Method and system for meteorological normalization and fire point constrained identification of non-meteorological contributions of particulate matter

Abstract

The application discloses a meteorological normalization and fire point constraint particle non-meteorological contribution identification method and system. The particle concentration, meteorology, fire point remote sensing and geographical space data in a target area are obtained, and after spatial matching and time alignment, a pollution feature set containing meteorological characteristics, time characteristics, geographical space constraint characteristics and fire point exposure characteristics is constructed; a concentration prediction model is established by using a machine learning model, meteorological normalization concentration is obtained by multiple random resampling of meteorological characteristics, and meteorological contribution and non-meteorological contribution are decomposed accordingly; multi-dimensional constraint discrimination is carried out by comprehensively considering non-meteorological contribution, fire point exposure and geographical space constraint, and pollution anomaly identification and non-meteorological contribution attribution results are output. The application can effectively strip meteorological interference, accurately identify fire point related pollution in combination with time synchronization, spatial distance, upwind direction, land cover and other constraints, and is suitable for regional particle pollution diagnosis and prevention and control decision support.

Description

Technical Field

[0001] This invention relates to the fields of atmospheric environmental monitoring, environmental data processing, and pollutant source attribution technology, and in particular to a method, system, device, and storage medium for identifying non-meteorological contributions of particulate matter based on meteorological normalization and geospatial fire point multidimensional constraint discrimination. Background Technology

[0002] Particulate matter concentration changes are influenced by both meteorological conditions and emission activities. Meteorological conditions (boundary layer height, temperature, humidity, wind speed, wind direction, precipitation, etc.) affect the formation, diffusion, and removal of particulate matter, while non-meteorological factors such as fire activity can also lead to abnormal increases in concentration. Therefore, it is difficult to distinguish the dominant factor of pollution increase based on the original concentration alone.

[0003] Existing attribution methods mainly fall into three categories. Statistical analysis methods are simple to implement but struggle to characterize the nonlinear relationship between meteorological events and pollution sources; chemical transport models rely on complex emission inventories, resulting in high computational costs and difficulty in meeting rapid diagnostic needs; machine learning models, while capable of fitting complex relationships, are primarily used for concentration prediction and lack sufficient integration of meteorological disturbance stripping and non-meteorological contribution extraction. Regarding meteorological attribution, methods using fixed baseline years or emission scenarios are easily influenced by baseline selection and are insensitive to spatially discrete and temporally sudden fire activity. While fire point remote sensing data can provide information such as location, time, and power, the sheer number of fire points alone cannot determine whether they truly correspond to an abnormal increase in particulate matter due to non-meteorological contributions.

[0004] Therefore, there is an urgent need for a unified method for processing particulate matter concentration, meteorological, fire point remote sensing, and geospatial data. This method should reduce meteorological interference through meteorological normalization and combine multi-dimensional constraints such as fire point exposure, spatial distance, upwind relationship, time lag, and land cover to achieve reliable identification of fire point-related pollution anomalies and accurate attribution of non-meteorological contributions. Summary of the Invention

[0005] (a) Purpose of the invention: The purpose of this invention is to provide a method and system for identifying non-meteorological contributions to particulate matter based on meteorological normalization and fire point constraints, so as to solve the problems in existing particulate matter pollution attribution methods, such as the difficulty in separating meteorological disturbances, insufficient identification of sudden non-meteorological sources such as fire points, insufficient utilization of geospatial constraints, and weak automatic identification capability of pollution anomalies.

