Dynamic inversion method of pollution source intensity based on air quality machine learning rolling prediction model

Abstract

The present invention relates to a method for strongly dynamic inversion of pollution sources based on an air quality machine learning rolling prediction model, comprising the following steps: step S1, acquisition and processing of an air quality modeling data set; step S2, construction of an air quality machine learning prediction model; step S3, meteorological standardization assessment of air quality; step S4, rolling prediction of air quality and meteorological standardization iterative calculation. The present invention not only solves the problem that existing meteorological normalization technology is limited by insufficient timeliness and is mainly used for post-analysis of policy interventions, but also reduces the biased estimation caused by random meteorological normalization, thereby improving the accuracy of the results. The present invention is particularly suitable for environmental management application scenarios with high timeliness requirements, such as emergency assessment of heavy pollution weather, and has important application value.

Description

Technical Field

[0001] The present invention relates to air pollution prevention and control technology, and in particular to a pollution source strong dynamic inversion method based on an air quality machine learning rolling prediction model. Background Art

[0002] Traditional methods for assessing pollution source emission intensity analyze ambient air quality monitoring data. By using appropriate mathematical and statistical methods to adjust for biases caused by variations in meteorological covariates in pollutant concentration time series, the contribution of pollution source intensity to air quality changes is revealed. However, applying these traditional mathematical and statistical methods in practice presents technical difficulties: 1) Classical mathematical statistics imposes stringent parameter testing requirements, while atmospheric environmental monitoring data often do not fully conform to the Gaussian normal distribution and homogeneity of variance requirements; 2) Meteorological factors and air pollution interact in complex nonlinear ways, and classical statistical models are insufficiently robust for these complex nonlinear problems.

[0003] With the rapid development of artificial intelligence (AI) machine learning algorithms, a series of new algorithms suitable for modeling complex problems have been continuously proposed. David K. Carslaw and his colleagues at York University in the UK developed a meteorological normalization technique based on random forest modeling to separate meteorological and emission information from air quality fluctuations. This technique assumes that air pollution emissions have certain cyclical characteristics. Using machine learning algorithms such as random forests, which have good fitting performance, a regression model is trained based on historical data, including meteorological parameters and variables representing the cyclical nature of pollution source intensity, to correlate these with monitored atmospheric pollutant concentrations. To reduce the impact of meteorological factors on pollutant concentrations at different times, at any given moment, several sets of meteorological data are randomly selected from the historical data. The trained model is used to predict the ambient concentration of the pollutant under these historical meteorological conditions. The concentration of the pollutant under these historical meteorological conditions is then calculated for each moment during the study period. The time series variation in the arithmetic mean of these multiple predicted concentrations reflects the change in emission intensity. At each moment, the pollutant is normalized under the average meteorological conditions. This technique overcomes the shortcomings and limitations of classical statistical evaluation methods. Currently, this technique has been widely used to study trends in air pollution emissions in hundreds of cities worldwide.

[0004] However, in addition to routine control measures, air pollution prevention and control also includes temporary and unpredictable measures, such as emergency response to heavy pollution and ensuring air quality for major events. These temporary measures significantly increase the timeliness of these measures. Current meteorological standardization technologies based on machine learning modeling are primarily used for retrospective analysis of scenarios such as policy interventions. These analyses are time-consuming, difficult to automate, and result in delayed results, making them ineffective for time-sensitive decision-making and management. Summary of the Invention

[0005] In order to meet the above-mentioned high-timeliness management needs, the present invention proposes a dynamic inversion method for pollution source strength based on an air quality machine learning rolling prediction model. By designing a machine learning pollutant rolling prediction model and designing an algorithm to effectively control the disturbance of changes in meteorological conditions, it is possible to quickly invert the changes in local pollutant emission source strength from real-time air quality monitoring data, so as to provide timely and efficient automated feedback evaluation for environmental management scenarios such as heavy pollution emergency control effects.

