Regional heavy pollution discriminating and forecasting system and method based on big data

Abstract

The invention discloses a regional heavy pollution discriminating and forecasting system and method based on big data, and relates to the technical field of environmental pollution monitoring. Comprising a multi-source data acquisition module, an adaptive weight distribution module, a feature extraction module, a heavy pollution discrimination module, a pollution diffusion path simulation module, an intelligent early warning and protection suggestion module, an event-driven dynamic monitoring module and a multi-scale meteorological prediction optimization module. And the pollution diffusion path simulation module performs prediction analysis on a pollution diffusion path and an influence range by using a diffusion path simulation algorithm. By designing the pollution diffusion path simulation module and the intelligent early warning and protection suggestion module, the pollutant diffusion path prediction and early warning functions are realized, and the problem that early warning lags due to the fact that the pollution diffusion path simulation and early warning module is separated and cannot realize timely linkage of prediction results and protection suggestions is solved; and a more timely and effective prevention and control strategy is provided, and the capability of coping with heavy pollution events is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental pollution monitoring, and in particular to a regional severe pollution identification and forecasting system and method based on big data. Background Art

[0002] In recent years, with the acceleration of urbanization and industrialization, air pollution has become increasingly serious. In particular, frequent regional severe pollution incidents have brought great challenges to public health, ecological environment and social economy. The generation of air pollution is affected by many factors, including emissions from pollution sources, as well as complex meteorological conditions, topography, traffic flow and other dynamic factors. The existing pollution monitoring and early warning system mainly relies on real-time data from fixed monitoring points. It is usually difficult to accurately predict the diffusion path and development trend of pollutants in real time, and it is impossible to respond to sudden changes in pollution in a timely manner.

[0003] To address this problem, a regional severe pollution identification and forecasting system that can integrate multiple data sources, dynamically allocate weights, and respond quickly is urgently needed. The system needs to be able to not only integrate multi-source data and achieve dynamic monitoring, but also provide high-precision pollution warnings through meteorological forecasts and pollution diffusion path simulations, and provide effective protection recommendations for the public and decision-making agencies.

[0004] 1. Patent document CN107292417B discloses a method and device for distinguishing and predicting regional heavy pollution based on a heavy pollution sequence case library. The above patent can improve the accuracy of predicting whether the target area will have heavy pollution in the future selected period, but the above patent cannot realize the diffusion path prediction and early warning function of pollutants.

[0005] 2. Patent document CN105738974B discloses a method and system for forecasting heavy air pollution weather. The above patent realizes the first combination of weather classification and meteorological element discrimination equation to discriminate heavy pollution processes that may occur in the future; the number of representative meteorological factors selected by the established discrimination equation is small, and the discrimination effect of heavy pollution cases is good; it is an effective supplement to the statistical forecasting technology and numerical forecasting technology of heavy pollution, the method is simple and easy, and the cost investment is small, which can effectively provide technical support for heavy pollution warning; a relatively simple and practical operation interface has been established, but the above patent cannot realize the dynamic discrimination and adjustment function of pollution.

[0006] 3. Patent document CN116381719B discloses a method for enhanced identification of polluted areas based on laser radar. The above patent avoids solving the extinction coefficient, and adjusting the iterative discrimination threshold can change the sensitivity of pollution identification, which can improve the efficiency and accuracy of identifying pollutants from laser radar detection data. In the case where the uneven disturbance of the horizontal atmosphere caused by air pollution is easily submerged in the background uniform atmosphere of a larger magnitude, it solves the problem of insufficient accuracy and low efficiency of traditional pollution identification methods, but the above patent cannot realize the function of dynamically adjusting the meteorological forecast results.

[0007] 4. Patent document CN114510823B discloses a method for rapidly assessing the degree of heavy metal pollution in soil. The above patent avoids the traditional way of thinking of assessing the "precise" content of heavy metals in soil and instead assesses the relative degree of heavy metal pollution in soil. The assessment efficiency is higher than that of traditional precise prediction methods, the assessment results are reliable, and the assessment accuracy can meet the needs of macro-decision-making. It can provide technical support for management departments to make macro-decisions on regional soil pollution prevention and control. However, the above patent cannot realize the analysis and processing function of multi-source data.

[0008] In summary, the above patents cannot realize the functions of predicting and warning the diffusion path of pollutants, dynamically distinguishing and adjusting pollution, dynamically adjusting meteorological forecast results, and analyzing and processing multi-source data, resulting in delayed data updates, difficulty in accurately predicting the pollution diffusion path and development trend in real time, and failure to respond in time to the problem of pollution not changing;

[0009] To this end, the present application proposes a regional heavy pollution identification and forecasting system and method based on big data, which can realize the functions of predicting and warning the diffusion path of pollutants, dynamically identifying and adjusting pollution, dynamically adjusting meteorological forecast results, and analyzing and processing multi-source data. Summary of the invention

[0010] The purpose of the present invention is to provide a regional heavy pollution identification and forecasting system and method based on big data, so as to solve the technical problems proposed in the above background technology, such as the inability to realize the prediction and early warning functions of the diffusion path of pollutants, the dynamic identification and adjustment function of pollution, the dynamic adjustment function of meteorological forecast results and the analysis and processing function of multi-source data, resulting in delayed data update, difficulty in real-time and accurate prediction of pollution diffusion paths and development trends, and inability to respond in time to the technical problems that pollution cannot change.