[0006] (II) Technical Solution: A method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints, characterized by the following steps: S1. Acquire particulate matter concentration data, meteorological data, fire point remote sensing data, and geospatial data from multiple cities, administrative regions, or monitoring stations within the target area; S2. Based on geospatial data, spatial matching and temporal alignment are performed on particulate matter concentration data, meteorological data and fire point remote sensing data to obtain spatiotemporal sample data corresponding to each city, administrative region or monitoring station in the target area. S3. Construct a particulate matter pollution feature set based on spatiotemporal sample data. The particulate matter pollution feature set includes meteorological features, temporal features, geospatial constraint features, and fire point exposure features. S4. Using particulate matter concentration as the target variable, establish a particulate matter concentration prediction model using a machine learning model. S5. While keeping the time characteristics and geospatial constraints unchanged, the meteorological characteristics are randomly resampled multiple times, and the resampled samples are input into the particulate matter concentration prediction model to obtain multiple sets of particulate matter predicted concentrations. S6. Calculate the meteorological normalized particulate matter concentration based on multiple sets of predicted particulate matter concentrations, and determine the meteorological contribution and non-meteorological contribution of particulate matter based on the difference between the observed particulate matter concentration and the meteorological normalized particulate matter concentration. S7. Construct multidimensional constraint discrimination results for geospatial fire points based on particulate matter non-meteorological contribution, fire point exposure characteristics, and geospatial constraint characteristics; S8. Based on the meteorological contribution of particulate matter, the non-meteorological contribution of particulate matter, and the multi-dimensional constraint discrimination results of geospatial fire points, output the anomaly identification results of particulate matter pollution and the attribution results of non-meteorological contributions.

[0007] Preferably, the particulate matter concentration data in S1 includes one or more of PM2.5, PM10, and PM2.5-10; the meteorological data includes one or more of boundary layer height, temperature, humidity, dew point temperature, air pressure, wind speed, wind direction, precipitation, solar radiation, and cloud cover; the fire point remote sensing data includes one or more of fire point latitude and longitude, fire point occurrence time, fire point radiation power, fire point confidence level, and satellite sensor identifier; and the geospatial data includes one or more of the following: urban administrative boundary, monitoring station latitude and longitude, urban center location, land cover type, farmland distribution, forest distribution, construction land distribution, road density, population density, nighttime light intensity, terrain height, and slope.

[0008] Preferably, the geospatial constraint features in S3 include one or more of the following: spatial mean features within city boundaries, preset buffer zone features, inter-city distance features, distance attenuation weight features, upwind spatial relationship features, land cover ratio features, and human activity intensity features; the fire point exposure features include one or more of the following: number of fire points within a preset spatial range, fire point density, total fire point radiation power, mean fire point radiation power, distance-weighted fire point intensity, upwind fire point intensity, number of fire points on different land cover types, and lag fire point exposure intensity within different time windows.

[0009] Preferably, the machine learning model in S4 includes one or more of the following: LightGBM model, XGBoost model, random forest model, gradient boosting tree model, support vector regression model, neural network model, and ensemble learning model; the input features of the machine learning model also include one or more of the following: periodic time encoding, long-term time variables, weekday variables, seasonal variables, and historical pollutant variation features; the machine learning model is used to establish the mapping relationship between meteorological features, time features, geospatial constraint features, fire point exposure features, and particulate matter concentration.

[0010] Preferably, in S5, multiple random resampling of meteorological features includes: extracting meteorological features from historical meteorological samples of the target city, administrative region, or monitoring station according to month, season, pollution level, or preset time window to form multiple meteorological disturbance samples; inputting the multiple meteorological disturbance samples into the particulate matter concentration prediction model; and performing mean, median, or quantile statistics on the multiple prediction results to obtain the meteorological normalized particulate matter concentration.

[0011] Preferably, the multidimensional constraint discrimination results of geospatial fire points in S7 are determined by fire point constraint discrimination indicators, which include one or more of the following: time synchronicity indicators, time lag indicators, spatial distance attenuation indicators, upwind consistency indicators, land cover type consistency indicators, and non-meteorological contribution anomaly intensity indicators. When the fire point constraint discrimination indicators meet the preset threshold conditions, the particulate matter pollution anomalies of the corresponding city, administrative region, or monitoring station in the corresponding time period are marked as fire point-related non-meteorological pollution anomalies.

[0012] Preferably, the non-meteorological contribution attribution results in S8 include one or more of the following: non-meteorological contribution intensity at the city or regional scale, meteorological contribution intensity, fire point exposure contribution level, pollution anomaly level, anomaly occurrence time, ranking of major driving factors, and spatial distribution results.