[0006] To achieve the above objectives, the present invention provides a method for dynamic inversion of pollution sources based on an air quality machine learning rolling prediction model, the method comprising the following steps:

[0007] Step S1, acquisition and processing of air quality modeling datasets: based on hourly continuous monitoring data of six pollutant concentrations of SO2, NO2, CO, O3, PM10, and PM2.5 from air quality monitoring stations within the study area, as well as synchronized environmental meteorological parameters and time trend variables, the time trend variables being used to characterize pollution emissions or atmospheric physical and chemical processes with periodic changes;

[0008] Step S2, construction of air quality machine learning prediction model: for any air pollutant, at the studied time t, use machine learning regression algorithm to compare the pollutant concentration C with the ambient meteorological parameters and time trend variables Conduct modeling, compare the consistency and difference between the pollutant concentrations predicted by the model and the actual monitored concentrations, calculate the correlation coefficient and root mean square error to evaluate the model fitting effect;

[0009] in is the model residual term, and the pollutant prediction model trained at the current time t is :

[0010]

[0011] ,

[0012] Where i represents the meteorological parameters of the model, including temperature T, relative humidity RH, boundary layer height BLH, etc.; j represents the time trend variable, including the daily hourly time series hour, the day of the Gregorian calendar date, and the linear trend variable wait;

[0013] Step S3, meteorological standardization assessment of air quality: Use the trained model to predict the pollutants at time t under a fixed set of historical meteorological conditions, replacing the average value as the representation of emission intensity, that is, the meteorological standardization concentration of pollutants at time t for:

[0014]

[0015] Where k is the kth sample of the selected fixed N groups of historical meteorological data, is the pollutant concentration predicted by the model under the kth historical meteorological data of the selected fixed N groups;

[0016] Step S4: air quality rolling forecast and meteorological standardization iterative calculation:

[0017] For the pollutant monitoring concentration at the next moment t+1, after the monitoring data collection is completed, the pollutant, meteorological parameters and time variable data at time t+1 are included as a new sample in the original training data set of the same length, and steps S2 and S3 are repeated using the new data set. The new training model at time t+1 is used. Calculate the meteorological normalized concentration of pollutants at time t and t+1 respectively, and keep N sets of historical meteorological conditions unchanged; as new monitoring data continues to synchronize with time, repeat the steps of t+1 for time t+2, and incorporate the new data at time t+2 to retrain a pollutant prediction model , using the model Update the meteorological standardized concentrations of pollutants at time t, t+1, and t+2, and the newly trained model Used to predict the concentration of various pollutants under fixed N groups of weather conditions, and discard the previous round of utilization Repeat the above steps until rolling iterative calculation is performed at every moment.

[0018] Preferably, the environmental meteorological parameters include ground temperature T, relative humidity RH, wind speed WS, wind direction WD, air pressure SP, radiation intensity SSR, mixing layer height BLH, total cloud cover TCC, precipitation Prep, trajectory category Air cluster and trajectory length Air length .

[0019] Preferably, the time trend variables include a timestamp Unix time, a lunar day number LunarDay, a solar day number Day-of-Year, a day of the week Day-of-Week, and a daily hour sequence Hour.

[0020] Preferably, the regression algorithm is a random forest algorithm, a neural network algorithm, a gradient regression tree algorithm or an extreme gradient boosting tree algorithm.

[0021] Preferably, the environmental meteorological parameters It includes a variety of easily accessible meteorological observation data or meteorological reanalysis data, and the variables used to characterize the emission source intensity are any air-related data related to pollution emission activities.

[0022] Preferably, in step S4, the fixed meteorological conditions selected for the fixed meteorological standardization are N groups of meteorological conditions fixed within any period of time.