[0011] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a regional heavy pollution discrimination and forecasting system based on big data, comprising a multi-source data acquisition module, an adaptive weight allocation module, a feature extraction module, a heavy pollution discrimination module, a pollution diffusion path simulation module, an intelligent early warning and protection suggestion module, an event-driven dynamic monitoring module and a multi-scale meteorological forecast optimization module, wherein the heavy pollution discrimination module is connected to the pollution diffusion path simulation module through a data interface, the heavy pollution discrimination module provides the pollution source location and meteorological data, and the pollution diffusion path simulation module uses a diffusion path simulation algorithm to perform a prediction and analysis of the pollution diffusion path and the impact range;

[0012] The pollution diffusion path simulation module provides diffusion path simulation algorithm, pollution source tracing function and three-dimensional diffusion map generation function;

[0013] The diffusion path simulation algorithm simulates the pollutant diffusion path based on wind speed, wind direction, temperature and humidity meteorological data using a fluid dynamics model / Lagrangian diffusion model;

[0014] The pollution source tracing function locates possible pollution sources based on the pollution diffusion model and the reverse calculation algorithm of pollution concentration, and provides real-time feedback on tracing information;

[0015] The three-dimensional diffusion map generation function generates a three-dimensional pollution diffusion map based on GIS technology, showing the diffusion range, concentration changes and affected areas of pollutants in space;

[0016] The diffusion map and pollution range data generated by the pollution diffusion path module are transmitted to the intelligent early warning and protection suggestion module through the data interface, providing technical data support for personalized protection suggestions.

[0017] Preferably, the heavy pollution discrimination module receives the high-dimensional feature data provided by the feature extraction module through a data interface, and performs heavy pollution discrimination analysis based on a deep learning model;

[0018] The heavy pollution identification module provides a deep learning identification model, pollution level output mechanism and adaptive identification tuning mechanism;

[0019] The deep learning discrimination model uses a combination of convolutional neural network and long short-term memory network models to perform multi-layer analysis on real-time feature data to identify pollution status and determine pollution levels;

[0020] The pollution level output mechanism divides the pollution discrimination results output by the deep learning discrimination model into light, moderate and heavy pollution levels, and generates corresponding warning signals;

[0021] The adaptive discriminant tuning mechanism dynamically adjusts the parameters of the deep learning discriminant model according to the error between the predicted results and the actual data, thus improving the adaptability of the discriminant model.

[0022] The judgment results output by the heavy pollution judgment module are transmitted to the intelligent early warning and protection suggestion module through the data interface for generating early warnings, and are also transmitted to the event-driven dynamic monitoring module to trigger specific event response measures.

[0023] Preferably, the feature extraction module obtains pollutant concentration, meteorological data and multi-dimensional feature data of traffic flow extracted from weighted data from the adaptive weight allocation module through a data interface;

[0024] The feature extraction module provides feature extraction algorithms and high-dimensional feature mapping functions;

[0025] The feature extraction algorithm is based on the multidimensional matrix construction method to extract and construct the multidimensional feature matrix of pollutants;

[0026] The high-dimensional feature mapping function uses machine learning algorithms to map pollution features into high-dimensional space to capture complex pollution causes and transmission characteristics;

[0027] The feature extraction module sends the high-latitude feature data to the heavy pollution identification module and the pollution diffusion path simulation module through the data interface for pollution level judgment and calculation of pollution diffusion simulation input.

[0028] Preferably, the adaptive weight allocation module adjusts the weight of each data through a dynamic weight allocation algorithm based on the multi-source data transmitted by the multi-source data acquisition module through the data interface, so as to adapt to different pollution scenarios;

[0029] The dynamic weight allocation module provides weight allocation algorithm and weight optimization feedback mechanism;

[0030] The dynamic weight allocation algorithm automatically adjusts the weight of data sources in major pollution incidents and daily monitoring situations through a weight optimization mechanism based on historical data, making data fusion scenario-adaptive.

[0031] The weight optimization feedback mechanism is to pass the real-time feedback data after the pollution event to the weight optimization mechanism to improve the dynamic response of weight allocation;

[0032] The adaptive weight distribution module transmits the weighted fusion data to the feature extraction module and the heavy pollution discrimination module, providing accurate data input for feature extraction and pollution discrimination.

[0033] Preferably, the multi-source data acquisition module is used to collect real-time data from multiple data sources such as air quality monitoring stations, meteorological stations, traffic flow monitoring equipment and satellite remote sensing systems, and convert the data into a unified format through standardized processing, cleaning and normalization methods to facilitate analysis and processing by subsequent modules. The data generated by the multi-source data acquisition module is transmitted to the adaptive weight allocation module and the feature extraction module through the data interface to realize weighted fusion and feature analysis of the data.

[0034] Preferably, the event-driven dynamic monitoring module automatically triggers the event-driven mechanism based on the warning signal of the intelligent warning and protection suggestion module;

[0035] The event-driven dynamic monitoring module provides event trigger condition setting, real-time dynamic monitoring and model automatic correction functions;

[0036] Event trigger condition setting Set event trigger rules for traffic peak, industrial emissions and weather deterioration conditions, and start real-time dynamic monitoring when the trigger conditions are met;

[0037] Real-time dynamic monitoring means that once the event-driven mechanism is triggered, the system will monitor pollutant changes in a specific area and feed back real-time data to the model's adaptive tuning mechanism;

[0038] The automatic model correction function dynamically adjusts the parameters of the heavy pollution identification model and the pollution diffusion path simulation module based on monitoring data to improve the accuracy of the response to emergencies;

[0039] The real-time monitoring data of the event-driven dynamic monitoring module is transmitted to the multi-scale meteorological forecast optimization module through the data interface to optimize the input parameters of pollution diffusion prediction.