[0013] A particulate matter non-meteorological contribution identification system based on meteorological normalization and fire point constraints, characterized in that the method for implementing particulate matter non-meteorological contribution identification based on meteorological normalization and fire point constraints includes: The input data layer includes a data acquisition module, which is used to acquire particulate matter concentration data, meteorological data, fire point remote sensing data, and geospatial data from multiple cities, administrative regions, or monitoring stations within the target area. The data processing layer includes a spatiotemporal matching module. The spatiotemporal matching module receives the output data of the data acquisition module and performs spatial matching and temporal alignment on the particulate matter concentration data, meteorological data and fire point remote sensing data based on the geospatial data to obtain spatiotemporal sample data corresponding to each city, administrative region or monitoring station in the target area. The model calculation layer includes a feature construction module, a model training module, and a meteorological normalization module. The feature construction module receives the output data from the spatiotemporal matching module and constructs a particulate matter pollution feature set based on the spatiotemporal sample data. The particulate matter pollution feature set includes meteorological features, temporal features, geospatial constraint features, and fire point exposure features. The model training module receives the output data from the feature construction module and uses a machine learning model to establish a particulate matter concentration prediction model with particulate matter concentration as the target variable. The meteorological normalization module receives the output data from the model training module and, while keeping the temporal and geospatial constraint features unchanged, performs multiple random resampling of the meteorological features. The resampled samples are then input into the particulate matter concentration prediction model to obtain multiple sets of predicted particulate matter concentrations. The meteorological normalized particulate matter concentration is then calculated based on these multiple sets of predicted particulate matter concentrations. The attribution discrimination layer includes a contribution decomposition module and a fire point multidimensional constraint discrimination module. The contribution decomposition module receives the output data from the meteorological normalization module and determines the meteorological contribution and non-meteorological contribution of particulate matter based on the difference between the observed particulate matter concentration and the meteorological normalized particulate matter concentration. The fire point multidimensional constraint discrimination module receives the output data from the contribution decomposition module and constructs a geospatial fire point multidimensional constraint discrimination result based on the non-meteorological contribution of particulate matter, fire point exposure characteristics, and geospatial constraint characteristics. The result service layer includes a result output module. The result output module receives the output data of the fire point multidimensional constraint discrimination module and outputs particulate matter pollution anomaly identification results and non-meteorological contribution attribution results based on the particulate matter meteorological contribution, particulate matter non-meteorological contribution and geospatial fire point multidimensional constraint discrimination results.

[0014] An electronic device for performing a method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints, characterized in that it includes a processor for performing the above method, a memory, and a computer program stored in the memory and executable by the processor.

[0015] (III) Beneficial Effects: Compared with existing technical solutions, the present invention has the following beneficial effects: 1. This invention obtains meteorological normalized particulate matter concentration by performing multiple random resampling of meteorological characteristics, which can reduce the influence of a single meteorological condition or a single base year selection on the particulate matter attribution results, and make the separation of meteorological contribution and non-meteorological contribution more stable and reliable.

[0016] 2. This invention determines the non-meteorological contribution of particulate matter based on meteorological normalization, avoiding the direct judgment of the impact of fire points based on the original particulate matter concentration, thereby reducing the interference of adverse meteorological conditions on the identification of fire point-related pollution anomalies and improving the accuracy of fire point identification.

[0017] 3. This invention combines fire point exposure characteristics with multi-dimensional constraints such as time synchronization, time lag, spatial distance attenuation, upwind consistency, and land cover type, enabling multi-angle comprehensive judgment of fire point-related non-meteorological pollution anomalies, effectively improving the spatial specificity of pollution source identification and the ability to interpret results.

[0018] 4. This invention incorporates geospatial factors such as urban boundaries, buffer zones, land cover, intensity of human activities, and topography into a unified processing flow, making it suitable for anomaly analysis of particulate matter pollution at urban, regional, and multi-timescale levels, and possessing good scenario adaptability.

[0019] 5. This invention forms a complete automated process from multi-source data acquisition, spatiotemporal matching, feature construction, model training, meteorological normalization, contribution decomposition, fire point constraint discrimination to result output. It can be used for regional particulate matter pollution diagnosis, non-meteorological source identification, and pollution prevention and control auxiliary decision-making, and has high practical value and promotion significance.

