Atmospheric pollution space-time dynamic emission reduction method based on meta-heuristic optimization algorithm and sensitivity algorithm

Abstract

The invention discloses an atmospheric pollution space-time dynamic emission reduction method based on a meta-heuristic optimization algorithm and a sensitivity algorithm. The method comprises the following steps: (1) constructing a multi-source heterogeneous data set and carrying out preprocessing; (2) carrying out mathematical statistical analysis on the data set, extracting meteorological conditions and atmospheric pollutant spatio-temporal change characteristic curves in different target areas, and generating a concentration field driving data file of the accompanying model; (3) predicting three-dimensional space-time distribution of a concentration field by using an atmospheric chemical transmission model, and evaluating and correcting by using the extracted concentration data per hour of the observation station; making a high-resolution atmospheric pollutant mass concentration three-dimensional space-time grid data set; (4) introducing a sensitivity algorithm into an atmospheric chemical transmission model, constructing and accompanying a meta-heuristic optimization algorithm model, and calculating an absolute contribution rate three-dimensional space-time grid data set of different emission source types; (5) inputting the absolute contribution rate three-dimensional space-time grid data sets of different emission source types into the adjoint-meta-heuristic optimization algorithm model to obtain predicted three-dimensional space-time distribution of optimal emission reduction of different emission types; the method provides a certain reference basis for improving the regional atmospheric environment quality and improving the cost effectiveness.

Description

Technical Field

[0001] The present invention relates to the technical field of energy conservation and environmental protection, and in particular to a method for spatiotemporal dynamic emission reduction of atmospheric pollution based on a meta-heuristic optimization algorithm and a sensitivity algorithm. Background Art

[0002] At present, in order to adapt to the new goals and new demands in the field of atmospheric environmental protection, various regions usually adopt the total amount control method of atmospheric environmental quality targets to achieve the purpose of improving environmental quality. From the perspective of the relationship between the total amount control of atmospheric pollutants and the improvement of environmental quality, pollutant emission reduction is the most important direct means. The total amount control policy is based on the premise of the total allowable emissions of the total pollution sources within the target area. The source-receptor response relationship of pollutants is constructed through the atmospheric chemical transport model, the allocation share of the total allowable emissions is formulated, and the total amount is scientifically and rationally allocated to each pollution source, ultimately achieving effective pollution emission limitation and reduction.

[0003] At present, it is still unclear whether the total amount control policy for pollutant emissions can achieve optimal performance. It is necessary to seek the optimal emission reduction strategy that meets the total amount control constraints on its basis. Most previous studies have used atmospheric chemical transport models to predict the mass concentration of atmospheric chemical pollutants and formulate atmospheric particulate matter and ozone prevention and control strategies based on the combination of regional emission species reduction ratios guided by the minimum control cost. However, they lack consideration of the spatial differences in the emission rates of different species and the impact of the spatiotemporal heterogeneity of chemical generation, which greatly reduces the effectiveness of atmospheric pollution prevention and control strategies within the system. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a spatiotemporal dynamic reduction method for atmospheric pollution based on a metaheuristic optimization algorithm and a sensitivity algorithm, to construct a source-receptor response model to quantify the nonlinear response relationship of atmospheric pollutant concentrations when emission sources change, to calculate the contribution of all emission sources at different times and locations, and to achieve the maximum economic benefit of regional atmospheric pollutant prevention and control strategies.

[0005] Technical solution: The method for spatiotemporal dynamic reduction of atmospheric pollution based on a meta-heuristic optimization algorithm and a sensitivity algorithm of the present invention comprises the following steps:

[0006] (1) Obtain local weather and meteorological observation data, ambient air quality monitoring data, and emission inventories to construct a multi-source heterogeneous data set and perform preprocessing;

[0007] (2) Perform mathematical and statistical analysis on the data set, extract the temporal and spatial variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify key control areas and key emission reduction periods in different seasons and months, and generate concentration field driving data files accompanying the model;

[0008] (3) Use the atmospheric chemical transport model to predict the three-dimensional spatiotemporal distribution of the concentration field, and use the extracted hourly concentration data of the observation station to evaluate and correct it; adjust the physical and chemical parameterization scheme of the optimization model until the prediction results meet the accuracy requirements, and finally produce a high-resolution three-dimensional spatiotemporal grid dataset of atmospheric pollutant mass concentrations;

[0009] (4) Based on the prepared concentration field dataset and model-driven data file, the sensitivity algorithm is introduced into the atmospheric chemical transport model, and a meta-heuristic optimization algorithm model is constructed to calculate the absolute contribution rate of different emission source types in a three-dimensional spatiotemporal grid dataset;

[0010] (5) The three-dimensional spatiotemporal grid dataset of the absolute contribution rates of different emission source types is input into the adjoint-meta-heuristic optimization algorithm model to obtain the predicted three-dimensional spatiotemporal distribution of the optimal emission reduction for different emission types.