[0040] Preferably, the multi-scale meteorological forecast optimization module realizes dynamic optimization prediction of pollution diffusion based on real-time monitoring data and multi-level meteorological data provided by the event-driven dynamic monitoring module;

[0041] The multi-scale meteorological forecast optimization module provides multi-scale meteorological data fusion, fine particle diffusion model optimization algorithm and pollution trend map generation functions;

[0042] Multi-scale meteorological data fusion is the fusion of local, regional and national meteorological data to provide forecast accuracy for meteorology at different scales;

[0043] The fine particulate matter diffusion model optimization algorithm combines meteorological data and particulate matter characteristics (particle size and composition) to dynamically predict the diffusion rate, concentration distribution and diffusion radius of pollutants;

[0044] The pollution trend map generation function generates pollution diffusion trend maps for multiple time periods, showing the pollution concentration distribution and impact range at different time points in the future;

[0045] The pollution trend data of the multi-scale meteorological forecast optimization module is transmitted to the intelligent early warning and protection recommendation module through the data interface, providing future pollution change forecasts and providing management departments with predictions and decision support for prevention and control strategies.

[0046] Preferably, the intelligent early warning and protection suggestion module is used to generate graded early warning information and personalized protection suggestions based on the structure of the heavy pollution identification module and the pollution diffusion path simulation module. The output data of the intelligent early warning and protection suggestion module is transmitted to the event-driven dynamic monitoring module through the data interface as a trigger condition to execute real-time monitoring and early warning response measures.

[0047] Preferably, the discrimination prediction method comprises the following steps:

[0048] S1, obtain real-time air quality monitoring data, meteorological data, traffic flow data and satellite remote sensing data from the multi-source data acquisition module, and pre-process the data;

[0049] S2. In the feature extraction module, multi-dimensional features of pollutant concentration, meteorological factors and traffic flow are extracted from the comprehensive data stream, and the data is mapped to a multi-dimensional space through a high-dimensional feature construction algorithm to generate a feature matrix;

[0050] S3, input the feature matrix into the heavy pollution discrimination module, and analyze the data through the combination model of convolutional neural network (CNN) and long short-term memory network (LSTM). The model discriminates the pollution level based on the input high-dimensional feature data, judges the pollution status in the current area, and classifies the results into light, moderate and heavy pollution;

[0051] S4. In the pollution diffusion path simulation module, based on meteorological data and the location of pollution sources, the pollution diffusion path simulation algorithm is used to predict the pollutant diffusion trend and generate a three-dimensional pollution diffusion map.

[0052] Preferably, the discrimination prediction method further comprises the following steps:

[0053] S11. All kinds of data are weighted and integrated through the adaptive weight distribution module, and the weights of different data sources are dynamically adjusted according to the pattern of historical pollution events to generate a comprehensive data stream, providing high-quality data input for subsequent identification and prediction;

[0054] S21, the feature matrix extracts key features related to heavy pollution events through a multi-layer feature extraction algorithm;

[0055] S31. If a severe pollution state is detected, the corresponding warning signal is triggered, and the system automatically optimizes the parameters of each model based on the feedback of historical and real-time data to improve the adaptability to sudden severe pollution events;

[0056] S41. Combined with the multi-scale meteorological data fusion and optimization module, the system makes short-term and long-term predictions on pollution diffusion trends and generates pollution trend graphs for multiple time periods. At the same time, the intelligent early warning and protection suggestion module provides early warning information to the public based on the diffusion prediction results and pushes personalized protection suggestions.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] 1. The present invention realizes the prediction and warning functions of the diffusion path of pollutants by designing a pollution diffusion path simulation module and an intelligent early warning and protection suggestion module, solves the problem of the separation of the pollution diffusion path simulation and the early warning module, which makes it impossible to realize the timely linkage between the prediction results and the protection suggestions, resulting in the problem of delayed early warning, provides a more timely and effective prevention and control strategy, and enhances the ability to respond to serious pollution events;

[0059] 2. The present invention realizes the dynamic discrimination and adjustment function of pollution by designing a dynamic weight allocation module and a heavy pollution discrimination module, solves the problem that the traditional pollution monitoring system cannot respond to sudden changes in real time and cause data updates, improves the monitoring accuracy and response speed of pollution events, and thus improves the timeliness and adaptability of the overall system;

[0060] 3. The present invention realizes the function of dynamically adjusting the meteorological forecast results by designing a multi-scale meteorological forecast optimization module and an event-driven dynamic monitoring module, which solves the problem that the existing meteorological forecast model lacks multi-scale compatibility in time and space, resulting in delayed response to short-term meteorological changes, improves the prediction accuracy of the system at different time and space scales, helps to accurately simulate the diffusion path of pollutants, and enhances the early warning capability of pollution events;

[0061] 4. The present invention realizes the analysis and processing function of multi-source data by designing a multi-source data acquisition module and a feature extraction module, solves the problem of fixed data source and difficulty in extracting key information, improves the sensitivity of pollution event identification and data reliability, and thus improves the accuracy of pollution monitoring and prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the pollution identification and prediction process of the present invention; DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] See also Figure 1 , an embodiment provided by the present invention: a regional heavy pollution discrimination and forecasting system based on big data, comprising a multi-source data acquisition module, an adaptive weight allocation module, a feature extraction module, a heavy pollution discrimination module, a pollution diffusion path simulation module, an intelligent early warning and protection suggestion module, an event-driven dynamic monitoring module and a multi-scale meteorological forecast optimization module, the heavy pollution discrimination module is connected to the pollution diffusion path simulation module through a data interface, the heavy pollution discrimination module provides the pollution source location and meteorological data, and the pollution diffusion path simulation module uses a diffusion path simulation algorithm to perform prediction and analysis of the pollution diffusion path and the impact range;

[0065] The pollution diffusion path simulation module provides diffusion path simulation algorithm, pollution source tracing function and three-dimensional diffusion map generation function;

[0066] The diffusion path simulation algorithm simulates the pollutant diffusion path based on wind speed, wind direction, temperature and humidity meteorological data using a fluid dynamics model / Lagrangian diffusion model;