[0020] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method and system for identifying non-meteorological contributions of particulate matter under meteorological normalization and fire point constraints according to the present invention. Figure 2 The system structure diagram of the particulate matter non-meteorological contribution identification method and system based on meteorological normalization and fire point constraint of the present invention is shown. Figure 3 The meteorological normalization flowchart is shown in the meteorological normalization and fire point constraint particulate matter non-meteorological contribution identification method and system of the present invention. Figure 4 The flowchart shows the multidimensional constraint discrimination process for fire point of the particulate matter non-meteorological contribution identification method and system based on meteorological normalization and fire point constraint of this invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the following description is provided in conjunction with the appendix. Figures 1-4 The present invention will be further described in detail with reference to specific embodiments.

[0023] Example 1: This embodiment provides a method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints. See also Figure 1 The method specifically includes the following steps: S1. Acquire particulate matter concentration data, meteorological data, fire point remote sensing data, and geospatial data from multiple cities, administrative regions, or monitoring stations within the target area; S2. Based on geospatial data, spatial matching and temporal alignment are performed on particulate matter concentration data, meteorological data and fire point remote sensing data to obtain spatiotemporal sample data corresponding to each city, administrative region or monitoring station in the target area. S3. Construct a particulate matter pollution feature set based on spatiotemporal sample data. The particulate matter pollution feature set includes meteorological features, temporal features, geospatial constraint features, and fire point exposure features. S4. Using particulate matter concentration as the target variable, establish a particulate matter concentration prediction model using a machine learning model. S5. While keeping the time characteristics and geospatial constraints unchanged, the meteorological characteristics are randomly resampled multiple times, and the resampled samples are input into the particulate matter concentration prediction model to obtain multiple sets of particulate matter predicted concentrations. S6. Calculate the meteorological normalized particulate matter concentration based on multiple sets of predicted particulate matter concentrations, and determine the meteorological contribution and non-meteorological contribution of particulate matter based on the difference between the observed particulate matter concentration and the meteorological normalized particulate matter concentration. S7. Construct multidimensional constraint discrimination results for geospatial fire points based on particulate matter non-meteorological contribution, fire point exposure characteristics, and geospatial constraint characteristics; S8. Based on the meteorological contribution of particulate matter, the non-meteorological contribution of particulate matter, and the multi-dimensional constraint discrimination results of geospatial fire points, output the anomaly identification results of particulate matter pollution and the attribution results of non-meteorological contributions.

[0024] Furthermore, particulate matter concentration data includes one or more of PM2.5, PM10, and PM2.5-10; meteorological data includes one or more of boundary layer height, temperature, humidity, dew point temperature, air pressure, wind speed, wind direction, precipitation, solar radiation, and cloud cover; fire point remote sensing data includes one or more of fire point latitude and longitude, fire point occurrence time, fire point radiation power, fire point confidence level, and satellite sensor identifier; geospatial data includes one or more of urban administrative boundaries, monitoring station latitude and longitude, urban center location, land cover type, farmland distribution, forest distribution, construction land distribution, road density, population density, nighttime light intensity, terrain height, and slope.

[0025] Furthermore, the geospatial constraint features include one or more of the following: spatial mean features within city boundaries, preset buffer zone features, inter-city distance features, distance attenuation weight features, upwind spatial relationship features, land cover ratio features, and human activity intensity features; the fire point exposure features include one or more of the following: number of fire points within the preset spatial range, fire point density, total fire point radiation power, mean fire point radiation power, distance-weighted fire point intensity, upwind fire point intensity, number of fire points on different land cover types, and lag fire point exposure intensity within different time windows.

[0026] Furthermore, the machine learning model includes one or more of the following: LightGBM model, XGBoost model, random forest model, gradient boosting tree model, support vector regression model, neural network model, and ensemble learning model; the input features of the machine learning model also include one or more of the following: periodic time encoding, long-term time variables, weekday variables, seasonal variables, and historical pollutant variation features; the machine learning model is used to establish the mapping relationship between meteorological features, time features, geospatial constraint features, fire point exposure features, and particulate matter concentration.

[0027] Furthermore, multiple random resampling of meteorological characteristics includes: extracting meteorological characteristics from historical meteorological samples of the target city, administrative region, or monitoring station according to month, season, pollution level, or preset time window to form multiple meteorological disturbance samples; inputting multiple meteorological disturbance samples into the particulate matter concentration prediction model; and performing mean, median, or quantile statistics on multiple prediction results to obtain the meteorological normalized particulate matter concentration.

