Urban fine particulate matter traceability system and method based on multiple atmosphere super stations

Abstract

The invention provides an urban fine particulate matter traceability system and method based on multiple atmosphere super stations, and relates to the technical field of atmosphere pollution monitoring and traceability, and the system comprises a data collection module which is used for collecting observation data in real time through multiple atmosphere super stations; the data preprocessing module is used for preprocessing the observation data; the traceability model construction module is used for constructing a traceability model based on a positive definite matrix factor receptor model and a random forest algorithm; the real-time collaborative traceability module is used for inputting the preprocessed data into a traceability model, outputting a pollution source contribution ratio and a regional emission weight through parallel calculation, and introducing an atmospheric diffusion model to dynamically correct a traceability result; and the result display and analysis module is used for displaying the traceability result through a dynamic pie chart and a 3D histogram based on an interactive visualization tool of WebGL. According to the method, the defects of a traditional method are overcome, and efficient decision support is provided for urban pollution control.

Description

Technical Field

[0001] The present invention relates to the technical field of air pollution monitoring and source tracing, and particularly relates to an urban fine particulate matter source tracing system and method based on multiple air super stations. Background Art

[0002] At present, with the accelerating urbanization process, the air pollution problem is becoming increasingly serious. Fine particulate matter (PM 2.5 ) has become a key pollutant endangering human health and the ecological environment. It can penetrate deep into the human respiratory system, causing various diseases, and at the same time has a negative impact on climate, visibility, etc. Therefore, accurately and quickly determining the source of PM 2.5 has become the core of formulating effective pollution control strategies.

[0003] Early air pollution source tracing technologies mainly relied on the analysis of pollutant chemical components, and inferred pollution sources by comparing the characteristics of pollutants from different sources. However, the information obtained by this method is limited, and it is impossible to accurately quantify the contribution degree of each pollution source, making it difficult to meet actual needs.

[0004] With the development of computer technology, receptor models have become the main tools for urban air pollution source tracing, among which positive matrix factorization and chemical mass balance models are widely used. However, these traditional models have many limitations:

[0005] Insufficient spatial representativeness: Traditional models mostly rely on the observation data of a single urban air super station or manual off-line observation data. The monitoring range of a single air super station is limited and can only cover the surrounding area, making it difficult to represent the air pollution situation of the entire city or region. The sources of urban air pollution are extensive and complex, including industrial emissions, vehicle exhaust, biomass burning, dust, etc. Scattered pollution sources may not be effectively identified and traced, resulting in the lack of comprehensiveness and pertinence of treatment measures.

[0006] Limited vertical detection ability: Conventional air super stations mainly monitor near-surface air pollutants and have insufficient ability to detect the vertical distribution of high-altitude pollutants. In fact, the distribution of pollutants in the vertical direction is not uniform, and high-altitude pollutants can affect near-surface air quality through atmospheric vertical exchange. The fixed monitoring points of a single air super station are difficult to capture these dynamic changes, making the source tracing results lagging and uncertain, and unable to timely and accurately reflect the impact of the vertical structure of air pollution on ground pollution.

[0007] Limited time resolution: The update frequency of manual off-line monitoring data is low, usually sampling once a day or every few days. In the urban environment where air quality changes rapidly, this low-frequency monitoring cannot reflect the changes, diffusion, and migration processes of air quality in real time. In the face of sudden pollution events, traditional models are difficult to quickly trace the source, resulting in the inability to take effective response measures in time and delaying the treatment opportunity.

[0008] These deficiencies in the prior art in the aspect of urban fine particulate matter source tracing seriously restrict the development of air pollution control work. Therefore, it is extremely urgent to develop a new method for efficient and accurate fine particulate matter source tracing, which can not only make up for the deficiencies of traditional technologies, but also provide strong technical support for urban air pollution control, and has important practical significance. Summary of the Invention

[0009] To this end, the present invention provides a system and method for urban fine particulate matter source tracing based on multiple atmospheric superstations, which is used to solve the problem that in the prior art, traditional source tracing methods rely on single-station or manual off-line observation data, with insufficient spatial representativeness, limited vertical detection ability, and limited time resolution, and it is difficult to meet the requirements of rapid and accurate source tracing of fine particulate matter in complex urban environments.