[0023] Based on the above technical solution, the advantages of the present invention are:

[0024] The strong dynamic inversion method of pollution sources based on the air quality machine learning rolling prediction model of the present invention realizes the rolling prediction of pollutants through hourly iterative modeling, and uses each round of newly trained pollutant prediction model to normalize the pollutants in the study period (such as heavy pollution process) to fixed preset meteorological conditions, realizing dynamic inversion while ensuring that the meteorological normalized pollutant concentrations are comparable in time series.

[0025] The pollution source intensity dynamic inversion method of the present invention not only solves the problem that the existing meteorological normalization technology is limited by insufficient timeliness and is mainly used for post-analysis of policy interventions, but also reduces the biased estimation caused by random meteorological normalization, thereby improving the accuracy of the results. It is particularly suitable for environmental management application scenarios with high timeliness requirements, such as emergency assessment of heavy pollution weather, and has important application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0027] Figure 1 This is a flow chart of the pollution source strong dynamic inversion method of the present invention;

[0028] Figure 2 The root mean square error and correlation coefficient of the SO2 concentration output model fitting data in Example 1;

[0029] Figure 3 This is the SO2 hourly concentration data after the infinite loop modeling and prediction system is established in Example 1. DETAILED DESCRIPTION

[0030] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments.

[0031] The present invention provides a method for the strong dynamic inversion of pollution sources based on an air quality machine learning rolling prediction model, such as Figure 1 As shown, the pollution source strong dynamic inversion method includes the following steps:

[0032] Step S1, acquisition and processing of air quality modeling data sets: based on the hourly continuous monitoring data of six pollutant concentrations of SO2, NO2, CO, O3, PM10 and PM2.5 at air quality monitoring stations in the study area, as well as synchronized environmental meteorological parameters and time trend variables, the time trend variables are used to characterize pollution emissions or atmospheric physical and chemical processes with periodic changes.

[0033] Specifically, based on the hourly continuous monitoring of six pollutants (SO2, NO2, CO, O3, PM10, PM2.5) concentration data at the air quality monitoring stations in the study area, as well as synchronized environmental meteorological parameters and time trend variables, the environmental meteorological parameters include ground temperature T, relative humidity RH, wind speed WS, wind direction WD, air pressure SP, radiation intensity SSR, mixing layer height BLH, total cloud cover TCC, precipitation Prep, trajectory category Air cluster and trajectory length Air length Time trend variables are used to characterize pollutant emissions or atmospheric physical and chemical processes with periodic changes. For example, the day of the year, the day of the week, and the hourly time series are used to indicate pollution source emission activities with seasonal, weekly, and daily variations on an annual time scale. Long-term interdecadal pollution emission trends are indicated by linearly increasing hourly variables, such as the Unix time (i.e., the accumulated seconds since midnight, January 1, 1970, Greenwich Mean Time). The dataset used to train pollutant models is based on at least one month of continuous monitoring data, with the principle that the longer the coverage period, the better.

[0034] Step S2, construction of air quality machine learning prediction model: for the six pollutants, at the studied time t, the concentration of pollutants is analyzed one by one using regression algorithm. and environmental meteorological parameters and time trend variables Modeling is performed to compare the consistency and difference between the pollutant concentrations predicted by the model and the actual monitored concentrations, and the correlation coefficient and root mean square error are calculated to evaluate the model fitting effect; preferably, the regression algorithm is a random forest algorithm, a neural network algorithm, a gradient regression tree algorithm or an extreme gradient boosting tree algorithm.

[0035] in is the model residual term, and the pollutant prediction model trained at the current time t is :

[0036]

[0037] ,

[0038] Where i represents the meteorological parameters of the model, including temperature T, relative humidity RH, boundary layer height BLH, etc.; j represents the time trend variable, including the daily hourly time series hour, the day of the Gregorian calendar date, and the linear trend variable wait.

[0039] Taking a full calendar year of monitoring data as an example, the pollutant prediction model is trained on the entire hourly dataset. Considering the timeliness requirements of practical applications, when the training sample is sufficiently large (e.g., more than a full year), the model's generalization ability on the test set is not required. However, the root mean square error and correlation coefficient of the model's fit to the entire data set must be reported.