[0028] Furthermore, the multidimensional constraint discrimination results of geospatial fire points are determined through fire point constraint discrimination indicators, which include one or more of the following: time synchronicity indicators, time lag indicators, spatial distance attenuation indicators, upwind consistency indicators, land cover type consistency indicators, and non-meteorological contribution anomaly intensity indicators. When the fire point constraint discrimination indicators meet the preset threshold conditions, the particulate matter pollution anomalies of the corresponding city, administrative region, or monitoring station in the corresponding time period are marked as fire point-related non-meteorological pollution anomalies.

[0029] Furthermore, the non-meteorological contribution attribution results include one or more of the following: non-meteorological contribution intensity at the city or regional scale, meteorological contribution intensity, fire point exposure contribution level, pollution anomaly level, anomaly occurrence time, ranking of major driving factors, and spatial distribution results.

[0030] Example 2: This embodiment provides a specific method for calculating fire point exposure characteristics and geospatial constraint characteristics. See also... Figure 4 After spatial matching and buffer analysis, the following features were calculated from the fire point remote sensing data, city boundary or monitoring station location, land cover data, and wind direction and distance information: Regarding the characteristics of fire point exposure, the number of fire points, fire point density, total fire point radiation power and average value within a preset spatial range are statistically analyzed; the distance-weighted fire point intensity is calculated based on the distance between the fire point and the monitoring station or city center; the upwind fire point intensity is calculated in combination with wind direction data; the number of fire points on different land surface types is statistically analyzed according to land cover type; and the fire point exposure intensity within different lag periods is calculated based on the time window settings.

[0031] Regarding geospatial constraints, the calculation includes the spatial mean of particulate matter concentration within city boundaries, concentration characteristics within a preset buffer zone, spatial distance characteristics between cities, spatial influence characteristics based on distance attenuation weights, upwind spatial relationship characteristics based on wind direction, area proportion characteristics of various types of land cover, and human activity intensity characteristics based on population density and nighttime light intensity.

[0032] The above characteristics together constitute the geospatial constraints and fire point exposure characteristics of particulate matter pollution feature set.

[0033] Example 3: This embodiment provides a specific method for determining the multi-dimensional constraint discrimination result of geospatial fire points. See also Figure 4 The multidimensional constraint judgment result of geospatial fire points is not determined solely by the number of fire points, but is determined by a combination of the following multidimensional indicators: The time synchronization index is used to determine whether the time of fire occurrence and the time of abnormal increase in particulate matter concentration are synchronized; the time lag index is used to determine whether an abnormal concentration response occurs within several hours after the fire occurs; the spatial distance attenuation index is used to determine whether the fire intensity shows a reasonable attenuation pattern with increasing distance; the upwind consistency index is used to determine whether the fire is located upwind of the monitoring station by combining wind direction data; the land cover type consistency index is used to determine whether the land cover type (such as farmland, forest land, grassland, etc.) where the fire is located is consistent with the expected type of open combustion; and the non-meteorological contribution anomaly intensity index is used to determine whether the non-meteorological contribution of particulate matter exceeds a preset threshold.

[0034] When the above-mentioned discrimination indicators meet their respective preset threshold conditions, the particulate matter pollution anomalies of the corresponding city, administrative region or monitoring station within the corresponding time period will be marked as fire-related non-meteorological pollution anomalies.

[0035] Example 4: This embodiment provides a specific calculation method for meteorological normalization. See also... Figure 3 In the target time sample, while maintaining the temporal and geospatial constraints, meteorological features are randomly resampled multiple times. Specifically, meteorological features are extracted from historical meteorological samples of the target city, administrative region, or monitoring station according to month, season, pollution level, or preset time window to form multiple meteorological disturbance samples.

[0036] Multiple meteorological disturbance samples are input into a pre-trained particulate matter concentration prediction model to obtain multiple sets of predicted particulate matter concentration results. The mean, median, or quantiles of these prediction results are then statistically calculated to obtain the meteorological normalized particulate matter concentration. This normalized concentration represents the baseline level of particulate matter concentration under current temporal characteristics and geographic spatial constraints, after eliminating the influence of random fluctuations in meteorological conditions.