[0011] Furthermore, in step (1), preprocessing is performed as follows: data cleaning and quality control are performed on the multi-source heterogeneous data set; abnormal, missing, invalid and duplicate entries in the multi-source heterogeneous data are initially eliminated based on the temporal and spatial continuity, typical size, data structure and file integrity characteristics of the data set, and the problematic data are cleared and corrected to produce a complete local multi-source observation data set.

[0012] Furthermore, step (2) is specifically as follows: the three-dimensional spatial grid resolution of the atmospheric chemical transport model is set to 3km×3km, and the local multi-source observation station data is spatially resampled using the bilinear interpolation method according to the central longitude and latitude of the three-dimensional spatial grid to generate a spatiotemporal standard gridded meteorological and emission data set; mathematical and statistical analysis is performed on the data set to extract the spatiotemporal variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify key control areas and key emission reduction periods in different seasons and months, and generate a concentration field driving data file accompanying the model, the file format of which is the network common data format, namely, netCDF format.

[0013] Furthermore, in step (3), the atmospheric chemical transport model WRF-chem is run to predict the atmospheric chemical pollutant concentration field, and a three-dimensional spatiotemporal grid dataset of hourly atmospheric pollutant mass concentrations is obtained. The model prediction data time series values ​​are extracted according to the latitude and longitude of the observation site, and the data are compared and verified point-to-point with the hourly air quality monitoring data. The physical and chemical parameterization scheme of the optimization model is adjusted until the prediction results meet the accuracy requirement of more than 90%, and finally a high-resolution three-dimensional spatiotemporal grid dataset of atmospheric pollutant mass concentrations is produced.

[0014] Furthermore, step (4) includes the following steps:

[0015] (41) At each step, the slope of the response function of concentration change to emission (dy / dx i ), the dy / dx of each grid i The value is compared with the slope change value (Δy / Δx) obtained by the finite difference method, where the determination coefficient R 2 =1;

[0016] (42) Set the accompanying model driving data file as the spatial distribution value y(i,j) of the hourly average mass concentration of ozone in the target area, where i and j are the grid point index row and column numbers respectively; construct the three-dimensional spatiotemporal distribution of the accompanying simulation calculation source-receptor response relationship function dy=f'(x)dx, where dx is the total grid emission change value; f'(x) is the three-dimensional spatiotemporal distribution of the accompanying sensitivity, which is multiplied by the intensity of the emission source to obtain the three-dimensional spatiotemporal grid data set Contri of the absolute contribution rate of the corresponding emission type;

[0017] (43) The two constraints that the optimization model must satisfy are defined as follows: a. the emission of any emission source at any three-dimensional grid point cannot exceed the original emission after being predicted by the optimization algorithm, i.e. the model prediction result does not accept the increase of emissions after optimization; b. the spatial sum of the different emission types at all three-dimensional grid points after being predicted by the optimization algorithm must be equal to the total allowable emission, i.e. theoretically the total allowable emission after optimization remains unchanged and no error is allowed.

[0018] The atmospheric pollution spatiotemporal dynamic emission reduction system based on the meta-heuristic optimization algorithm and the sensitivity algorithm of the present invention comprises:

[0019] Acquisition preprocessing module: used to obtain local weather and meteorological observation data, ambient air quality monitoring data and emission inventory to construct multi-source heterogeneous data sets and perform preprocessing;

[0020] Driving data module: used to perform mathematical and statistical analysis on the data set, extract the temporal and spatial variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify key control areas and key emission reduction periods in different seasons and months, and generate concentration field driving data files accompanying the model;

[0021] Three-dimensional spatiotemporal grid module: used to predict the three-dimensional spatiotemporal distribution of concentration field using the atmospheric chemical transport model, and to evaluate and correct it using the extracted hourly concentration data of the observation station; to adjust the physical and chemical parameterization scheme of the optimization model until the prediction results meet the accuracy requirements, and finally to produce a high-resolution three-dimensional spatiotemporal grid dataset of atmospheric pollutant mass concentration;

[0022] Adjoint-Meta-heuristic Optimization Module: It is used to introduce the sensitivity algorithm into the atmospheric chemical transport model based on the prepared concentration field dataset and model-driven data file, build and accompany the meta-heuristic optimization algorithm model, and calculate the absolute contribution rate of different emission source types in the three-dimensional space-time grid dataset;

[0023] Three-dimensional spatiotemporal distribution module: It is used to input the three-dimensional spatiotemporal grid data set of absolute contribution rates of different emission source types into the adjoint-meta-heuristic optimization algorithm model to obtain the predicted three-dimensional spatiotemporal distribution of optimal emission reduction for different emission types.