[0040] The environmental meteorological parameters used in the modeling may include a variety of easily accessible meteorological observation data or meteorological reanalysis data, such as ground temperature, relative humidity, wind speed, wind direction, air pressure, radiation intensity, mixing layer height, total cloud cover, precipitation, trajectory category and trajectory length, etc. The variables used to characterize the emission source intensity can be any air-related data related to pollution emission activities, including but not limited to motor vehicle traffic, key source pollutant emission monitoring data, etc.

[0041] Step S3, meteorologically standardized air quality assessment: The trained model is used to predict the pollutant concentration at time t under a fixed set of historical meteorological conditions, replacing the numerical mean as the representation of emission intensity. That is, the meteorologically standardized concentration of the pollutant at time t is:

[0042]

[0043] Where k is the kth sample of the selected fixed N groups of historical meteorological data, It is the pollutant concentration predicted by the model under the kth historical meteorological data of the selected fixed N groups.

[0044] Specifically, the purpose of meteorological standardization is to normalize the interference of meteorological conditions that change over time on pollutant concentrations to average meteorological conditions, thereby eliminating the disturbance of meteorological conditions that reflects changes in pollution emission intensity. The existing technology uses a trained random forest pollutant prediction model f to predict the concentration of pollutants at time t under N randomly selected groups of historical meteorological conditions, and takes the algebraic mean of the predicted pollutant concentrations. . The N randomly selected groups of meteorological data are extracted from the historical environmental meteorological data in the original training data set. According to the law of large numbers and the central limit theorem, when N is large enough, the meteorological disturbance in the pollutant concentration at that moment will tend to 0. In theory, the change in the meteorological standardized concentration of pollutants over time can reflect the change in the intensity of pollution emissions. The more common number of random meteorological samples is 300 to 1000 groups. The larger the meteorological samples extracted, the longer the calculation time. At present, this meteorological standardization method still has shortcomings, that is, the randomly extracted historical meteorological data introduces random errors, resulting in the risk that the randomly extracted "average meteorological conditions" at different times are not completely consistent; when the extracted meteorological samples are large enough to eliminate random risks, the cost of computing resources is greatly increased.

[0045] To meet the timeliness requirements and eliminate meteorological disturbances in time series with as small a historical meteorological sample as possible, the present invention proposes using the trained pollutant model to predict the ambient concentration of pollutants at time t under a fixed set of historical meteorological conditions (non-random), such as 24 sets of hourly meteorological conditions fixed on a certain historical day, and taking the algebraic mean as a representation of emission intensity.

[0046] Step S4: air quality rolling forecast and meteorological standardization iterative calculation:

[0047] For the pollutant monitoring concentration at the next moment t+1, after the monitoring data collection is completed, the pollutant, meteorological parameters and time variable data at time t+1 are included as a new sample in the original training data set of the same length (such as a complete natural year) (the first sample of the original data set is removed).

[0048] Repeat steps S2 and S3 using the new dataset, and use the new training model at time t+1 Calculate the meteorological standardized concentration of pollutants at time t and t+1 respectively, and keep the K group of historical meteorological conditions unchanged; then the model used to describe the relationship between pollutant concentration, emission and meteorological mapping at time t+1 is for:

[0049]

[0050] Using the newly trained model at time t+1 Calculate the meteorological standardized concentrations of pollutants at time t and time t+1 respectively, and the K groups of historical meteorological conditions can remain unchanged:

[0051]

[0052] As new monitoring data continues to synchronize over time, repeat the steps of t+1 for time t+2, and retrain a pollutant prediction model by incorporating the new data at time t+2. , using the model Update the meteorological standardized concentrations of pollutants at time t, t+1, and t+2, and the newly trained model It is used to predict the concentration of various pollutants under fixed N groups of meteorological conditions. Therefore, the meteorological standardization results of the new prediction model constructed in each round are comparable in time series. At this time, the meteorological standardization results of the previous round of prediction models are discarded. Repeat the above steps until rolling iterative calculation is performed at every moment.