[0037] The difference between the observed particulate matter concentration and the meteorological normalized particulate matter concentration is taken as the non-meteorological contribution of particulate matter. The larger the difference, the more significant the contribution of non-meteorological factors to the increase in particulate matter concentration, and vice versa, the meteorological contribution is dominant.

[0038] Example 5: See Figure 2 This embodiment provides a particulate matter non-meteorological contribution identification system based on meteorological normalization and fire point constraints. It includes: a data acquisition module, a spatiotemporal matching module, a feature construction module, a model training module, a meteorological normalization module, a contribution decomposition module, a fire point multidimensional constraint discrimination module, and a result output module.

[0039] The data acquisition module acquires particulate matter concentration data, meteorological data, fire point remote sensing data, and geospatial data from multiple cities, administrative regions, or monitoring stations within the target area. The spatiotemporal matching module performs spatial matching and temporal alignment of particulate matter concentration data, meteorological data, and fire point remote sensing data based on geospatial data, obtaining spatiotemporal sample data corresponding to each city, administrative region, or monitoring station within the target area. The feature construction module constructs a particulate matter pollution feature set based on the spatiotemporal sample data, including meteorological features, temporal features, geospatial constraint features, and fire point exposure features. The model training module uses particulate matter concentration as the target variable and a machine learning model to establish a particulate matter concentration prediction model. The meteorological normalization module maintains temporal and geospatial characteristics while adhering to the normalization principle. Under the condition that the bundle characteristics remain unchanged, meteorological characteristics are randomly resampled multiple times, and the resampled samples are input into the particulate matter concentration prediction model to obtain multiple sets of predicted particulate matter concentrations. The meteorological normalized particulate matter concentration is calculated based on the multiple sets of predicted particulate matter concentrations. The contribution decomposition module is used to determine the meteorological contribution and non-meteorological contribution of particulate matter based on the difference between the observed particulate matter concentration and the meteorological normalized particulate matter concentration. The multi-dimensional constraint discrimination module for fire points is used to construct the multi-dimensional constraint discrimination result of geospatial fire points based on the non-meteorological contribution of particulate matter, the exposure characteristics of fire points, and the geospatial constraint characteristics. The result output module is used to output the anomaly identification result of particulate matter pollution and the non-meteorological contribution attribution result based on the meteorological contribution of particulate matter, the non-meteorological contribution of particulate matter, and the multi-dimensional constraint discrimination result of geospatial fire points.

[0040] Furthermore, the particulate matter non-meteorological contribution identification system based on meteorological normalization and fire point constraints is functionally divided into an input data layer, a data processing layer, a model calculation layer, an attribution discrimination layer, and a result service layer.

[0041] The input data layer is used to access particulate matter concentration data, meteorological data, fire point remote sensing data, and geospatial data, supporting unified access and management of multi-source heterogeneous data. The data processing layer handles data access, quality control, spatiotemporal matching, and sample library management, ensuring alignment and consistency of various data types across time and space. The model calculation layer performs feature construction, model training, model invocation, and meteorological normalization calculation, enabling the establishment of particulate matter concentration prediction models and batch calculation of meteorological normalized concentrations. The attribution and discrimination layer performs contribution decomposition, fire point exposure calculation, geospatial constraint processing, and multi-dimensional constraint discrimination, achieving quantitative decomposition of meteorological and non-meteorological contributions to particulate matter and automatic identification of fire point-related pollution anomalies. The results service layer outputs pollution anomaly identification results, non-meteorological contribution attribution results, spatial distribution results, and reporting results, providing data support for environmental management and pollution prevention and control decisions.