[0024] Furthermore, in the acquisition preprocessing module, the preprocessing is as follows: data cleaning and quality control are performed on the multi-source heterogeneous data sets; abnormal, missing, invalid and duplicate entries in the multi-source heterogeneous data are preliminarily eliminated based on the time and space continuity, typical size, data structure and file integrity characteristics in the data sets, and the problematic data are cleared and corrected to produce a complete local multi-source observation data set.

[0025] Furthermore, in the driving data module, the details are as follows: the three-dimensional spatial grid resolution of the atmospheric chemical transport model is set to 3km×3km, and the bilinear interpolation method is used to spatially resample the local multi-source observation station data according to the longitude and latitude of the three-dimensional spatial grid center to generate a spatiotemporal standard gridded meteorological and emission data set; mathematical and statistical analysis is performed on the data set to extract the spatiotemporal variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify key control areas and key emission reduction periods in different seasons and months, and generate a concentration field driving data file for the accompanying model. The file format is the network common data format, namely netCDF format.

[0026] Furthermore, in the three-dimensional space-time grid module, the atmospheric chemical transport model WRF-chem is run to predict the atmospheric chemical pollutant concentration field, and the three-dimensional space-time grid dataset of hourly atmospheric pollutant mass concentration is obtained. The data time series values ​​are predicted according to the longitude and latitude extraction mode of the observation site, and are compared and verified point-to-point with the hourly air quality monitoring data. The physical and chemical parameterization scheme of the optimization model is adjusted until the prediction results meet the accuracy requirement of more than 90%, and finally a high-resolution three-dimensional space-time grid dataset of atmospheric pollutant mass concentration is produced.

[0027] Furthermore, the adjoint-meta-heuristic optimization module includes the following steps:

[0028] (41) At each step, the slope of the response function of concentration change to emission (dy / dx i ), the dy / dx of each grid i The value is compared with the slope change value (Δy / Δx) obtained by the finite difference method, where the determination coefficient R 2=1;

[0029] (42) Set the accompanying model driving data file as the spatial distribution value y(i,j) of the hourly average mass concentration of ozone in the target area, where i and j are the grid point index row and column numbers respectively; construct the three-dimensional spatiotemporal distribution of the accompanying simulation calculation source-receptor response relationship function dy=f'(x)dx, where dx is the total grid emission change value; f'(x) is the three-dimensional spatiotemporal distribution of the accompanying sensitivity, which is multiplied by the intensity of the emission source to obtain the three-dimensional spatiotemporal grid data set Contri of the absolute contribution rate of the corresponding emission type;

[0030] (43) The two constraints that the optimization model must satisfy are defined as follows: a. the emission of any emission source at any three-dimensional grid point cannot exceed the original emission after being predicted by the optimization algorithm, i.e. the model prediction result does not accept the increase of emissions after optimization; b. the spatial sum of the different emission types at all three-dimensional grid points after being predicted by the optimization algorithm must be equal to the total allowable emission, i.e. theoretically the total allowable emission after optimization remains unchanged and no error is allowed.

[0031] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: It provides a spatiotemporal dynamic emission reduction method for atmospheric pollutants that integrates a metaheuristic optimization algorithm and an accompanying sensitivity algorithm, which satisfies the premise of allowable total emissions and establishes the optimal emission reduction method for atmospheric pollutants in different emission sectors, different emission types, and different grid points in the target area at different times, providing a certain reference basis for improving regional atmospheric environmental quality and enhancing cost-effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The prediction accuracy comparison and verification results of the present invention are as follows;

[0033] Figure 2 is the genetic algorithm of the present invention, wherein Figure 2 (a) is NIND and Figure 2 (b) in the figure is the result of MAXGEN parameter sensitivity test;

[0034] Figure 3 It is a schematic diagram of the genetic algorithm flow of the present invention;

[0035] Figure 4 The prediction results of the genetic algorithm of the present invention are compared with those of the simulated annealing algorithm and the random hill climbing algorithm;

[0036] Figure 5 It is the statistical result of the effectiveness evaluation of the spatiotemporal dynamic reduction scheme of nitrogen oxide emissions of the present invention. DETAILED DESCRIPTION

[0037] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0038] like Figure 1 As shown, the embodiment of the present invention provides a method for spatiotemporal dynamic reduction of atmospheric pollution based on a meta-heuristic optimization algorithm and a sensitivity algorithm, comprising the following steps:

[0039] (1) Obtain local weather and meteorological observation data, ambient air quality monitoring data, and emission inventories to construct a multi-source heterogeneous data set and perform preprocessing. The preprocessing is as follows: perform data cleaning and quality control on the multi-source heterogeneous data set; preliminarily eliminate abnormal, missing, invalid, and duplicate entries in the multi-source heterogeneous data based on the temporal and spatial continuity, typical size, data structure, and file integrity characteristics of the data set, clear and correct the problematic data, and produce a complete local multi-source observation data set.