[0010] In order to solve the above problems, an embodiment of the present invention provides a system for urban fine particulate matter source tracing based on multiple atmospheric superstations, and the system includes: A data acquisition module, which is used to collect observation data in real time through multiple atmospheric superstations. The observation data includes fine particulate matter concentration, chemical composition and meteorological parameter data. The atmospheric superstations include urban stations, suburban stations and regional stations, and are arranged according to a weighted scoring model of urban scale, pollution source distribution, population density and topography; A data preprocessing module, which is used to preprocess the observation data; A source tracing model construction module, which is used to construct a source tracing model based on the positive matrix factorization receptor model and the random forest algorithm; A real-time collaborative source tracing module, which is used to input the preprocessed data into the source tracing model, and output the source contribution ratio and regional emission weight through parallel computing, and at the same time introduce an atmospheric diffusion model to dynamically correct the source tracing result; A result display and analysis module, which is used for an interactive visualization tool based on WebGL to display the source tracing result through a dynamic pie chart and a 3D bar chart.

[0011] Preferably, constructing the source tracing model based on the positive matrix factorization receptor model and the random forest algorithm includes: Using the positive matrix factorization receptor model to decompose the concentration matrix of each chemical component in the observed fine particulate matter into , where is the number of source factors, m is the number of samples, n is the type of chemical components, is the factor contribution matrix, is the factor spectrum matrix; Iteratively optimizing the positive matrix factorization receptor model through the Bootstrap resampling technique; Selecting the optimal number of source factors through cross-validation , maximize the model explained variance and minimize the sum of squared residuals , subject to ; Utilize the random forest algorithm combined with knowledge graph technology to fuse the prior knowledge of the spatio-temporal distribution of pollution sources, the emission characteristics of pollution sources, and meteorological conditions, optimize the model input features, and jointly evaluate the model performance through the leave-one-out method and K-fold cross-validation. The objective function is: ; where is the total number of samples, is the true label of the -th sample, is the predicted label of the -th sample, is the regularization coefficient, is the total number of feature dimensions, is the weight coefficient of the -th feature.

[0012] Preferably, the iterative optimization of the positive definite matrix factorization receptor model by the Bootstrap resampling technique includes: Randomly draw subsamples from the original data and generate a Bootstrap dataset by repeating a set number of times; Perform positive definite matrix factorization on each sub-dataset and calculate the mean and confidence interval of the factor contribution matrix; Constrain all elements of the factor contribution matrix and the factor spectrum matrix to be non-negative to ensure the physical rationality of the pollution source contribution.

[0013] Preferably, the calculation formula for the sum of squared residuals is: ; where is the sum of squared residuals; m represents the number of samples; n represents the number of chemical component types; p represents the number of pollution source factors; is an element in the original data matrix, representing the concentration value of the -th chemical component in the -th sample; is an element in the factor contribution matrix G, representing the contribution of the -th pollution source to the total concentration in the -th sample; is an element in the factor spectrum matrix F, representing the concentration ratio of the -th chemical component in the -th pollution source; is an element in the uncertainty matrix U, representing the -th sample in the Measurement uncertainty of chemical components.

[0014] Preferably, the formula of the weighted scoring model is: ; Where is the weight of the th atmospheric superstation, is the representative score of the th atmospheric superstation, is the coverage score of the th atmospheric superstation, is the data quality score of the th atmospheric superstation, and n is the number of atmospheric superstations.

[0015] Preferably, the introduction of the atmospheric diffusion model to dynamically correct the traceability result includes: ; Where is the pollutant concentration at the horizontal coordinate , vertical height z and time , is the total number of pollution sources, is the horizontal position coordinate of the th pollution source, is the emission intensity of the th pollution source, and are diffusion parameters, is the wind speed, is the emission height of the th pollution source.