[0042] In summary, this invention provides a method and system for identifying non-meteorological contributions to particulate matter based on meteorological normalization and multi-dimensional constraint discrimination of geospatial fire points, belonging to the field of atmospheric environmental monitoring and pollutant source attribution technology. The method first acquires particulate matter concentration, meteorological, fire point remote sensing, and geospatial data. After spatial matching and temporal alignment, a pollution feature set is constructed, including meteorological features, temporal features, geospatial constraint features, and fire point exposure features. Then, a concentration prediction model is established using a machine learning model. Meteorological normalization is achieved through multiple random resampling of meteorological features, thereby decomposing the meteorological and non-meteorological contributions of particulate matter. Finally, multi-dimensional constraint discrimination of fire points is performed by integrating non-meteorological contributions, fire point exposure, and geospatial constraints, outputting pollution anomaly identification and non-meteorological contribution attribution results. This invention can effectively eliminate meteorological interference and, combined with multi-dimensional constraints such as temporal synchronicity, spatial distance attenuation, upwind consistency, and land cover type, achieve accurate identification of fire point-related pollution anomalies, providing reliable technical support for regional particulate matter pollution diagnosis and pollution control decision-making.

[0043] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints, characterized in that, Includes the following steps: S1. Acquire particulate matter concentration data, meteorological data, fire point remote sensing data, and geospatial data from multiple cities, administrative regions, or monitoring stations within the target area; S2. Based on geospatial data, spatial matching and temporal alignment are performed on particulate matter concentration data, meteorological data and fire point remote sensing data to obtain spatiotemporal sample data corresponding to each city, administrative region or monitoring station in the target area. S3. Construct a particulate matter pollution feature set based on spatiotemporal sample data. The particulate matter pollution feature set includes meteorological features, temporal features, geospatial constraint features, and fire point exposure features. S4. Using particulate matter concentration as the target variable, establish a particulate matter concentration prediction model using a machine learning model. S5. While keeping the time characteristics and geospatial constraints unchanged, the meteorological characteristics are randomly resampled multiple times, and the resampled samples are input into the particulate matter concentration prediction model to obtain multiple sets of particulate matter predicted concentrations. S6. Calculate the meteorological normalized particulate matter concentration based on multiple sets of predicted particulate matter concentrations, and determine the meteorological contribution and non-meteorological contribution of particulate matter based on the difference between the observed particulate matter concentration and the meteorological normalized particulate matter concentration. S7. Construct multidimensional constraint discrimination results for geospatial fire points based on particulate matter non-meteorological contributions, fire point exposure characteristics, and geospatial constraint characteristics; S8. Based on the meteorological contribution of particulate matter, the non-meteorological contribution of particulate matter, and the multi-dimensional constraint discrimination results of geospatial fire points, output the anomaly identification results of particulate matter pollution and the attribution results of non-meteorological contributions.

2. The method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints according to claim 1, characterized in that, The particulate matter concentration data in S1 includes one or more of PM2.5, PM10, and PM2.5-10; the meteorological data includes one or more of boundary layer height, temperature, humidity, dew point temperature, air pressure, wind speed, wind direction, precipitation, solar radiation, and cloud cover. Fire point remote sensing data includes one or more of the following: fire point latitude and longitude, fire point occurrence time, fire point radiation power, fire point confidence level, and satellite sensor identifier; geospatial data includes one or more of the following: urban administrative boundaries, monitoring station latitude and longitude, urban center location, land cover type, farmland distribution, forest distribution, construction land distribution, road density, population density, nighttime light intensity, terrain height, and slope.

3. The method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints according to claim 1, characterized in that, The geospatial constraint features in S3 include one or more of the following: spatial mean features within city boundaries, preset buffer zone features, inter-city distance features, distance attenuation weight features, upwind spatial relationship features, land cover ratio features, and human activity intensity features; the fire point exposure features include one or more of the following: number of fire points within a preset spatial range, fire point density, total fire point radiation power, mean fire point radiation power, distance-weighted fire point intensity, upwind fire point intensity, number of fire points on different land cover types, and lag fire point exposure intensity within different time windows.

4. The method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints according to claim 1, characterized in that, The machine learning models in S4 include one or more of the following: LightGBM model, XGBoost model, random forest model, gradient boosting tree model, support vector regression model, neural network model, and ensemble learning model; the input features of the machine learning models also include one or more of the following: periodic time encoding, long-term time variables, weekday variables, seasonal variables, and historical pollutant variation features; the machine learning models are used to establish the mapping relationship between meteorological features, time features, geospatial constraint features, fire point exposure features, and particulate matter concentration.