[0040] (2) Perform mathematical and statistical analysis on the data set, extract the spatiotemporal variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify the key control areas and key emission reduction periods in different seasons and months, and generate the concentration field driving data file of the accompanying model; specifically, set the three-dimensional spatial grid resolution of the atmospheric chemical transport model to 3km×3km, and use the bilinear interpolation method to spatially resample the local multi-source observation station data according to the longitude and latitude of the three-dimensional spatial grid center to generate a spatiotemporal standard gridded meteorological and emission data set; perform mathematical and statistical analysis on the data set, extract the spatiotemporal variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify the key control areas and key emission reduction periods in different seasons and months, and generate the concentration field driving data file of the accompanying model. The file format is the network common data format, i.e., netCDF format.

[0041] (3) The atmospheric chemical transport model is used to predict the three-dimensional spatiotemporal distribution of the concentration field, and the hourly concentration data extracted from the observation stations are used for evaluation and correction; the physical and chemical parameterization scheme of the optimization model is adjusted until the prediction results meet the accuracy requirements, and finally a high-resolution three-dimensional spatiotemporal grid dataset of atmospheric pollutant mass concentration is produced; the atmospheric chemical transport model WRF-chem is run to predict the atmospheric chemical pollutant concentration field, and a three-dimensional spatiotemporal grid dataset of hourly atmospheric pollutant mass concentration is obtained. The time series values ​​of the predicted data are extracted according to the longitude and latitude of the observation station, and a point-to-point comparison is performed with the hourly air quality monitoring data. The physical and chemical parameterization scheme of the optimization model is adjusted until the prediction results meet the accuracy requirements of more than 90%, and finally a three-dimensional spatiotemporal grid dataset of high-resolution atmospheric pollutant mass concentration is produced.

[0042] (4) Based on the prepared concentration field dataset and model-driven data file, the sensitivity algorithm is introduced into the atmospheric chemical transport model, and a meta-heuristic optimization algorithm model is constructed to calculate the absolute contribution rate of different emission source types in a three-dimensional spatiotemporal grid dataset; the following steps are included:

[0043] (41) At each step, the slope of the response function of concentration change to emission (dy / dx i ), the dy / dx of each grid i The value is compared with the slope change value (Δy / Δx) obtained by the finite difference method, where the determination coefficient R 2 =1; the comparison and verification results are as follows Figure 1 shown.

[0044] (42) Set the accompanying model driving data file as the spatial distribution value y(i,j) of the hourly average mass concentration of ozone in the target area, where i and j are the grid point index row and column numbers respectively; construct the three-dimensional spatiotemporal distribution of the accompanying simulation calculation source-receptor response relationship function dy=f'(x)dx, where dx is the total grid emission change value; f'(x) is the three-dimensional spatiotemporal distribution of the accompanying sensitivity, which is multiplied by the intensity of the emission source to obtain the three-dimensional spatiotemporal grid data set Contri of the absolute contribution rate of the corresponding emission type;

[0045] (43) The two constraints that the optimization model must satisfy are defined as follows: a. the emission of any emission source at any three-dimensional grid point cannot exceed the original emission after being predicted by the optimization algorithm, i.e. the model prediction result does not accept the increase of emissions after optimization; b. the spatial sum of the different emission types at all three-dimensional grid points after being predicted by the optimization algorithm must be equal to the total allowable emission, i.e. theoretically the total allowable emission after optimization remains unchanged and no error is allowed.

[0046] The genetic algorithm (GA) model in the typical metaheuristic optimization algorithm was run, and the crossover probability was defined as 0.7 and the mutation probability was defined as 0.01. Sensitivity experiments were carried out on the key parameters of the model such as population size (NIND) and evolutionary generations (MAXGEN), and the optimal values ​​of NIND and MAXGEN were determined to be 400 and 180, respectively.

[0047] The GA optimization process is as follows: (S1) Parameter initialization: randomly generate an initial population, each individual represents a potential solution to the problem; (S2) Fitness evaluation: calculate the fitness value of each individual in the initial population, and the fitness value indicates the performance of each individual in the population in solving the problem; (S3) Select excellent individuals: define the survival of the fittest selection strategy, select a part of individuals with higher fitness values ​​as parents, and use them to generate the next generation of population; (S4) Gene crossover: select two random individuals from the excellent parents in (S3), apply the crossover operator and crossover rate, and perform gene crossover to generate new individuals; (S5) Mutation: apply the mutation operator and mutation rate to the newly generated individuals in (S4) in order to maintain the diversity of the population; (S6) Iteration: add the newly generated offspring in (S5) to the original population to form a new generation of population, and continue to repeat the above iterative steps starting from the new generation of population until the iteration meets the stopping condition.

[0048] The optimization process was repeated 100 times, and the average of the 100 hourly emission reduction spatial distribution prediction results was calculated to reduce the accuracy error caused by random initialization of GA. Finally, the average result was multiplied by the obtained sensitivity coefficient to produce a three-dimensional grid spatiotemporal dynamic data set for the optimal emission reduction of different emission types, while providing the optimal prediction data sets at different time scales and spatial scales.