[0067] The pollution source tracing function locates possible pollution sources based on the pollution diffusion model and the reverse calculation algorithm of pollution concentration, and provides real-time feedback on tracing information;

[0068] The three-dimensional diffusion map generation function generates a three-dimensional pollution diffusion map based on GIS technology, showing the diffusion range, concentration changes and affected areas of pollutants in space;

[0069] The diffusion map and pollution range data generated by the pollution diffusion path module are transmitted to the intelligent early warning and protection suggestion module through the data interface to provide technical data support for personalized protection suggestions;

[0070] The intelligent early warning and protection suggestion module is used to generate graded early warning information and personalized protection suggestions based on the structure of the heavy pollution identification module and the pollution diffusion path simulation module. The output data of the intelligent early warning and protection suggestion module is transmitted to the event-driven dynamic monitoring module through the data interface as a trigger condition to execute real-time monitoring and early warning response measures;

[0071] Furthermore, after receiving the heavy pollution discrimination signal, the pollution diffusion path simulation module performs a preliminary calculation of the diffusion path through the Lagrangian diffusion model and meteorological data in the system. At this time, the module obtains the real-time meteorological data of the area, including customs, wind direction, humidity and temperature. The system combines the initial concentration data of pollutants with the current meteorological conditions to calculate the real-time path of pollutant diffusion. If the system detects an obvious pollution source, it traces the pollution source based on the reverse calculation algorithm to determine the location of the main pollution source. Using GIS technology, the system generates a three-dimensional model of pollutant diffusion in the area, showing the diffusion of pollutants at different time points such as 1 hour, 3 hours and 12 hours in the future, covering the regional distribution of high pollution concentration areas, low concentration impact areas and boundary areas, forming a dynamic pollution impact range map.

[0072] The system continuously collects real-time meteorological data and dynamically adjusts the diffusion path of pollutants. According to changes in current meteorological conditions, such as sudden changes in wind direction or increased airflow speed, the module updates the diffusion simulation path in real time and generates a new round of pollution diffusion forecast maps, so as to facilitate timely adjustments to the content of protection recommendations;

[0073] Based on the forecast data provided by the pollution diffusion simulation module, the intelligent early warning and protection suggestion module divides the early warning fans in the area into three levels: "early warning notification", "warning" and "serious warning", corresponding to the different early warning needs of mild, moderate and severe pollution. The system pushes the early warning information through various means, including mobile applications, text messages, emails and social media, to ensure that different groups of people can obtain information in a timely manner. At the same time, the system will push corresponding emergency response suggestions (such as traffic flow restrictions, industrial production restrictions, etc.) to relevant management departments;

[0074] The system collects feedback data after the warning information is released, such as actual pollution concentration and public response. If there is a large deviation between the actual pollution level and the predicted result, the system records the error and transmits it back to the pollution diffusion path simulation module as a basis for adaptive optimization of model parameters; based on the feedback data, the system optimizes future pollution path simulation and warning effects by adjusting the diffusion model parameters and the discrimination threshold of the intelligent warning module, so as to improve the accuracy and response speed of model prediction in the next pollution event.

[0075] See also Figure 1 , an embodiment provided by the present invention: a regional heavy pollution discrimination and forecasting system based on big data, wherein the heavy pollution discrimination module receives high-dimensional feature data provided by the feature extraction module through a data interface, and performs heavy pollution discrimination analysis based on a deep learning model;

[0076] The heavy pollution identification module provides a deep learning identification model, pollution level output mechanism and adaptive identification tuning mechanism;

[0077] The deep learning discrimination model uses a combination of convolutional neural network and long short-term memory network models to perform multi-layer analysis on real-time feature data to identify pollution status and determine pollution levels;

[0078] The pollution level output mechanism divides the pollution discrimination results output by the deep learning discrimination model into light, moderate and heavy pollution levels, and generates corresponding warning signals;

[0079] The adaptive discriminant tuning mechanism dynamically adjusts the parameters of the deep learning discriminant model according to the error between the predicted results and the actual data, thus improving the adaptability of the discriminant model.

[0080] The discrimination results output by the heavy pollution discrimination module are transmitted to the intelligent early warning and protection suggestion module through the data interface to generate early warnings, and are also transmitted to the event-driven dynamic monitoring module to trigger specific event response measures;

[0081] The feature extraction module obtains pollutant concentration, meteorological data and multi-dimensional feature data of traffic flow extracted from weighted data from the adaptive weight allocation module through a data interface;

[0082] The feature extraction module provides feature extraction algorithms and high-dimensional feature mapping functions;

[0083] The feature extraction algorithm is based on the multidimensional matrix construction method to extract and construct the multidimensional feature matrix of pollutants;

[0084] The high-dimensional feature mapping function uses machine learning algorithms to map pollution features into high-dimensional space to capture complex pollution causes and transmission characteristics;

[0085] The feature extraction module sends the high-latitude feature data to the heavy pollution identification module and the pollution diffusion path simulation module through the data interface for pollution level judgment and calculation of pollution diffusion simulation input;

[0086] Furthermore, the feature extraction module constructs multi-dimensional features for the pre-processed data, and extracts features that are highly relevant to pollution events through the multi-source data fusion module, including: pollutant features: concentrations of different pollutants, concentration growth rates, and concentration peaks; meteorological features: such as customs, wind direction trends, and temperature and humidity; traffic flow features: such as peak hours, vehicle density, and average speed; spatial features: based on satellite remote sensing data, extract the geographical features of pollution sources and regions; integrate all feature data into a high-dimensional feature matrix to form a set of feature vectors that can describe the current pollution status, providing input for the deep learning discrimination algorithm;