5. The method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints according to claim 1, characterized in that, In S5, multiple random resampling of meteorological features includes: extracting meteorological features from historical meteorological samples of the target city, administrative region, or monitoring station according to month, season, pollution level, or preset time window to form multiple meteorological disturbance samples; inputting multiple meteorological disturbance samples into the particulate matter concentration prediction model; and performing mean, median, or quantile statistics on multiple prediction results to obtain the meteorological normalized particulate matter concentration.

6. The method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints according to claim 1, characterized in that, In S7, the multidimensional constraint discrimination results of geospatial fire points are determined by fire point constraint discrimination indicators, which include one or more of the following: time synchronicity indicators, time lag indicators, spatial distance attenuation indicators, upwind consistency indicators, land cover type consistency indicators, and non-meteorological contribution anomaly intensity indicators. When the fire point constraint discrimination indicators meet the preset threshold conditions, the particulate matter pollution anomalies of the corresponding city, administrative region, or monitoring station in the corresponding time period are marked as fire point-related non-meteorological pollution anomalies.

7. The method for identifying non-meteorological contributions of particulate matter based on meteorological normalization and fire point constraints according to claim 6, characterized in that, The non-meteorological contribution attribution results in S8 include one or more of the following: non-meteorological contribution intensity at the city or regional scale, meteorological contribution intensity, fire point exposure contribution level, pollution anomaly level, anomaly occurrence time, ranking of major driving factors, and spatial distribution results.

8. A particulate matter non-meteorological contribution identification system based on meteorological normalization and fire point constraints, characterized in that, A method for identifying non-meteorological contributions of particulate matter as described in any one of claims 1-7, comprising: The input data layer includes a data acquisition module, which is used to acquire particulate matter concentration data, meteorological data, fire point remote sensing data, and geospatial data from multiple cities, administrative regions, or monitoring stations within the target area. The data processing layer includes a spatiotemporal matching module. The spatiotemporal matching module receives the output data of the data acquisition module and performs spatial matching and temporal alignment on the particulate matter concentration data, meteorological data and fire point remote sensing data based on the geospatial data to obtain spatiotemporal sample data corresponding to each city, administrative region or monitoring station in the target area. The model calculation layer includes a feature construction module, a model training module, and a meteorological normalization module. The feature construction module receives the output data from the spatiotemporal matching module and constructs a particulate matter pollution feature set based on the spatiotemporal sample data. The particulate matter pollution feature set includes meteorological features, temporal features, geospatial constraint features, and fire point exposure features. The model training module receives the output data from the feature construction module and uses a machine learning model to establish a particulate matter concentration prediction model with particulate matter concentration as the target variable. The meteorological normalization module receives the output data from the model training module and, while keeping the temporal and geospatial constraint features unchanged, performs multiple random resampling of the meteorological features. The resampled samples are then input into the particulate matter concentration prediction model to obtain multiple sets of predicted particulate matter concentrations. The meteorological normalized particulate matter concentration is then calculated based on these multiple sets of predicted particulate matter concentrations. The attribution discrimination layer includes a contribution decomposition module and a fire point multidimensional constraint discrimination module. The contribution decomposition module receives the output data from the meteorological normalization module and determines the meteorological contribution and non-meteorological contribution of particulate matter based on the difference between the observed particulate matter concentration and the meteorological normalized particulate matter concentration. The fire point multidimensional constraint discrimination module receives the output data from the contribution decomposition module and constructs a geospatial fire point multidimensional constraint discrimination result based on the non-meteorological contribution of particulate matter, fire point exposure characteristics, and geospatial constraint characteristics. The result service layer includes a result output module. The result output module receives the output data of the fire point multidimensional constraint discrimination module and outputs particulate matter pollution anomaly identification results and non-meteorological contribution attribution results based on the particulate matter meteorological contribution, particulate matter non-meteorological contribution and geospatial fire point multidimensional constraint discrimination results.

9. An electronic device for performing the particulate matter non-meteorological contribution identification method based on meteorological normalization and fire point constraints as described in any one of claims 1-7, characterized in that, This includes a processor that performs the above methods, a memory, and a computer program stored in the memory and executable by the processor.

Images