are randomly sampled multiple times;

[0060] In the specific implementation, a new feature variable is generated for each piece of data by resampling the meteorological variables. The new feature variable contains the original time variable and the resampled meteorological variable. Specifically, the resampled meteorological variables in the new feature variable are randomly extracted from the meteorological variable data of the two weeks before and after a specific day. Based on the historical meteorological variable data obtained, random sampling is performed. Assuming that the new meteorological variable that needs to be resampled on March 15, 2016 in place A can be taken out by random sampling from the meteorological observation data from March 1 to 29 of all years in the historical meteorological variable data, replacing the original meteorological data on March 15, 2016, to form a new input variable. Repeat the sampling 1000 times for each day's data to obtain a new data set. At this time, the daily data in place A corresponds to 1000 randomly extracted new feature vectors;

[0061] Step 4: Input the sampled meteorological features into the trained random forest model to obtain the air pollutant data without the influence of meteorological factors and send it to the data center;

[0062] In the specific implementation, based on steps 2 and 3, the random forest model of the pollutants in place A itself changing with the weather and the new feature vector after data standardization have been obtained. Based on the trained random forest model and the input of the new feature vector, the results after eliminating the impact of meteorology can be obtained. Since the sampling is repeated 1000 times, there are also 1000 results. Finally, the average value is calculated by combining the 1000 results to obtain the pollution data after eliminating the impact of meteorology and send it to the data center.

[0063] Step 5: Match the central city and the surrounding cities according to the geographical distance and terrain factors between the cities to form different city pairs between all the cities;

[0064] Based on the above embodiment, step 5 specifically includes:

[0065] Taking each city as the central city, the geographical distance between it and other surrounding cities is calculated. At the same time, the terrain factors between cities are taken into consideration, and the central city, surrounding cities, geographical distance and terrain factors are regarded as fixed city pairs.

[0066] In specific implementation, the central city and the surrounding cities are matched according to the geographical distance between the cities to form different city pairs between all cities. First, the longitude and latitude information of all cities is obtained from the map software; secondly, the geographical distance between all cities is calculated using the distance function according to the longitude and latitude information; thirdly, according to the terrain information of the map software, it is determined whether there are large mountains and other terrains that are not conducive to the spread of pollution; finally, each city is taken as the central city in turn to match the city pairs formed by the central city-surrounding city-distance between cities-topographic factors. For example, A-B-40km-hills. If the distance between the two cities is small, but there is a large mountain range in between, such as: there is a large mountain range between C and D, it will affect the spread of pollutants, so C and D cannot form a city pair.

[0067] Step 6, establishing a pollution diffusion function estimation model between cities based on the wind direction data and wind speed data of the city pairs and each city;

[0068] Furthermore, the expression of the pollution diffusion function estimation model between cities is:

[0069]

[0070]

[0071] in, represents the monitoring concentration of air pollutants in the central city f on day t of m month of year y, It represents the monitoring concentration of air pollutants in the central city n of the surrounding cities on day t of month m of year y. represents the set of all cities, d is the distance band, Represents a positive integer. is a dummy variable. If the distance between nearby city n and central city f is in distance band d, The value is 1. represents the intercept of each distance band d, represents the cross-border transmission coefficient of atmospheric pollutants, It represents the angle between the straight line direction from the central city to the surrounding cities and the wind direction of the central city on day t of year m month y. The value range is 0-180 degrees. represents the city fixed effect, are year-month-day fixed effects, represents standard error.

[0072] During the specific implementation, a pollution diffusion function estimation model between cities is established. Specifically, a concentric circle model is used to estimate the air pollution diffusion function. The concentric circle model describes the spillover effect between a location and several concentric circles extending outward from the location, allowing each concentric ring to have a different intercept and slope. The present invention uses d=1, 2, 3, ..., D as an index and defines a ring every 50 kilometers. Therefore, all city pairs are divided into distances from 0 to 50, 50 to 100..., (D-1)*50 to D*50 kilometers. In the model, the key influencing conditions are wind direction, wind speed, and the geographical distance from the central city to the surrounding cities. When the maximum transmission distance from the central city to the surrounding cities calculated by wind speed on a certain day is greater than the geographical distance between them, the data will be included in the pollution diffusion function estimation model. Otherwise, these data will not be included in the model. For example, if the wind speed in the central city is 2 meters per second on a certain day, the maximum transport distance on that day is 172.8 (2*60*60*24) kilometers. If the distance between the surrounding cities and the central city is less than 172.8 kilometers, they will be included in the model, otherwise they will not be included in the model. The detailed pollution diffusion function estimation model is as follows:

[0073]

[0074]