[0016] Preferably, in the result display and analysis module, the pollution source contribution ratio and regional emission weight are displayed by superimposing a dynamic pie chart and a 3D heat map, supporting users to interactively screen the spatial-temporal range, and calculating the regional pollution contribution difference through the following formula: ; Where and are the pollution source contribution rates of urban stations, suburban stations or regional stations respectively, is the regional pollution contribution difference value.

[0017] The data preprocessing module includes a data cleaning unit, a standardization processing unit and a feature extraction unit. The data cleaning unit uses 3 Combined with an adaptive low-pass filter, outliers are removed according to the principle; the normalization processing unit uses the Z-score normalization method with a dynamic threshold for concentration data and the Min-Max normalization method for chemical composition data; the feature extraction unit extracts key features through a deep neural network combined with principal component analysis and linear discriminant analysis.

[0018] Preferably, the formula for the Z-score normalization method with a dynamic threshold is: ; where represents the normalized concentration value, represents the real-time concentration information of the current monitoring station, represents the dynamically calculated real-time mean value, represents the dynamically calculated real-time standard.

[0019] An embodiment of the present invention also provides a method for tracing the source of fine particulate matter in a city based on multiple atmospheric super stations. The method uses the above-mentioned system for tracing the source of fine particulate matter in a city based on multiple atmospheric super stations, and includes the following steps: Observation data are collected in real time through multiple atmospheric super stations. The observation data include fine particulate matter concentration, chemical composition, and meteorological parameter data. The atmospheric super stations include urban stations, suburban stations, and regional stations, and are arranged according to a weighted scoring model of urban scale, pollution source distribution, population density, and topography and landform. The observation data are preprocessed. A source tracing model is constructed based on the positive definite matrix factorization receptor model and the random forest algorithm. The preprocessed data are input into the source tracing model, and the source contribution ratio and regional emission weight are output through parallel computing. At the same time, an atmospheric diffusion model is introduced to dynamically correct the source tracing result. Based on the WebGL-based interactive visualization tool, the source tracing result is displayed through a dynamic pie chart and a 3D bar chart.

[0020] It can be seen from the above technical solutions that the present invention application has the following beneficial effects: First, multi-dimensional data fusion and high-precision processing. By scientifically arranging urban stations, suburban stations, and regional stations, combining a dynamic weighted scoring model to optimize the site weights, and using the Z-score normalization with a dynamic threshold and the Min-Max normalization method, the representativeness and comparability of the data are effectively improved. The deep neural network combined with principal component analysis (PCA) and linear discriminant analysis (LDA) automatically extracts key features, significantly reducing redundant information and providing a high-quality data basis for the model.

[0021] Second, intelligent model and rapid response ability. Innovatively integrate the Positive Matrix Factorization receptor model (PMF) with the random forest algorithm. Through Bootstrap resampling iterative optimization and cross-validation techniques, ensure the physical rationality of pollution source contributions and the high interpretability of the model. Combined with a parallel computing architecture, single-site source tracing analysis can be completed within 5 minutes, with an accuracy rate of over 90%, significantly superior to the timeliness and accuracy of traditional methods.

[0022] Third, dynamic visualization and scientific decision-making support. Based on a WebGL-based interactive tool, realize multi-dimensional visual displays such as 3D heat maps and dynamic pie charts, support users to screen the spatio-temporal range in real time and calculate the differences in regional pollution contributions. Combine the dynamic correction results of the atmospheric diffusion model and intelligent early warning algorithms to provide accurate pollution source distributions, contribution ratios, and trend predictions for environmental management departments, and help formulate efficient governance strategies. Brief Description of the Drawings

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly describe the drawings required in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be construed as imposing any limitations on the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them: Figure 1 It is a block diagram of a system for tracing urban fine particulate matter based on multiple atmospheric supersites provided by the present invention; Figure 2 It is a flowchart of a method for tracing urban fine particulate matter based on multiple atmospheric supersites provided by the present invention. Detailed Embodiments