[0053] Preferably, in the fixed meteorological standardization, the selected fixed meteorological conditions are N groups of meteorological conditions fixed within any period of time.

[0054] The strong dynamic inversion method of pollution sources based on the air quality machine learning rolling prediction model of the present invention realizes the rolling prediction of pollutants through hourly iterative modeling, and uses each round of newly trained pollutant prediction model to normalize the pollutants in the study period (such as heavy pollution process) to fixed preset meteorological conditions, realizing dynamic inversion while ensuring that the meteorological normalized pollutant concentrations are comparable in time series.

[0055] The pollution source intensity dynamic inversion method of the present invention not only solves the problem that the existing meteorological normalization technology is limited by insufficient timeliness and is mainly used for post-analysis of policy interventions, but also reduces the biased estimation caused by random meteorological normalization, thereby improving the accuracy of the results. It is particularly suitable for environmental management application scenarios with high timeliness requirements, such as emergency assessment of heavy pollution weather, and has important application value.

[0056] Example 1

[0057] The present invention can be implemented through program software operation. Taking the rolling prediction of PM2.5 concentration hourly monitoring data in Tianjin as an example to invert emission intensity, the specific implementation method includes the following steps:

[0058] Step 101 uses Python code to preprocess (for missing values, outliers, etc.) and load data based on hourly SO2 concentration observations from 2015 to the present in Tianjin, along with synchronized ambient meteorological parameters (surface temperature, relative humidity, wind speed, wind direction, air pressure, radiation intensity, mixing layer height, total cloud cover, precipitation, trajectory type, and trajectory length) and temporal trend variables (timestamp, lunar day number, solar day number, day of the week, and daily hourly time series). The selected ambient meteorological variables can include a variety of readily available meteorological observations or meteorological reanalysis data. Variables used to characterize emission source intensity can include air-related data related to pollution emission activities, including but not limited to motor vehicle traffic and key source pollutant emission monitoring data.

[0059] Step 102: For the hourly concentration of SO2, based on the Python code, the random forest algorithm is used to model the SO2 concentration, environmental meteorological parameters and time trend variables at the studied time t, and the root mean square error (RMSE) and correlation coefficient (R2) of the model fitting data are output, such as Figure 2 The modeling algorithm used is random forest, but other regression algorithms can also be neural networks, gradient regression trees, extreme gradient boosting trees (XGBoost), etc.

[0060] In step 103, all dates with a complete 24-hour day of data are screened out based on the Python code, 24 sets of complete meteorological data (hours) for a certain historical day are randomly selected, and 24 sets of ambient concentrations of SO2 under the fixed meteorological conditions at time t are predicted based on the random forest model. The average value is used to represent the emission intensity at that time.

[0061] Step 104: Establish an infinite loop modeling and prediction system, such as completing the processing of steps 101 to 103 at time t+1, and outputting the corresponding data and image files of the pollutant observation concentration and emission intensity in the last 7 days, and iterating every hour, such as Figure 3 As shown in the figure, the output at 19:00 on August 31, 2024 is the hourly SO2 concentration data from 19:00 on August 24, 2024 to 18:00 on August 31, 2024. The solid line is the monitored concentration (Observed SO2) and the dotted line is the predicted concentration after fixed meteorological standardization (Predicted SO2).

[0062] like Figure 3 As shown in the figure, the solid line represents the monitored ambient concentration, which is highly volatile due to meteorological influences. After the meteorological standardization of the present invention, the SO2 volatility of the dotted line is significantly reduced, and it shows a stable change every day, reflecting the stable daily cycle of pollution emission intensity. Taking the continuous increase in SO2 monitored from August 27 to August 29 as an example, it can be found that the increase in SO2 ambient concentration is not caused by increased emissions from pollution sources, but rather by the accumulation of SO2 pollution caused by unfavorable meteorological conditions. This verifies the feasibility of the method of the present invention.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present invention. They should all be included in the scope of the technical solution for protection of the present invention.