[0087] The system inputs the feature matrix into the heavy pollution discrimination module and uses pre-trained deep learning models, such as a combination of convolutional neural networks and long short-term memory networks, to perform discriminant analysis. The model can automatically learn the pollution feature patterns in the high-dimensional feature matrix, and then identify the regional pollution level, and classify the input feature matrix into light, moderate or heavy pollution levels. For the situation where the initial detected pollutant concentration increases significantly, the model weightedly analyzes the pollutant concentration, meteorological conditions and traffic conditions to generate a preliminary judgment of the pollution level: Light pollution: The air quality has slightly decreased, but has not yet had a significant impact on human health; Moderate pollution: The air quality has reached a level that affects health, and vigilance is needed; Severe pollution: The pollutant concentration is high and may affect a wide range of people, and prevention and control measures must be taken immediately;

[0088] The heavy pollution discrimination module continuously monitors each feature data when the pollution is in a heavy pollution state, and makes judgments based on the updated iteration of real-time data, and feeds back the latest pollution level information to the feature extraction module to optimize the parameter configuration of feature extraction. If the heavy pollution discrimination model detects frequent changes in pollution levels, the system records the changes in feature data in the feature matrix and automatically adjusts the parameters of the convolution layer and memory network to optimize the adaptability of the model under different pollution conditions;

[0089] When the discrimination module determines that the current area is in a state of severe pollution, the results are transmitted to the early warning module and the pollution diffusion path simulation module, triggering the corresponding pollution early warning mechanism and simultaneously starting the diffusion path simulation analysis. The discrimination results are released to relevant departments and the public through the system, providing timely and accurate reference for subsequent pollution protection decisions.

[0090] See also Figure 1 , an embodiment provided by the present invention: a regional heavy pollution discrimination and forecasting system based on big data, wherein the adaptive weight allocation module adjusts the weight of each data through a dynamic weight allocation algorithm based on the multi-source data transmitted by the multi-source data acquisition module through the data interface, thereby adapting to different pollution scenarios;

[0091] The dynamic weight allocation module provides weight allocation algorithm and weight optimization feedback mechanism;

[0092] The dynamic weight allocation algorithm automatically adjusts the weight of data sources in major pollution incidents and daily monitoring situations through a weight optimization mechanism based on historical data, making data fusion scenario-adaptive.

[0093] The weight optimization feedback mechanism is to pass the real-time feedback data after the pollution event to the weight optimization mechanism to improve the dynamic response of weight allocation;

[0094] The adaptive weight distribution module transmits weighted fusion data to the feature extraction module and the heavy pollution identification module, providing accurate data input for feature extraction and pollution identification;

[0095] The multi-source data acquisition module is used to collect real-time data from multiple data sources such as air quality monitoring stations, weather stations, traffic flow monitoring equipment and satellite remote sensing systems, and convert the data into a unified format through standardized processing, cleaning and normalization methods to facilitate analysis and processing by subsequent modules. The data generated by the multi-source data acquisition module is transmitted to the adaptive weight allocation module and the feature extraction module through the data interface to achieve weighted fusion and feature analysis of the data;

[0096] Furthermore, the multi-source data acquisition module is connected to multiple data source interfaces, including: air quality monitoring data: such as PM2.5, PM10, sulfur dioxide and other pollutant concentrations, provided by fixed monitoring stations; meteorological data: including customs, wind direction, temperature and humidity, etc., collected in real time by meteorological stations; traffic data: obtaining information such as traffic flow and congestion index from the urban traffic monitoring platform; satellite remote sensing data: providing a wide range of pollution distribution image data, especially detailed information on highly polluted areas; the system sets the acquisition frequency according to the characteristics of the data source, such as air quality and meteorological data are updated every 15 minutes, traffic data are updated every 5 minutes, and satellite remote sensing data are acquired regularly according to the satellite transit frequency;

[0097] At the beginning of data collection, the system sets an initial weight for each data source. For example, air quality monitoring data and meteorological data usually have higher weights, while traffic data and satellite remote sensing data are supplemented with slightly lower initial weights. This initial weight is set based on the degree of influence of each data source on pollution discrimination in historical pollution events;

[0098] When the system detects a significant change in pollutant concentration, such as a sharp increase in PM2.5 or PM10, or when the system enters a specific high-pollution weather mode, such as calm wind and inversion weather conditions, the dynamic weight allocation module automatically triggers the weight adjustment process: if the collection of a data source is abnormal, such as data fluctuations or loss at an air quality monitoring station, the system reduces the weight of the data source; the system analyzes the characteristics of the pollution event, and if the current pollution event is greatly affected by meteorological conditions, the weight of meteorological data is appropriately increased; if it is detected that road traffic is the main factor of pollution, the weight of traffic data is increased; when the system dynamically allocates weights, it will also refer to historical data of similar pollution events and automatically adjust the weight allocation model to match the characteristics of the current event; the system uses a weight optimization algorithm to allocate and adjust each data source and calculate the contribution of each data source. For example, the weight of air quality data may be increased from 40% to 60%, meteorological data from 30% to 25%, traffic data from 10%, and satellite remote sensing data from 5%. The updated weights are applied to the data fusion process in real time;

[0099] The dynamic weight allocation module weights and fuses the real-time data of each data source according to the adjusted weights to generate a comprehensive data stream. Through the real-time update of the comprehensive data stream, the system inputs the obtained weighted data into the pollution trend modeling module to calculate the future pollution trend and diffusion range. Dynamic weight allocation ensures the rationality of the contribution of different data sources in pollution events and improves the prediction accuracy of the pollution trend model.