[0049] (5) The three-dimensional spatiotemporal grid dataset of the absolute contribution rates of different emission source types is input into the adjoint-meta-heuristic optimization algorithm model to obtain the predicted three-dimensional spatiotemporal distribution of the optimal emission reduction for different emission types.

[0050] Based on the optimized prediction data set of spatiotemporal dynamic emission reduction in three-dimensional grids, the spatiotemporal quotas of emission reduction of each pollutant in different target areas are calculated after the emission reduction amount of each pollutant is scientifically and reasonably allocated under the condition of meeting the total emission reduction control condition. According to the quota size, the optimal emission reduction plan is formulated and the effectiveness is evaluated for different emission types in different locations, different time periods, different industries and sectors; according to the optimal emission reduction plan, the key emission types and key time periods in the key control areas are further identified, and the emission reduction benefits of different emission industries and types are statistically analyzed to achieve spatiotemporal dynamic optimization of emission reduction and rapid evaluation of multiple pollutants, providing a certain reference basis for improving regional air environmental quality and enhancing cost-effectiveness.

[0051] Specific experimental data:

[0052] Taking the Pearl River Delta as an example, the nitrogen oxides NO x The upper limit of total emission reduction is 18%. The obtained Contri is used as the input of GA model. Assuming that NO xA random spatial distribution of emissions e(x, y) is used as the initial solution to obtain the corresponding initial objective function value M0. After random perturbation of e(x, y) in each iteration, a new M is generated. i Value, according to NO x The 18% reduction target, after optimization, the ozone mass concentration value M in the target area i Reach the global minimum; the specific process is as follows:

[0053] First, we conducted sensitivity experiments on key parameters of the GA model, such as population size (NIND) and evolutionary generations (MAXGEN). We defined the NIND value range as 50 to 500 (interval 50), the MAXGEN value range as 20 to 200 (interval 20), and performed a sensitivity analysis on NIND when MAXGEN was 200. The results are as follows: Figure 2 (a) shows; when NIND is defined as 400, a sensitivity analysis of MAXGEN is performed, and the results are as follows Figure 2 As shown in (b), the optimal combination of NIND and MAXGEN is finally determined to be 400 and 180.

[0054] As shown Figure 3 The GA technical flow chart shown in the figure has the following specific implementation steps:

[0055] (S1) Parameter initialization: Randomly generate an initial population e(x,y), each individual represents a potential solution to the problem;

[0056] (S2) Evaluating fitness: Substituting each individual in the initial population into the objective function to calculate its fitness value, the fitness value indicates the performance of each individual in the population in solving the problem. In this embodiment, the fitness value is the change value Mi of the ozone mass concentration;

[0057] (S3) Select excellent individuals: define the survival of the fittest selection strategy, select a part of individuals with higher fitness values ​​as parents, and use them to generate the next generation of population;

[0058] (S4) Genetic crossover: Select two random individuals from the excellent parents in (S3), define the crossover probability as 0.7, apply the crossover operator and crossover rate, and perform genetic crossover to generate new individuals;

[0059] (S5) Mutation: Apply the mutation operator and mutation rate to the newly generated individuals in (S4). The mutation probability is defined as 0.01 to maintain the diversity of the population.

[0060] (S6) Iteration: Add the newly generated offspring in (S5) to the original population to form a new generation population, and continue to repeat the above iterative steps starting from the new generation population until the iteration meets the stopping condition.

[0061] The optimization process was repeated 100 times, and the average of the 100 hourly emission reduction spatial distribution prediction results was calculated to reduce the precision error caused by random initialization of GA.

[0062] Run other typical meta-inspired optimization algorithms such as random hill climbing algorithm and simulated annealing algorithm, and use the prediction results to compare the prediction results of GA model. The comparison results are as follows: Figure 4 As shown in the figure, the prediction accuracy of the GA model is much higher than that of the other two algorithm models. The GA prediction results are multiplied by the obtained sensitivity coefficient to produce a three-dimensional grid spatiotemporal dynamic data set for optimal emission reduction of different emission types, and at the same time provide the optimal prediction data set at different time scales and spatial scales.

[0063] Based on the three-dimensional grid spatiotemporal dynamic data set of optimal emission reduction, the effectiveness evaluation of the optimal emission reduction scheme for different emission types in different industries and sectors at different locations and time periods is realized, and the key control areas for different emission types are further identified based on this. Figure 5 The NO of each city in the Pearl River Delta obtained in this implementation case x The statistical results of the effectiveness evaluation of the optimal emission reduction plan for pollutants show that through the NO x Emission reductions can more effectively reduce ozone concentration levels and improve policy cost-effectiveness.