[0075] is the monitored concentration of air pollutants in central city f on day t of month m of year y, It is the monitored concentration of air pollutants in the central city n of the surrounding city on day t of month m of year y. is the set of all cities, d is the distance band, Represents a positive integer. is a dummy variable. If the distance between nearby city n and central city f is in distance band d, The value is 1. is the intercept of each distance band d, It is the cross-border transmission coefficient of atmospheric pollutants, also known as the diffusion coefficient, which reflects the average physical transmission of air pollution concentration from the central city to nearby cities in each band. It is the angle between the straight line direction from the central city to the surrounding cities and the wind direction of the central city on day t of month y of year y. The value range is 0-180 degrees. When it is greater than 90 degrees, the wind direction in the central city is opposite to the straight line direction from the central city to the surrounding cities, which can define that air pollution will not spread from the central city to the surrounding cities. is the city fixed effect, is the year-month-day fixed effect. Considering the serial correlation at different times in nearby cities, Standard errors are clustered at the level of nearby cities.

[0076] Step 7, using air pollutant data to estimate the pollution diffusion function estimation model to obtain the existing pollution diffusion function;

[0077] Furthermore, the step 7 specifically includes:

[0078] The pollution diffusion function estimation model is estimated using air pollutant data that eliminates meteorological influences, and the cross-border transmission coefficient of atmospheric pollutants is obtained. , different distance bands d have different corresponding values, and finally a piecewise linear function that changes with the distance band is obtained as the existing pollution diffusion function.

[0079] In specific implementation, the pollution diffusion function is estimated using the pollution data that eliminates the meteorological influence, that is, the accurate expression of the existing pollution diffusion function is obtained. The coefficients obtained by estimating the pollution diffusion function using the pollution data that eliminates the meteorological influence are the coefficients estimated in step 6. , different distance bands d have different corresponding values, and the final pollution diffusion function is a piecewise linear function that changes with the distance band. For example, the coefficient of the air pollution diffusion function within a 50-kilometer range is 0.256, which means that for every 1 microgram / cubic meter increase in SO2 in the central city, the SO2 in the surrounding cities within a 50-kilometer range will increase by about 0.256 micrograms / cubic meter. And the farthest transmission distance can be found from the attenuation function. When the geographical distance between cities reaches a certain level, the coefficient is positive but not statistically significant, indicating that the farthest transmission distance of pollutants is less than this distance.

[0080] Step 8: construct a local air pollution concentration calculation model based on the existing pollution diffusion function;

[0081] Furthermore, the expression of the local air pollution concentration calculation model is:

[0082]

[0083]

[0084] in, is the air pollution concentration at the central city monitoring station on day t of year y, is the coefficient corresponding to the existing air pollution diffusion function, It is the angle between the wind direction of the surrounding cities and the straight line direction from the surrounding cities to the central city. and They represent the air pollution concentrations generated locally in the central city and surrounding cities respectively.

[0085] In the specific implementation, a local atmospheric pollution concentration measurement model is constructed. Through multiple linear equation modeling based on the pollution diffusion function, the pollution concentration generated locally in each city on day t of year m month y is obtained. In the model, each city is regarded as a point, that is, an air pollution source. The air pollution generated by the central city can be transported to the surrounding cities by wind, and the air pollution generated by the surrounding cities can also be transported to the central city. The amount of pollution transmission between cities depends on the wind direction, wind speed and distance. Therefore, the monitored pollution concentration of a central city is equal to the pollution concentration generated locally plus the pollution transported in from nearby cities, minus the pollution blowing to nearby cities.

[0086] The specific local air pollution concentration calculation model is as follows:

[0087]

[0088]

[0089] is the air pollution concentration at the central city monitoring station on day t of year y, is the coefficient corresponding to the air pollution diffusion function, It is the angle between the wind direction of the surrounding cities and the straight line direction from the surrounding cities to the central city. and Represents the concentration of air pollution generated locally in the central city and surrounding cities respectively. The meanings of other variables are the same as those in the pollution diffusion function estimation model.

[0090] It is worth noting that when the maximum transportation distance on a certain day is less than the geographical distance between the central city and the surrounding cities, these data are not considered in the calculation.

[0091] Step 9, solving the local air pollution concentration calculation model to obtain the local air pollution concentration of all cities;

[0092] Based on the above embodiment, step 9 specifically includes:

[0093] The air pollutant data without the influence of meteorological factors are input into the local air pollution concentration calculation model, and the concentration of air pollutants generated locally in all cities is obtained through model solution.

[0094] In specific implementation, the air pollutant data that eliminates the influence of meteorological factors can be input into the local atmospheric pollution concentration calculation model, and the local atmospheric pollution concentration calculation model can be solved to obtain the local atmospheric pollution concentration of all cities. The pollution data that eliminates the influence of meteorological factors can be input into the local atmospheric pollution concentration calculation model, and the atmospheric pollutant concentration generated locally in all cities can be obtained through the model solution.

[0095] Step 10: Display the local air pollution concentration in a digital way.