[0100] See also Figure 1 , an embodiment provided by the present invention: a regional heavy pollution identification and forecasting system based on big data, the intelligent early warning and protection suggestion module is used to generate graded early warning information and personalized protection suggestions based on the structure of the heavy pollution identification module and the pollution diffusion path simulation module, and the output data of the intelligent early warning and protection suggestion module is transmitted to the event-driven dynamic monitoring module through the data interface as a trigger condition to execute real-time monitoring and early warning response measures;

[0101] The event-driven dynamic monitoring module automatically triggers the event-driven mechanism based on the warning signal of the intelligent warning and protection suggestion module;

[0102] The event-driven dynamic monitoring module provides event trigger condition setting, real-time dynamic monitoring and model automatic correction functions;

[0103] Event trigger condition setting Set event trigger rules for traffic peak, industrial emissions and weather deterioration conditions, and start real-time dynamic monitoring when the trigger conditions are met;

[0104] Real-time dynamic monitoring means that once the event-driven mechanism is triggered, the system will monitor pollutant changes in a specific area and feed back real-time data to the model's adaptive tuning mechanism;

[0105] The automatic model correction function dynamically adjusts the parameters of the heavy pollution identification model and the pollution diffusion path simulation module based on monitoring data to improve the accuracy of the response to emergencies;

[0106] The real-time monitoring data of the event-driven dynamic monitoring module is transmitted to the multi-scale meteorological forecast optimization module through the data interface to optimize the input parameters of pollution diffusion prediction;

[0107] Furthermore, when the event-driven dynamic monitoring module is started, the system administrator pre-sets a set of trigger conditions, including but not limited to: pollutant concentration thresholds: when the PM2.5 or PM10 concentration reaches a certain value, event monitoring is triggered; meteorological conditions: specific meteorological conditions (such as calm winds, inversion layers, low pressure, etc.) will increase the risk of pollution, and the system will start dynamic monitoring when these meteorological conditions are detected; traffic rush hour: when the traffic data collection module detects a surge in traffic flow or congestion, the monitoring module is triggered to evaluate the impact of traffic on air quality; the system scans the collected multi-source data (pollutant concentration, meteorological data, traffic flow, etc.) in real time, and when any trigger condition is met, the module enters the event monitoring state. At this time, the module starts the dynamic monitoring process, collects relevant data in real time, and closely observes the progress of pollution events;

[0108] After the event is triggered, the dynamic monitoring module conducts multi-dimensional data analysis on the current pollution event. The module's built-in anomaly recognition algorithm will mark the event as "abnormal" when it detects a sharp increase in pollutant concentration, deteriorating weather conditions, or a continuous increase in traffic flow. If the abnormal marking condition is triggered, an alarm signal will be immediately sent to other related modules (such as the early warning module); during the initial triggering of the event and the abnormal identification stage, the module will adjust the monitoring frequency from the normal mode to the high-frequency mode. In high-frequency mode, the frequency of data collection is accelerated to ensure rapid capture of changes in pollution events: Normal mode: collect data every 15 minutes / 30 minutes; High-frequency mode: increase the collection frequency to once every 5 minutes, or even once every 1 minute, and the specific frequency is determined according to the severity of the pollution; the event-driven dynamic monitoring module provides various data and trend change reports to the dynamic weight allocation module, the pollution diffusion path simulation module, and the intelligent early warning module, so that these modules can be adjusted and optimized in real time according to the latest monitoring data.

[0109] See also Figure 1 , an embodiment provided by the present invention: a regional heavy pollution discrimination and forecasting system based on big data, wherein the multi-scale meteorological forecasting optimization module realizes dynamic optimization prediction of pollution diffusion based on real-time monitoring data and multi-level meteorological data provided by an event-driven dynamic monitoring module;

[0110] The multi-scale meteorological forecast optimization module provides multi-scale meteorological data fusion, fine particle diffusion model optimization algorithm and pollution trend map generation functions;

[0111] Multi-scale meteorological data fusion is the fusion of local, regional and national meteorological data to provide forecast accuracy for meteorology at different scales;

[0112] The fine particulate matter diffusion model optimization algorithm combines meteorological data and particulate matter characteristics (particle size and composition) to dynamically predict the diffusion rate, concentration distribution and diffusion radius of pollutants;

[0113] The pollution trend map generation function generates pollution diffusion trend maps for multiple time periods, showing the pollution concentration distribution and impact range at different time points in the future;

[0114] The pollution trend data of the multi-scale meteorological forecast optimization module is transmitted to the intelligent early warning and protection suggestion module through the data interface, providing future pollution change forecasts and providing management departments with prediction and decision support for prevention and control strategies;

[0115] Furthermore, the multi-scale meteorological forecast optimization module obtains meteorological data of various scales through the multi-source data acquisition module, including: local real-time meteorological data: conventional meteorological data obtained from ground meteorological stations; medium-scale meteorological data: regional meteorological condition information obtained from meteorological satellites and radar images, covering an area of ​​50-500 kilometers; large-scale meteorological data: long-term meteorological forecast data obtained from the global meteorological model or meteorological database GFS;

[0116] Construct multi-scale meteorological forecast models: Short-term model from the ground: Use local real-time meteorological data and combine machine learning models (such as support vector machines and decision trees) to construct a short-term meteorological change prediction model for the pollution site, and predict local meteorological conditions such as wind speed and wind direction in the next 1 to 3 hours; Regional mesoscale model: Based on mesoscale meteorological data, use deep neural networks or convolutional neural networks (CNN) to predict meteorological conditions within a range of 50-500 kilometers, and predict wind field changes, temperature distribution, and humidity gradients in the next 3 to 24 hours; Global large-scale model: Combine global meteorological data to build a large-scale meteorological forecast model, and use long-term forecasting algorithms (such as long short-term memory networks LSTM) to predict weather patterns in the next 1 to 3 days, including the formation and movement of large cyclones and inversion layers; The module uses a dynamic weighting algorithm to adjust the weights of models of different scales according to the needs of the forecast time and regional scope. For example, in the early stages of a pollution outbreak, the weight of the local short-term model is higher to ensure the accuracy of predictions in a short period of time; when the pollution event develops for more than 24 hours, the weights of the mesoscale and large-scale models are increased;