[0064] The embodiment of the present invention also provides an atmospheric pollution spatiotemporal dynamic emission reduction system based on a meta-heuristic optimization algorithm and a sensitivity algorithm, including:

[0065] Collection and preprocessing module: used to obtain local weather and meteorological observation data, ambient air quality monitoring data and emission inventories to construct multi-source heterogeneous data sets and perform preprocessing; the preprocessing is as follows: data cleaning and quality control of multi-source heterogeneous data sets; based on the time and space continuity, typical size, data structure, and file integrity characteristics of the data sets, preliminarily eliminate abnormal, missing, invalid, and duplicate entries in the multi-source heterogeneous data, clear and correct problematic data, and produce a complete local multi-source observation data set.

[0066] Driving data module: used to perform mathematical and statistical analysis on the data set, extract the temporal and spatial variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify the key control areas and key emission reduction periods in different seasons and months, and generate the concentration field driving data file of the accompanying model; specifically as follows: set the three-dimensional spatial grid resolution of the atmospheric chemical transport model to 3km×3km, and use the bilinear interpolation method to spatially resample the local multi-source observation station data according to the longitude and latitude of the three-dimensional spatial grid center to generate a temporal and spatial standard gridded meteorological and emission data set; perform mathematical and statistical analysis on the data set, extract the temporal and spatial variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify the key control areas and key emission reduction periods in different seasons and months, and generate the concentration field driving data file of the accompanying model. The file format is the network common data format, namely netCDF format.

[0067] Three-dimensional space-time grid module: used to use the atmospheric chemical transport model to predict the three-dimensional space-time distribution of the concentration field, and use the extracted hourly concentration data of the observation station for evaluation and correction; adjust the physical and chemical parameterization scheme of the optimization model until the prediction result meets the accuracy requirement, and finally produce a high-resolution three-dimensional space-time grid data set of atmospheric pollutant mass concentration; specifically as follows: run the atmospheric chemical transport model WRF-chem to predict the atmospheric chemical pollutant concentration field, obtain a three-dimensional space-time grid data set of hourly atmospheric pollutant mass concentration, extract the model prediction data time series value according to the longitude and latitude of the observation station, and compare it with the hourly air quality monitoring data for point-to-point verification, adjust the physical and chemical parameterization scheme of the optimization model until the prediction result meets the accuracy requirement of more than 90%, and finally produce a three-dimensional space-time grid data set of high-resolution atmospheric pollutant mass concentration.

[0068] Adjoint-meta-heuristic optimization module: used to introduce the sensitivity algorithm into the atmospheric chemical transport model based on the prepared concentration field dataset and model-driven data file, build and accompany the meta-heuristic optimization algorithm model, and calculate the absolute contribution rate of different emission source types in the three-dimensional space-time grid dataset; including the following steps:

[0069] (41) At each step, the slope of the response function of concentration change to emission (dy / dx i ), the dy / dx of each grid i The value is compared with the slope change value (Δy / Δx) obtained by the finite difference method, where the determination coefficient R 2 =1;

[0070] (42) Set the accompanying model driving data file as the spatial distribution value y(i,j) of the hourly average mass concentration of ozone in the target area, where i and j are the grid point index row and column numbers respectively; construct the three-dimensional spatiotemporal distribution of the accompanying simulation calculation source-receptor response relationship function dy=f'(x)dx, where dx is the total grid emission change value; f'(x) is the three-dimensional spatiotemporal distribution of the accompanying sensitivity, which is multiplied by the intensity of the emission source to obtain the three-dimensional spatiotemporal grid data set Contri of the absolute contribution rate of the corresponding emission type;

[0071] (43) The two constraints that the optimization model must satisfy are defined as follows: a. the emission of any emission source at any three-dimensional grid point cannot exceed the original emission after being predicted by the optimization algorithm, i.e. the model prediction result does not accept the increase of emissions after optimization; b. the spatial sum of the different emission types at all three-dimensional grid points after being predicted by the optimization algorithm must be equal to the total allowable emission, i.e. theoretically the total allowable emission after optimization remains unchanged and no error is allowed.

[0072] Three-dimensional spatiotemporal distribution module: It is used to input the three-dimensional spatiotemporal grid data set of absolute contribution rates of different emission source types into the adjoint-meta-heuristic optimization algorithm model to obtain the predicted three-dimensional spatiotemporal distribution of optimal emission reduction for different emission types.