[0096] During specific implementation, the air pollution data of various cities can be summarized and organized, and a policy evaluation module can be constructed that includes measurement methods such as dual machine learning, synthetic control method, double difference method, propensity score matching, etc. By selecting different measurement methods to analyze the data, the final analysis results can be obtained, and the effect of policy implementation can be evaluated in combination with specific policies. At the same time, an air pollution map display system can be constructed based on the transmitted data to realize dynamic monitoring of air pollution conditions in various places.

[0097] The method for separating exogenous and local air pollution based on edge computing provided in this embodiment can accurately separate pollution from foreign cities and calculate the concentration of air pollutants generated locally in the city by building a mechanism for eliminating the impact of local meteorological factors and a pollution diffusion function model based on random forests and a local air pollution concentration measurement model based on high-frequency air pollutant monitoring concentration data. At the same time, the implementation of this technology based on the edge computing framework further reduces the requirements of this technology for hardware equipment and the use environment, making the air pollution concentrations of different cities at different stages comparable, which is conducive to timely measuring the effects of atmospheric environmental protection measures and accurately evaluating the implementation effects of air pollution prevention and control policies. It provides a new idea for solving the problem that the concentration of existing air pollution monitoring stations cannot represent the local pollution emission level, and improves the recognition accuracy, adaptability and real-time performance.

[0098] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof.

[0099] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for separating exogenous and local atmospheric pollution based on edge computing, characterized in that: include: Step 1, obtaining atmospheric pollutant monitoring concentration data and meteorological data of different cities, wherein the monitoring concentration data includes daily monitoring concentrations of PM2.5, PM10, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone; The meteorological data include maximum temperature, minimum temperature, average temperature, precipitation, sunshine, surface temperature, air pressure and humidity; Step 2: Use monitoring data and meteorological data to train the random forest model; The step 2 specifically includes: The monitoring concentration data at the daily level is used as the target variable, and the time variable and meteorological variable are used as the characteristic variables, among which the meteorological variable is the meteorological data, and the time variable includes the specific time characteristics of a specific day; Based on the set target variables and feature variables, the data is divided into training set, validation set and test set according to the preset ratio, and the random forest model is pre-trained. The model with the best comprehensive effect in the three situations of training, validation and testing is taken as the optimal model, and the model parameters are saved; Step 3, under the condition that the time dimension remains unchanged, the meteorological characteristics of the meteorological data are randomly sampled multiple times; Step 4: Input the sampled meteorological features into the trained random forest model to obtain the air pollutant data without the influence of meteorological factors and send it to the data center; Step 5: Match the central city and the surrounding cities according to the geographical distance and terrain factors between the cities to form different city pairs between all the cities; The step 5 specifically includes: Take each city as the central city, calculate the geographical distance between it and other surrounding cities, and consider the topographic factors between cities, and regard the central city, surrounding cities, geographical distance and topographic factors as fixed city pairs; Step 6: Establish a pollution diffusion function estimation model between cities based on the city pairs and the wind direction data and wind speed data of each city, wherein the expression of the pollution diffusion function estimation model between cities is: ； in, represents the monitoring concentration of air pollutants in the central city f on day t of m month of year y, It represents the monitoring concentration of air pollutants in the central city n of the surrounding cities on day t of month m of year y. represents the set of all cities, d is the distance band, represents a positive integer, is a dummy variable. If the distance between nearby city n and central city f is in distance band d, The value is 1. represents the intercept of each distance band d, represents the cross-border transmission coefficient of atmospheric pollutants, It represents the angle between the straight line direction from the central city to the surrounding cities and the wind direction of the central city on day t of year m month y. The value range is 0-180 degrees. represents the city fixed effect, are year-month-day fixed effects, represents standard error; Step 7, using air pollutant data to estimate the pollution diffusion function estimation model to obtain the existing pollution diffusion function; The step 7 specifically includes: The pollution diffusion function estimation model is estimated using air pollutant data that eliminates the influence of meteorological factors to obtain the cross-border transmission coefficient of atmospheric pollutants. , different distance bands d have different corresponding values, and finally a piecewise linear function that changes with the distance band is obtained as the existing pollution diffusion function; Step 8: construct a local air pollution concentration calculation model based on the existing pollution diffusion function, wherein the expression of the local air pollution concentration calculation model is: ； in, is the air pollution concentration at the central city monitoring station on day t of year y, is the coefficient corresponding to the existing air pollution diffusion function, It is the angle between the wind direction of the surrounding cities and the straight line direction from the surrounding cities to the central city. and They represent the concentration of air pollution generated locally in the central city and surrounding cities respectively; Step 9, solving the local air pollution concentration calculation model to obtain the local air pollution concentration of all cities; The step 9 specifically includes: The air pollutant data without the influence of meteorological factors are input into the local air pollution concentration calculation model, and the concentration of air pollutants generated locally in all cities is obtained through model solution; Step 10: Display the local air pollution concentration in a digital way.

Images