[0117] The module generates meteorological forecast data for multiple time periods based on the forecast results of models of different scales, including forecast results of meteorological elements such as wind speed, wind direction, temperature and humidity for the next 1 hour, 3 hours, 12 hours, 24 hours and 72 hours; during the monitoring of pollution events, the module updates data and re-forecasts in units of 5 to 15 minutes to ensure that the system is sensitive to meteorological changes. For example, when the short-term model detects that the wind direction may change dramatically, the multi-scale meteorological forecast optimization module immediately recalculates the wind field forecast results and feeds back to other modules; if the multi-scale meteorological forecast optimization module identifies abnormal weather phenomena (such as strong inversion layer or extreme windy weather) during the forecast process, it will immediately trigger the early warning mechanism and notify other related modules through the system to adjust the diffusion path simulation or early warning strategy.

[0118] Working principle: First, the multi-source data acquisition module is used to obtain various types of raw data in real time, including air quality monitoring data, meteorological data, traffic flow information and satellite remote sensing data, and pre-process the data;

[0119] Then, the preprocessed data enters the dynamic weight allocation module and the multi-scale meteorological forecast optimization module. The dynamic weight allocation module assigns different weights to different data sources according to the real-time pollution status, historical pollution characteristics and meteorological conditions; the multi-scale meteorological forecast module generates future meteorological forecast data through a variety of meteorological models and dynamically adjusts weights and forecast parameters;

[0120] Finally, the system inputs the fused data into the heavy pollution identification module and the pollution diffusion path simulation module to identify high-risk areas and diffusion trends; the event-driven dynamic monitoring module monitors pollution and meteorological changes in real time during pollution events, and issues warnings through the intelligent early warning and protection recommendation module, and recommends corresponding protective measures.

[0121] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A regional heavy pollution identification and forecasting system based on big data, including a multi-source data acquisition module, an adaptive weight allocation module, a feature extraction module, a heavy pollution identification module, a pollution diffusion path simulation module, an intelligent early warning and protection suggestion module, an event-driven dynamic monitoring module and a multi-scale meteorological forecast optimization module, characterized in that: The heavy pollution identification module is connected to the pollution diffusion path simulation module through a data interface. The heavy pollution identification module provides the pollution source location and meteorological data. The pollution diffusion path simulation module uses a diffusion path simulation algorithm to predict and analyze the pollution diffusion path and impact range. The pollution diffusion path simulation module provides diffusion path simulation algorithm, pollution source tracing function and three-dimensional diffusion map generation function; The diffusion path simulation algorithm simulates the pollutant diffusion path based on wind speed, wind direction, temperature and humidity meteorological data using a fluid dynamics model / Lagrangian diffusion model; The pollution source tracing function locates possible pollution sources based on the pollution diffusion model and the reverse calculation algorithm of pollution concentration, and provides real-time feedback on tracing information; The three-dimensional diffusion map generation function generates a three-dimensional pollution diffusion map based on GIS technology, showing the diffusion range, concentration changes and affected areas of pollutants in space; The diffusion map and pollution range data generated by the pollution diffusion path module are transmitted to the intelligent early warning and protection suggestion module through the data interface, providing technical data support for personalized protection suggestions.

2. According to the big data-based regional heavy pollution identification and forecasting system according to claim 1, it is characterized by: The heavy pollution discrimination module receives the high-dimensional feature data provided by the feature extraction module through the data interface, and performs heavy pollution discrimination analysis based on the deep learning model; The heavy pollution identification module provides a deep learning identification model, pollution level output mechanism and adaptive identification tuning mechanism; The deep learning discrimination model uses a combination of convolutional neural network and long short-term memory network models to perform multi-layer analysis on real-time feature data to identify pollution status and determine pollution levels; The pollution level output mechanism divides the pollution discrimination results output by the deep learning discrimination model into light, moderate and heavy pollution levels, and generates corresponding warning signals; The adaptive discriminant tuning mechanism dynamically adjusts the parameters of the deep learning discriminant model according to the error between the predicted results and the actual data, thus improving the adaptability of the discriminant model. The judgment results output by the heavy pollution judgment module are transmitted to the intelligent early warning and protection suggestion module through the data interface for generating early warnings, and are also transmitted to the event-driven dynamic monitoring module to trigger specific event response measures.

3. According to the big data-based regional heavy pollution identification and forecasting system of claim 2, it is characterized by: The feature extraction module obtains pollutant concentration, meteorological data and multi-dimensional feature data of traffic flow extracted from weighted data from the adaptive weight allocation module through a data interface; The feature extraction module provides feature extraction algorithms and high-dimensional feature mapping functions; The feature extraction algorithm is based on the multidimensional matrix construction method to extract and construct the multidimensional feature matrix of pollutants; The high-dimensional feature mapping function uses machine learning algorithms to map pollution features into high-dimensional space to capture complex pollution causes and transmission characteristics; The feature extraction module sends the high-latitude feature data to the heavy pollution identification module and the pollution diffusion path simulation module through the data interface for pollution level judgment and calculation of pollution diffusion simulation input.