Claims

1. A method for spatiotemporal dynamic reduction of atmospheric pollution based on meta-heuristic optimization algorithm and sensitivity algorithm, characterized in that: The following steps are involved: (1) Obtain local weather and meteorological observation data, ambient air quality monitoring data, and emission inventories to construct a multi-source heterogeneous data set and perform preprocessing; (2) Perform mathematical and statistical analysis on the data set, extract the temporal and spatial variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify key control areas and key emission reduction periods in different seasons and months, and generate concentration field driving data files accompanying the model; (3) Use the atmospheric chemical transport model to predict the three-dimensional spatiotemporal distribution of the concentration field, and use the extracted hourly concentration data of the observation station to evaluate and correct it; adjust the physical and chemical parameterization scheme of the optimization model until the prediction results meet the accuracy requirements, and finally produce a high-resolution three-dimensional spatiotemporal grid dataset of atmospheric pollutant mass concentrations; (4) Based on the prepared concentration field dataset and model-driven data file, the sensitivity algorithm is introduced into the atmospheric chemical transport model, and a meta-heuristic optimization algorithm model is constructed to calculate the absolute contribution rate of different emission source types in a three-dimensional spatiotemporal grid dataset; (5) The three-dimensional spatiotemporal grid dataset of the absolute contribution rates of different emission source types is input into the adjoint-meta-heuristic optimization algorithm model to obtain the predicted three-dimensional spatiotemporal distribution of the optimal emission reduction for different emission types.

2. The method for spatiotemporal dynamic reduction of atmospheric pollution based on metaheuristic optimization algorithm and sensitivity algorithm according to claim 1 is characterized in that: In step (1), the preprocessing is as follows: data cleaning and quality control are performed on the multi-source heterogeneous data set; based on the temporal and spatial continuity, typical size, data structure, and file integrity characteristics of the data set, abnormal, missing, invalid, and duplicate entries in the multi-source heterogeneous data are initially eliminated, and problematic data are cleared and corrected to produce a complete local multi-source observation data set.

3. The method for spatiotemporal dynamic reduction of atmospheric pollution based on metaheuristic optimization algorithm and sensitivity algorithm according to claim 1 is characterized in that: Step (2) is as follows: set the three-dimensional spatial grid resolution of the atmospheric chemical transport model to 3km×3km, and use the bilinear interpolation method to spatially resample the local multi-source observation station data according to the longitude and latitude of the three-dimensional spatial grid center to generate a spatiotemporal standard gridded meteorological and emission data set; perform mathematical and statistical analysis on the data set, extract the spatiotemporal variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify key control areas and key emission reduction periods in different seasons and months, and generate a concentration field driving data file for the accompanying model. The file format is the network common data format, i.e., netCDF format.

4. The method for spatiotemporal dynamic reduction of atmospheric pollution based on metaheuristic optimization algorithm and sensitivity algorithm according to claim 1 is characterized in that: In step (3), the atmospheric chemical transport model WRF-chem is run to predict the atmospheric chemical pollutant concentration field, and a three-dimensional spatiotemporal grid dataset of hourly atmospheric pollutant mass concentrations is obtained. The model prediction data time series values ​​are extracted according to the latitude and longitude of the observation site, and the data are compared and verified point-to-point with the hourly air quality monitoring data. The physical and chemical parameterization scheme of the optimization model is adjusted until the prediction results meet the accuracy requirement of more than 90%, and finally a high-resolution three-dimensional spatiotemporal grid dataset of atmospheric pollutant mass concentrations is produced.

5. The method for spatiotemporal dynamic reduction of atmospheric pollution based on metaheuristic optimization algorithm and sensitivity algorithm according to claim 1 is characterized in that: Step (4) comprises the following steps: (41) At each step, the slope of the response function of concentration change to emission (dy / dx i ), the dy / dx of each grid i The value is compared with the slope change value (Δy / Δx) obtained by the finite difference method, where the determination coefficient R 2 =1; (42) Set the accompanying model driving data file as the spatial distribution value y(i,j) of the hourly average mass concentration of ozone in the target area, where i and j are the grid point index row and column numbers respectively; construct the three-dimensional spatiotemporal distribution of the accompanying simulation calculation source-receptor response relationship function dy=f'(x)dx, where dx is the total grid emission change value; f'(x) is the three-dimensional spatiotemporal distribution of the accompanying sensitivity, which is multiplied by the intensity of the emission source to obtain the three-dimensional spatiotemporal grid data set Contri of the absolute contribution rate of the corresponding emission type; (43) The two constraints that the optimization model must satisfy are defined as follows: a. the emission of any emission source at any three-dimensional grid point cannot exceed the original emission after being predicted by the optimization algorithm, i.e. the model prediction result does not accept the increase of emissions after optimization; b. the spatial sum of the different emission types at all three-dimensional grid points after being predicted by the optimization algorithm must be equal to the total allowable emission, i.e. theoretically the total allowable emission after optimization remains unchanged and no error is allowed.