Claims

1. A pollution source dynamic inversion method based on an air quality machine learning rolling prediction model, characterized by: The pollution source strong dynamic inversion method comprises the following steps: Step S1, acquisition and processing of air quality modeling datasets: based on hourly continuous monitoring data of six pollutant concentrations of SO2, NO2, CO, O3, PM10, and PM2.5 from air quality monitoring stations within the study area, as well as synchronized environmental meteorological parameters and time trend variables, the time trend variables being used to characterize pollution emissions or atmospheric physical and chemical processes with periodic changes; Step S2, construction of air quality machine learning prediction model: for the six pollutants, at the studied time t, the concentration of pollutants is analyzed one by one using regression algorithm. and environmental meteorological parameters and time trend variables Conduct modeling, compare the consistency and difference between the pollutant concentrations predicted by the model and the actual monitored concentrations, calculate the correlation coefficient and root mean square error to evaluate the model fitting effect; in is the model residual term, and the pollutant prediction model trained at the current time t is : ； ； ； In the formula represents the modeled ambient meteorological parameters; represents the time trend variable; T represents the ground temperature, RH represents the relative humidity, WS represents the wind speed, WD represents the wind direction, SP represents the air pressure, SSR represents the radiation intensity, BLH represents the height of the mixing layer, TCC represents the total cloud cover, Prep represents the precipitation, Air cluster Indicates the trajectory category, Air length Indicates the trajectory length; Hour indicates the daily hour sequence, Day-of-Week indicates the day of the week, Day-of-Year indicates the day of the year in the Gregorian calendar, LunarDay indicates the day of the year in the lunar calendar, and Unix time indicates the timestamp; Step S3, meteorologically standardized air quality assessment: The trained model is used to predict the pollutant concentration at time t under a fixed set of historical meteorological conditions, replacing the numerical mean as the representation of emission intensity. That is, the meteorologically standardized concentration of the pollutant at time t is: ； Where k is the kth sample of the selected fixed N groups of historical meteorological data, is the pollutant concentration predicted by the model under the kth historical meteorological data of the selected fixed N groups; Step S4: air quality rolling forecast and meteorological standardization iterative calculation: For the pollutant monitoring concentration at the next moment t+1, after the monitoring data collection is completed, the pollutant, meteorological parameters and time variable data at time t+1 are included as a new sample into the original training data set of the same length, and steps S2 and S3 are repeated using the new data set. The new training model at time t+1 is used. Calculate the meteorological normalized concentrations of pollutants at time t and t+1 respectively, and keep the K groups of historical meteorological conditions unchanged; as the new monitoring data continues to synchronize with time, repeat the steps of t+1 for time t+2, and incorporate the new data at time t+2 to retrain a pollutant prediction model , using the model Update the meteorological standardized concentrations of pollutants at time t, t+1, and t+2, and the newly trained model Used to predict the concentration of various pollutants under fixed N groups of weather conditions, and discard the previous round of utilization Repeat the above steps until rolling iterative calculation is performed at every moment.

2. The pollution source strong dynamic inversion method according to claim 1, characterized in that: The regression algorithm is a random forest algorithm, a neural network algorithm, a gradient regression tree algorithm or an extreme gradient boosting tree algorithm.

3. The pollution source strong dynamic inversion method according to claim 1, characterized in that: The environmental meteorological parameters include a variety of easily accessible meteorological observation data or meteorological reanalysis data, and the variables used to characterize the emission source intensity are any air-related data related to pollution emission activities.

4. The pollution source strong dynamic inversion method according to claim 1, characterized in that: In step S4, the fixed meteorological conditions selected for the fixed meteorological standardization are N groups of meteorological conditions fixed within any period of time.

Images