4. The regional severe pollution identification and forecasting system based on big data according to claim 3 is characterized by: The adaptive weight allocation module adjusts the weight of each data through a dynamic weight allocation algorithm based on the multi-source data transmitted by the multi-source data acquisition module through the data interface, so as to adapt to different pollution scenarios; The dynamic weight allocation module provides weight allocation algorithm and weight optimization feedback mechanism; The dynamic weight allocation algorithm automatically adjusts the weight of data sources in major pollution incidents and daily monitoring situations through a weight optimization mechanism based on historical data, making data fusion scenario-adaptive. The weight optimization feedback mechanism is to pass the real-time feedback data after the pollution event to the weight optimization mechanism to improve the dynamic response of weight allocation; The adaptive weight allocation module transfers the weighted fusion data to the feature extraction module and the heavy pollution discrimination module, providing accurate data input for feature extraction and pollution discrimination.

5. The regional severe pollution identification and forecasting system based on big data according to claim 4 is characterized by: The multi-source data acquisition module is used to collect real-time data from multiple data sources such as air quality monitoring stations, meteorological stations, traffic flow monitoring equipment and satellite remote sensing systems, and convert the data into a unified format through standardized processing, cleaning and normalization methods to facilitate analysis and processing by subsequent modules. The data generated by the multi-source data acquisition module is transmitted to the adaptive weight allocation module and the feature extraction module through the data interface to realize weighted fusion and feature analysis of the data.

6. The regional severe pollution identification and forecasting system based on big data according to claim 1 is characterized by: The event-driven dynamic monitoring module automatically triggers the event-driven mechanism based on the warning signal of the intelligent warning and protection suggestion module; The event-driven dynamic monitoring module provides event trigger condition setting, real-time dynamic monitoring and model automatic correction functions; Event trigger condition setting Set event trigger rules for traffic peak, industrial emissions and weather deterioration conditions, and start real-time dynamic monitoring when the trigger conditions are met; Real-time dynamic monitoring means that once the event-driven mechanism is triggered, the system will monitor pollutant changes in a specific area and feed back real-time data to the model's adaptive tuning mechanism; The automatic model correction function dynamically adjusts the parameters of the heavy pollution identification model and the pollution diffusion path simulation module based on monitoring data to improve the accuracy of the response to emergencies; The real-time monitoring data of the event-driven dynamic monitoring module is transmitted to the multi-scale meteorological forecast optimization module through the data interface to optimize the input parameters of pollution diffusion prediction.

7. The regional severe pollution identification and forecasting system based on big data according to claim 6 is characterized by: The multi-scale meteorological forecast optimization module realizes dynamic optimization prediction of pollution diffusion based on real-time monitoring data and multi-level meteorological data provided by the event-driven dynamic monitoring module; The multi-scale meteorological forecast optimization module provides multi-scale meteorological data fusion, fine particle diffusion model optimization algorithm and pollution trend map generation functions; Multi-scale meteorological data fusion is the fusion of local, regional and national meteorological data to provide forecast accuracy for meteorology at different scales; The fine particulate matter diffusion model optimization algorithm combines meteorological data and particulate matter characteristics (particle size and composition) to dynamically predict the diffusion rate, concentration distribution and diffusion radius of pollutants; The pollution trend map generation function generates pollution diffusion trend maps for multiple time periods, showing the pollution concentration distribution and impact range at different time points in the future; The pollution trend data of the multi-scale meteorological forecast optimization module is transmitted to the intelligent early warning and protection recommendation module through the data interface, providing future pollution change forecasts and providing management departments with predictions and decision support for prevention and control strategies.

8. The regional severe pollution identification and forecasting system based on big data according to claim 1 is characterized by: The intelligent early warning and protection suggestion module is used to generate graded early warning information and personalized protection suggestions based on the structure of the heavy pollution identification module and the pollution diffusion path simulation module. The output data of the intelligent early warning and protection suggestion module is transmitted to the event-driven dynamic monitoring module through the data interface as a trigger condition to execute real-time monitoring and early warning response measures.

9. A method for distinguishing and predicting regional severe pollution based on big data, applicable to a system for distinguishing and predicting regional severe pollution based on big data as claimed in any one of claims 1 to 8, characterized in that: The discrimination prediction method comprises the following steps: S1, obtain real-time air quality monitoring data, meteorological data, traffic flow data and satellite remote sensing data from the multi-source data acquisition module, and pre-process the data; S2. In the feature extraction module, multi-dimensional features of pollutant concentration, meteorological factors and traffic flow are extracted from the comprehensive data stream, and the data is mapped to a multi-dimensional space through a high-dimensional feature construction algorithm to generate a feature matrix; S3, input the feature matrix into the heavy pollution discrimination module, and analyze the data through the combination model of convolutional neural network (CNN) and long short-term memory network (LSTM). The model discriminates the pollution level based on the input high-dimensional feature data, judges the pollution status in the current area, and classifies the results into light, moderate and heavy pollution; S4. In the pollution diffusion path simulation module, based on meteorological data and the location of pollution sources, the pollution diffusion path simulation algorithm is used to predict the pollutant diffusion trend and generate a three-dimensional pollution diffusion map.

10. A regional severe pollution identification and forecasting method based on big data according to claim 9, characterized in that: The discrimination prediction method further comprises the following steps: S11. All kinds of data are weighted and integrated through the adaptive weight distribution module, and the weights of different data sources are dynamically adjusted according to the pattern of historical pollution events to generate a comprehensive data stream, providing high-quality data input for subsequent identification and prediction; S21, the feature matrix extracts key features related to heavy pollution events through a multi-layer feature extraction algorithm; S31. If a severe pollution state is detected, the corresponding warning signal is triggered, and the system automatically optimizes the parameters of each model based on the feedback of historical and real-time data to improve the adaptability to sudden severe pollution events; S41. Combined with the multi-scale meteorological data fusion and optimization module, the system makes short-term and long-term predictions on pollution diffusion trends and generates pollution trend graphs for multiple time periods. At the same time, the intelligent early warning and protection suggestion module provides early warning information to the public based on the diffusion prediction results and pushes personalized protection suggestions.

Images