6. A spatiotemporal dynamic atmospheric pollution reduction system based on meta-heuristic optimization algorithm and sensitivity algorithm, characterized in that: include: Acquisition preprocessing module: used to obtain local weather and meteorological observation data, ambient air quality monitoring data and emission inventory to construct multi-source heterogeneous data sets and perform preprocessing; Driving data module: used to perform mathematical and statistical analysis on the data set, extract the temporal and spatial variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify key control areas and key emission reduction periods in different seasons and months, and generate concentration field driving data files accompanying the model; Three-dimensional spatiotemporal grid module: used to predict the three-dimensional spatiotemporal distribution of concentration field using the atmospheric chemical transport model, and to evaluate and correct it using the extracted hourly concentration data of the observation station; to adjust the physical and chemical parameterization scheme of the optimization model until the prediction results meet the accuracy requirements, and finally to produce a high-resolution three-dimensional spatiotemporal grid dataset of atmospheric pollutant mass concentration; Adjoint-Meta-heuristic Optimization Module: It is used to introduce the sensitivity algorithm into the atmospheric chemical transport model based on the prepared concentration field dataset and model-driven data file, build and accompany the meta-heuristic optimization algorithm model, and calculate the absolute contribution rate of different emission source types in the three-dimensional space-time grid dataset; Three-dimensional spatiotemporal distribution module: It is used to input the three-dimensional spatiotemporal grid data set of absolute contribution rates of different emission source types into the adjoint-meta-heuristic optimization algorithm model to obtain the predicted three-dimensional spatiotemporal distribution of optimal emission reduction for different emission types.

7. The atmospheric pollution spatiotemporal dynamic emission reduction system based on meta-heuristic optimization algorithm and sensitivity algorithm according to claim 6 is characterized in that: In the acquisition preprocessing module, the preprocessing is as follows: data cleaning and quality control of multi-source heterogeneous data sets; based on the time and space continuity, typical size, data structure, and file integrity characteristics of the data sets, the abnormal, missing, invalid, and duplicate entries in the multi-source heterogeneous data are initially eliminated, and the problematic data are cleared and corrected to produce a complete local multi-source observation data set.

8. The atmospheric pollution spatiotemporal dynamic emission reduction system based on meta-heuristic optimization algorithm and sensitivity algorithm according to claim 6 is characterized in that: In the driving data module, the details are as follows: the three-dimensional spatial grid resolution of the atmospheric chemical transport model is set to 3km×3km, and the local multi-source observation station data is spatially resampled using the bilinear interpolation method according to the central longitude and latitude of the three-dimensional spatial grid to generate a spatiotemporal standard gridded meteorological and emission data set; mathematical and statistical analysis is performed on the data set to extract the spatiotemporal variation characteristic curves of meteorological conditions and atmospheric pollutants in different target areas, identify key control areas and key emission reduction periods in different seasons and months, and generate a concentration field driving data file for the accompanying model. The file format is the network common data format, namely netCDF format.

9. The atmospheric pollution spatiotemporal dynamic emission reduction system based on meta-heuristic optimization algorithm and sensitivity algorithm according to claim 6 is characterized in that: In the three-dimensional space-time grid module, the atmospheric chemical transport model WRF-chem is run to predict the atmospheric chemical pollutant concentration field, and a three-dimensional space-time grid dataset of hourly atmospheric pollutant mass concentrations is obtained. The data time series values ​​are predicted based on the longitude and latitude extraction mode of the observation site, and are compared and verified point-to-point with the hourly air quality monitoring data. The physical and chemical parameterization scheme of the optimization model is adjusted until the prediction results meet the accuracy requirement of more than 90%, and finally a high-resolution three-dimensional space-time grid dataset of atmospheric pollutant mass concentrations is produced.

10. The atmospheric pollution spatiotemporal dynamic emission reduction system based on meta-heuristic optimization algorithm and sensitivity algorithm according to claim 6, characterized in that: The adjoint-meta-heuristic optimization module includes the following steps: (41) At each step, the slope of the response function of concentration change to emission (dy / dx i ), the dy / dx of each grid i The value is compared with the slope change value (Δy / Δx) obtained by the finite difference method, where the determination coefficient R 2 =1; (42) Set the accompanying model driving data file as the spatial distribution value y(i,j) of the hourly average mass concentration of ozone in the target area, where i and j are the grid point index row and column numbers respectively; construct the three-dimensional spatiotemporal distribution of the accompanying simulation calculation source-receptor response relationship function dy=f'(x)dx, where dx is the total grid emission change value; f'(x) is the three-dimensional spatiotemporal distribution of the accompanying sensitivity, which is multiplied by the intensity of the emission source to obtain the three-dimensional spatiotemporal grid data set Contri of the absolute contribution rate of the corresponding emission type; (43) The two constraints that the optimization model must satisfy are defined as follows: a. the emission of any emission source at any three-dimensional grid point cannot exceed the original emission after being predicted by the optimization algorithm, i.e. the model prediction result does not accept the increase of emissions after optimization; b. the spatial sum of the different emission types at all three-dimensional grid points after being predicted by the optimization algorithm must be equal to the total allowable emission, i.e. theoretically the total allowable emission after optimization remains unchanged and no error is allowed.

Images