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0025] Embodiment 1: To solve the problem that in the prior art, traditional source tracing methods rely on single-site or manual off-line observation data, with insufficient spatial representativeness, limited vertical detection capabilities, and limited time resolution, making it difficult to meet the requirements for rapid and accurate source tracing of fine particulate matter in complex urban environments, as Figure 1 shown, the present invention proposes a system for tracing urban fine particulate matter based on multiple atmospheric supersites, and the system includes: The data acquisition module 100 is used to collect observation data in real time through multiple atmospheric super stations. The observation data includes fine particulate matter concentration, chemical composition, and meteorological parameter data. The atmospheric super stations include urban stations, suburban stations, and regional stations, and are arranged according to a weighted scoring model based on urban scale, pollution source distribution, population density, and topography and geomorphology. The data preprocessing module 200 is used to preprocess the observation data. The source tracing model construction module 300 is used to construct a source tracing model based on the positive definite matrix factorization receptor model and the random forest algorithm. The real-time collaborative source tracing module 400 is used to input the preprocessed data into the source tracing model, and output the source contribution ratio and regional emission weight through parallel computing. At the same time, an atmospheric diffusion model is introduced to dynamically correct the source tracing results. The result display and analysis module 500 is used for an interactive visualization tool based on WebGL to display the source tracing results through dynamic pie charts and 3D bar charts.

[0026] As can be seen from the above technical solutions, the present invention proposes an urban fine particulate matter source tracing system based on multiple atmospheric super stations. By arranging urban stations, suburban stations, and regional stations (covering multi-dimensional pollution sources and topography) through a weighted scoring model, using dynamic threshold Z-score and Min-Max normalization (improving data comparability) combined with a deep neural network and PCA / LDA (extracting key features) to efficiently preprocess multi-source data; constructing a PMF-random forest fusion model (physical rationality and high interpretability) to analyze the source contribution, combined with parallel computing and an atmospheric diffusion model (5-minute rapid response, dynamically correcting errors), and through WebGL interactive visualization (3D heat map, dynamic pie chart) to display the source tracing results in real time, realizing rapid and accurate source tracing of fine particulate matter, and providing dynamic and visual scientific decision-making support for pollution control.

[0027] In the data acquisition module 100, first, according to factors such as urban scale, pollution source distribution, population density, and topography and geomorphology, a weighted scoring model is used to determine the layout of the atmospheric super stations. The formula of the weighted scoring model is: ; Among them, is the weight of the th atmospheric super station, is the representative score of the th atmospheric super station, is the coverage score of the th atmospheric super station, is the data quality score of the th atmospheric super station, and n is the number of atmospheric super stations.

[0028] According to the calculation results, at least 3 atmospheric super stations should be reasonably arranged in different areas of the city, including urban stations, suburban stations and regional stations. The urban station is set at the center of the urban built-up area, responsible for monitoring the overall air quality status and change trends of the city; the suburban station is located in the suburbs around the city, used to compare the air quality differences between the city and the suburbs and monitor the urban pollution diffusion situation; the regional station is set outside the city to monitor the air quality status and pollutant transmission rules within a large area.

[0029] The present invention arranges stations by integrating multiple factors, enabling the monitoring network to cover different functional areas, enhancing the representativeness and comprehensiveness of the monitoring data, being able to accurately reflect the air quality status of different areas and the whole of the city, and providing a more reliable data basis for subsequent source tracing analysis.

[0030] The construction of each atmospheric super station strictly follows relevant standards and specifications, is equipped with professional monitoring instruments and equipment, such as fine particulate matter concentration monitors, chemical composition analyzers, meteorological parameter monitors, etc., and a perfect quality control and quality assurance system is established. The instruments and equipment are calibrated and maintained regularly to ensure their stable operation and guarantee the accuracy and reliability of the monitoring data.

[0031] The present invention uses professional instruments to ensure accurate data collection, and the quality control system guarantees the stable operation of the instruments, improves the accuracy and reliability of the data, provides high-quality data for the source tracing model, and avoids deviations in the source tracing results caused by data errors.

[0032] In this embodiment, each atmospheric super station is equipped with professional monitoring instruments, including: Fine particulate matter concentration monitor: The measurement range is 0 - 1000 μg / m³, and the detection limit is not less than 1 μg / m³, which is used to accurately measure the fine particulate matter concentration.

[0033] Chemical composition analyzer: It includes an on-line automatic monitor for water-soluble ions chromatography in the particle group, an on-line automatic monitor for inorganic elements, and an on-line automatic monitor for organic carbon and inorganic carbon. It can accurately detect the main chemical components in PM 2.5 such as 8 water-soluble components such as sulfate and nitrate, nearly 20 inorganic elements such as Fe, Zn, and Cu, as well as organic carbon and elemental carbon, and the detection limit is lower than 1.0 ng / m³.

[0034] Meteorological parameter monitor: It can accurately measure meteorological parameters such as temperature, humidity, wind speed, and wind direction in real time, and the measurement accuracy is better than ±5%.

[0035] Furthermore, the monitoring instruments of each atmospheric super station collect data such as fine particulate matter concentration, chemical composition, and meteorological parameters in real time at regular time intervals (such as every 60 minutes), and transmit the data to the data processing center through a high-speed and stable wireless transmission network. Encryption and other measures are taken during the transmission process to ensure the integrity and security of the data, preventing data loss or tampering.

[0036] The real-time collection and efficient transmission of the present invention enable the data processing center to obtain the latest monitoring data, provide support for rapid traceability, timely reflect the changes in air quality, and contribute to a rapid response in the event of a pollution incident.

[0037] In the data preprocessing module 200, after receiving the data, the data cleaning unit cleans the data by using the 3σ principle combined with an adaptive low-pass filter. Through this method, outliers are identified and removed, noise is filtered, and combined with the relevant monitoring data of multiple super stations, the parameter settings are adjusted according to the real-time monitoring data of each station to achieve more accurate data cleaning.

[0038] Furthermore, for the concentration data, the standardization processing unit adopts the Z-score standardization method with a dynamic threshold, and the formula is . Among them, represents the standardized concentration value, represents the real-time concentration information of the current monitoring station, represents the dynamically calculated real-time mean, represents the dynamically calculated real-time standard. Adopting the Z-score standardization based on a dynamic threshold can adjust the mean and standard deviation according to the real-time distribution of the data, making the standardized data more comparable.

[0039] For the chemical composition data, the Min-Max standardization method is used, fully considering the actual physical meaning of the data to avoid information loss during the standardization process.

[0040] Furthermore, the feature extraction unit extracts key features by using a method that combines a deep neural network with principal component analysis (PCA) and linear discriminant analysis (LDA). In this way, the features hidden in the data that are closely related to the sources of fine particulate matter are automatically mined, data redundancy is reduced, the true and effective information is retained, and the efficiency and accuracy of the traceability model are improved. The processed data is stored in a dedicated database to provide a high-quality data basis for the subsequent construction of the traceability model.

[0041] In the traceability model construction module 300, the present invention uses the positive matrix factorization receptor model (PMF) to decompose the concentration matrix of each chemical component in the observed atmospheric particulate matter into a factor contribution matrix and a factor spectrum matrix. At the same time, combined with the random forest algorithm, prior knowledge such as the spatio-temporal distribution of pollution sources, emission characteristics, and meteorological conditions is integrated. PMF can effectively analyze the contributions of pollution sources and chemical composition spectra, and the random forest algorithm integrates prior knowledge to optimize the model, improve the adaptability and accuracy of the model, and more accurately identify pollution sources.

[0042] Specifically, the concentration matrix of each chemical component in the observed fine particulate matter is decomposed by using the positive matrix factorization receptor model into , where is the number of pollution source factors, m is the number of samples, n is the type of chemical components, is the factor contribution matrix, is the factor spectrum matrix.

[0043] Furthermore, the positive matrix factorization receptor model is iteratively optimized by using the Bootstrap resampling technique. Sub-samples are randomly drawn from the original data, and the Bootstrap data set is generated by repeating a set number of times (1000 times); positive matrix factorization is performed on each sub-data set, and the mean and confidence interval of the factor contribution matrix are calculated; all elements of the factor contribution matrix and the factor spectrum matrix are constrained to be non-negative values to ensure the physical rationality of the pollution source contribution.

[0044] Furthermore, the optimal number of pollution source factors is selected through cross-validation, to maximize the model's explained variance and minimize the sum of squared residuals , satisfying . The formula for the sum of squared residuals is: where, is the sum of squared residuals; m represents the number of samples; n represents the type of chemical components; p represents the number of pollution source factors; is an element in the original data matrix, representing the concentration value of the th chemical component in the th sample; is an element in the factor contribution matrix G, representing the contribution of the th pollution source to the total concentration in the th sample; is an element in the factor spectrum matrix F, representing the concentration ratio of the th chemical component in the th pollution source; is an element in the uncertainty matrix U, representing the th sample Measurement uncertainty of chemical components

[0045] Furthermore, the random forest algorithm is used in combination with the knowledge graph technology to fuse the prior knowledge of the spatio-temporal distribution of pollution sources, the emission characteristics of pollution sources, and meteorological conditions, optimize the input features of the model, and jointly evaluate the performance of the model through the leave-one-out method and K-fold cross-validation. The objective function is: ; Wherein, is the total number of samples, is the true label of the th sample, is the predicted label of the th sample, is the regularization coefficient, is the total number of feature dimensions, is the th weight coefficient of the feature.

[0046] In the real-time collaborative source tracing module 400, the preprocessed data is input into the constructed source tracing model, and parallel computing architecture is used for rapid calculation to output the source contribution ratio of fine particulate matter and the regional emission weight. The present invention can complete the single-site source tracing analysis within 5 minutes, with an accuracy of over 90%, significantly superior to the timeliness and accuracy of traditional methods.

[0047] Furthermore, the present invention introduces an atmospheric diffusion model to dynamically correct the source tracing results. The formula of the atmospheric diffusion model is: ; Wherein, is the pollutant concentration at the horizontal coordinate , vertical height z and time , is the total number of pollution sources, is the horizontal position coordinate of the th pollution source, is the th emission intensity of the pollution source, and are diffusion parameters, is the wind speed, is the th emission height of the pollution source.

[0048] Furthermore, the source tracing results of the present invention are updated in real time and stored in the result database for subsequent query and analysis.

[0049] In the result display and analysis module 500, the traceability results are displayed in various intuitive forms such as dynamic pie charts, 3D bar charts, and 3D heat maps through a WebGL-based interactive visualization tool. These visualization charts support users to freely zoom, rotate, and filter the spatio-temporal range, facilitating users to deeply analyze the traceability results from different perspectives.

[0050] The present invention provides rich data analysis tools, including trend prediction analysis based on time series, pollution source correlation analysis based on association rule mining, pollution type classification analysis based on clustering analysis, etc. At the same time, through the formula calculate the regional pollution contribution difference, where, and are the pollution source contribution rates of urban stations, suburban stations or regional stations respectively, is the regional pollution contribution difference value, which helps the environmental management department deeply understand the laws behind the data and provides comprehensive data support for formulating scientific and reasonable pollution control strategies.

[0051] Embodiment 2: As Figure 2 shown, the present invention provides a method for tracing fine particulate matter in cities based on multiple atmospheric supersites. This method uses the system for tracing fine particulate matter in cities based on multiple atmospheric supersites in Embodiment 1 above, and includes the following steps: S1: Real-time collect observation data through multiple atmospheric supersites. The observation data includes fine particulate matter concentration, chemical composition, and meteorological parameter data. The atmospheric supersites include urban stations, suburban stations, and regional stations, and are arranged according to a weighted scoring model of urban scale, pollution source distribution, population density, and topography; S2: Preprocess the observation data; S3: Construct a traceability model based on the positive definite matrix factorization receptor model and the random forest algorithm; S4: Input the preprocessed data into the traceability model, and output the pollution source contribution ratio and regional emission weight through parallel computing. At the same time, introduce an atmospheric diffusion model to dynamically correct the traceability results; S5: Use a WebGL-based interactive visualization tool to display the traceability results through dynamic pie charts and 3D bar charts.

[0052] A method for tracing the source of urban fine particulate matter based on multiple atmospheric super stations in this embodiment uses the aforementioned system for tracing the source of urban fine particulate matter based on multiple atmospheric super stations. Therefore, the specific implementation manners in the method for tracing the source of urban fine particulate matter based on multiple atmospheric super stations can be seen in the embodiment part of the system for tracing the source of urban fine particulate matter based on multiple atmospheric super stations described above. For example, in steps S1, S2, S3, S4, and S5, the data acquisition module 100, data preprocessing module 200, source tracing model construction module 300, real-time collaborative source tracing module 400, and result display and analysis module 500 in the aforementioned system for tracing the source of urban fine particulate matter based on multiple atmospheric super stations are respectively used. Therefore, the specific implementation manners can refer to the descriptions of the corresponding embodiments of each part. To avoid redundancy, they will not be elaborated here.

[0053] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0054] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0055] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks. These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide for implementing the specified functions in the processFigure 1 one process or multiple processes and / or boxes Figure 1 steps of functions specified in one box or multiple boxes.

[0056] Obviously, the above embodiments are only examples for clear illustration and are not limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all implementation manners here. And the obvious changes or modifications derived therefrom are still within the protection scope of the present invention.

Claims

1. An urban fine particulate matter tracing system based on multiple atmospheric super stations, characterized in that, including: a data acquisition module for collecting observation data in real time through multiple atmospheric supersites, where the observation data includes fine particulate matter concentration, chemical composition, and meteorological parameter data, and the atmospheric supersites include urban sites, suburban sites, and regional sites, and are arranged according to a weighted scoring model of urban scale, pollution source distribution, population density, and topography; a data preprocessing module for preprocessing the observation data; a source apportionment model construction module for constructing a source apportionment model based on the positive matrix factorization receptor model and the random forest algorithm; a real-time collaborative source apportionment module for inputting the preprocessed data into the source apportionment model and outputting the source contribution ratio and regional emission weight through parallel computing, and at the same time introducing an atmospheric diffusion model to dynamically correct the source apportionment results; a result display and analysis module for an interactive visualization tool based on WebGL to display the source apportionment results through dynamic pie charts and 3D bar charts.

2. The urban fine particulate matter tracing system based on multiple atmospheric super stations according to claim 1, characterized in that, The construction of the source apportionment model based on the positive matrix factorization receptor model and the random forest algorithm includes: Using the positive definite matrix factorization receptor model, the concentration matrix of each chemical component in the observed fine particulate matter is decomposed into , where is the number of pollution source factors, m is the number of samples, n is the number of chemical component types, is the factor contribution matrix, is the factor spectrum matrix; iteratively optimizing the positive matrix factorization receptor model through the Bootstrap resampling technique; Select the optimal number of pollution source factors through cross-validation , so that the model explained variance is maximized and the sum of squared residuals is minimized, satisfying ; using the random forest algorithm combined with the knowledge graph technology to fuse the prior knowledge of the spatio-temporal distribution of pollution sources, the emission characteristics of pollution sources, and meteorological conditions, optimizing the model input features, and jointly evaluating the model performance through the leave-one-out method and K-fold cross-validation, and the objective function is: ； Among them, is the total number of samples, is the true label of the -th sample, is the predicted label of the -th sample, is the regularization coefficient, is the total number of feature dimensions, is the -th weight coefficient of the feature.

3. The urban fine particulate matter source tracing system based on multiple atmospheric super stations according to claim 2, wherein The iterative optimization of the positive matrix factorization receptor model through the Bootstrap resampling technique includes: randomly extracting subsamples from the original data and generating a Bootstrap data set by repeating a set number of times; performing positive matrix factorization on each sub-data set, and calculating the mean and confidence interval of the factor contribution matrix; constraining all elements of the factor contribution matrix and the factor spectrum matrix to be non-negative values to ensure the physical rationality of the source contribution.

4. The urban fine particulate matter traceability system based on multiple atmospheric super stations according to claim 2, characterized in that, The calculation formula of the residual sum of squares is: ； Among them, is the sum of squared residuals; m represents the number of samples; n represents the types of chemical components; p represents the number of pollution source factors; is an element in the original data matrix, representing the -th sample and the -th concentration value of the chemical component; is an element in the factor contribution matrix G, representing the contribution of the -th sample and the -th pollution source to the total concentration; is an element in the factor spectrum matrix F, representing the concentration ratio of the -th pollution source and the -th chemical component; is an element in the uncertainty matrix U, representing the measurement uncertainty of the -th sample and the -th chemical component.

5. The urban fine particulate matter traceability system based on multiple atmospheric super stations according to claim 1, characterized in that, The formula of the weighted scoring model is: ； Among them, is the weight of the th atmospheric super station, is the representative score of the th atmospheric super station, is the coverage score of the th atmospheric super station, is the data quality score of the th atmospheric super station, and n is the number of atmospheric super stations.

6. The urban fine particulate matter tracing system based on multiple atmospheric super stations according to claim 1, wherein The introduction of the atmospheric diffusion model to dynamically correct the source apportionment results includes: ； Among them, is the pollutant concentration at the horizontal coordinate , vertical height z, and time . is the total number of pollution sources, is the horizontal position coordinate of the th pollution source, is the th pollution source's emission intensity, and are dispersion parameters, is the wind speed, is the th pollution source's emission height.

7. The urban fine particulate matter traceability system based on multiple atmospheric super stations according to claim 1, characterized in that In the result display and analysis module, the source contribution ratio and regional emission weight are displayed by overlaying a dynamic pie chart and a 3D heat map, supporting users to interactively screen the spatio-temporal range, and calculating the regional pollution contribution difference through the following formula: ； Among them, and are the pollution source contribution rates of the urban station, suburban station or regional station respectively, is the regional pollution contribution difference value.

8. The urban fine particulate matter traceability system based on multiple atmospheric super stations according to claim 1, characterized in that The data preprocessing module includes a data cleaning unit, a standardization processing unit, and a feature extraction unit. The data cleaning unit uses 3 principles combined with an adaptive low-pass filter to remove outliers; the standardization processing unit uses the Z-score standardization method with a dynamic threshold for concentration data and the Min-Max standardization method for chemical composition data; the feature extraction unit extracts key features through a deep neural network combined with principal component analysis and linear discriminant analysis.

9. The urban fine particulate matter traceability system based on multiple atmospheric super stations according to claim 8, characterized in that, The formula of the Z-score normalization method for the dynamic threshold is: ； Among them, represents the standardized concentration value, represents the real-time concentration information of the current monitoring site, represents the real-time mean value calculated dynamically, represents the real-time standard calculated dynamically.

10. A method for tracing the sources of urban fine particulate matter based on multiple atmospheric super stations, characterized in that, The method uses the urban fine particulate matter source apportionment system based on multiple atmospheric supersites described in any one of claims 1 to 9, and includes the following steps: collecting observation data in real time through multiple atmospheric supersites, where the observation data includes fine particulate matter concentration, chemical composition, and meteorological parameter data, and the atmospheric supersites include urban sites, suburban sites, and regional sites, and are arranged according to a weighted scoring model of urban scale, pollution source distribution, population density, and topography; preprocessing the observation data; constructing a source apportionment model based on the positive matrix factorization receptor model and the random forest algorithm; inputting the preprocessed data into the source apportionment model and outputting the source contribution ratio and regional emission weight through parallel computing, and at the same time introducing an atmospheric diffusion model to dynamically correct the source apportionment results; An interactive visualization tool based on WebGL that displays the traceability results through dynamic pie charts and 3D bar charts.

Images