An industrial park air quality intelligent monitoring system based on multi-sensor fusion

Abstract

The application belongs to the field of environmental monitoring systems, and particularly relates to an industrial park air quality intelligent monitoring system based on multi-sensor fusion, which comprises a multi-parameter sensor network module, a data transmission and convergence module, a chemical coupling mechanism analysis module, a dynamic threshold and abnormality identification module, a pollution tracing and transmission analysis module and an intelligent early warning decision and release module; the chemical coupling mechanism analysis module integrates three modeling units of photochemical reaction path, oxidation reaction path and gas-particle conversion path, and quantitatively describes chemical reaction processes such as generation of ozone from volatile organic compounds and nitrogen oxides and generation of sulfates from sulfur dioxide; through the above technical solution, the application realizes cognitive leap from measurement of what concentration to why it is this concentration, and provides the environmental management department with causal explanation ability of pollution causes and quantitative technical support for precise tracing, differentiated control and the like.

Description

Technical Field

[0001] This invention belongs to the field of environmental monitoring systems, specifically a smart air quality monitoring system for industrial parks based on multi-sensor fusion. Background Technology

[0002] Industrial parks are home to a variety of industrial enterprises, resulting in a wide range of pollutant emissions. These pollutants are also prone to undergoing complex chemical reactions to generate secondary pollutants, which seriously endanger human health and the regional ecological environment. Accurately monitoring air quality and controlling pollution in industrial parks is a core task of regional ecological environmental protection. Air quality monitoring systems are essential equipment for pollution source tracing, environmental assessment, and early warning decision-making.

[0003] Existing air quality monitoring technologies in industrial parks mainly rely on multi-sensor nodes deployed throughout the area to collect pollutant concentration data. By processing multi-source monitoring data through conventional data fusion methods such as weighted averaging and principal component analysis, basic monitoring and anomaly alarms of pollutant concentrations in the park can be achieved, meeting the needs of routine environmental data statistical monitoring.

[0004] Existing monitoring data processing methods only physically overlay pollutant data, completely ignoring the chemical coupling and transformation mechanisms between atmospheric pollutants. They cannot distinguish the causes of pollutant exceedances, and can only output surface concentration data without providing causal explanations for pollution source tracing. This makes it difficult to support environmental management departments in formulating precise pollution control strategies, forming a core technological bottleneck for precise and intelligent pollution control in industrial parks.

[0005] Therefore, this invention provides an intelligent air quality monitoring system for industrial parks based on multi-sensor fusion. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve its technical problem is as follows: An intelligent air quality monitoring system for industrial parks based on multi-sensor fusion, comprising a multi-parameter sensor network module, a data transmission and aggregation module, a chemical coupling mechanism analysis module, a dynamic threshold and anomaly identification module, a pollution source tracing and transmission analysis module, and an intelligent early warning decision-making and dissemination module; the multi-parameter sensor network module is connected to the data transmission and aggregation module via a communication network; the data transmission and aggregation module is connected to the chemical coupling mechanism analysis module; the chemical coupling mechanism analysis module is communicatively connected to both the dynamic threshold and anomaly identification module and the pollution source tracing and transmission analysis module; the dynamic threshold and anomaly identification module is connected to the intelligent early warning decision-making and dissemination module; the pollution source tracing and transmission analysis module is connected to the intelligent early warning decision-making and dissemination module; and the intelligent early warning decision-making and dissemination module is communicatively connected to the data transmission and aggregation module.

[0008] In the multi-parameter sensor network module, the deployment density of monitoring micro-stations is set to 4 to 8 per square kilometer, and the horizontal distance between adjacent monitoring micro-stations is controlled within the range of 300 to 500 meters. The module is internally configured with a multi-parameter sensor module and a meteorological parameter sensor module. The multi-parameter sensor module includes a sulfur dioxide concentration sensor, a nitrogen oxide concentration sensor, a volatile organic compound concentration sensor, an ozone concentration sensor, and a fine particulate matter concentration sensor. The meteorological parameter sensor module includes a temperature sensor, a humidity sensor, a wind speed sensor, a wind direction sensor, and a solar radiation sensor. The sampling frequency of the monitoring micro-stations is set to... The sampling data is processed 3 to 10 times per minute, undergoing analog-to-digital conversion and preliminary filtering before being packaged and transmitted. In the data transmission and aggregation module, the deployment density of the edge computing gateway is set to one edge computing gateway for every 8 to 16 monitoring micro-stations. The edge computing gateway is internally configured with a data caching unit and a local analysis unit. The storage capacity of the data caching unit is not less than 32 gigabytes, and the execution capability of the local analysis unit is 500 million floating-point operations per second. The module adopts a dual-channel parallel transmission architecture of time-division duplex LTE network and low-power wide area network, with a data transmission latency of no more than 2 seconds and a reliability of no less than 99%.5%; In the chemical coupling mechanism analysis module, the coupling analysis engine is internally configured with a photochemical reaction path modeling unit, an oxidation reaction path modeling unit, and a gas-particle conversion path modeling unit. The photochemical reaction path modeling unit constructs a nonlinear response relationship between the concentration changes of volatile organic compounds and nitrogen oxides and the ozone generation rate based on the ozone generation standard EKMA curve. The oxidation reaction path modeling unit establishes a sulfate generation rate calculation model based on the oxidation reaction mechanism of sulfur dioxide and hydroxyl radicals. The gas-particle conversion path modeling unit establishes a calculation model for the distribution ratio of volatile organic compounds between the gas phase and the particulate phase based on the gas-particle distribution equilibrium model. The output results of the coupling analysis engine include the coupling strength coefficient and chemical conversion contribution ratio between various pollutants. In the dynamic threshold and anomaly identification module, the dynamic threshold calculation unit calculates the dynamic baseline concentration range based on the historical data sequence of pollutant concentrations at each monitoring point. The dynamic baseline concentration range includes two time scales: hourly baseline range and daily baseline range. The hourly baseline range is calculated based on historical data of the same period within the last 30 days, and the daily baseline range is calculated based on historical data within the last 90 days. The exceedance judgment unit calculates the real-time relative concentration of each monitoring point. Regarding the exceedance probability and exceedance magnitude of the dynamic baseline, the exceedance probability is calculated using a probability model based on Bayesian inference, while the exceedance magnitude is calculated using the standardized deviation obtained by dividing the difference between the real-time concentration and the mean of the dynamic baseline by the standard deviation of the dynamic baseline. In the pollution source tracing and transmission analysis module, the transmission trajectory tracking unit calculates the propagation direction and speed of pollutants based on the spatial topology of the monitoring network and the concentration gradient changes at each monitoring point. The air mass backward trajectory simulation unit uses a Lagrange air mass trajectory model combined with wind field data and mixing layer height data to simulate the transmission trajectory of the polluted air mass over the past 72 hours. The regional contribution analysis unit uses source-direction analysis results and the geographical coordinates of each pollution source to estimate the contribution ratio of each pollution source region to the pollution concentration in the monitoring area. In the intelligent early warning decision and release module, the decision generation engine performs multi-level early warning assessment, with early warning levels divided into four levels: blue, yellow, orange, and red. The decision generation engine comprehensively considers the exceedance of pollutant concentration, the type of pollution cause, the characteristics of pollution transmission, and the contribution ratio of each pollution source to generate early warning levels and response strategies. The response strategy library predefines response strategies corresponding to different early warning levels and different pollution cause types.

[0009] Preferably, in the multi-parameter sensor network module, each gaseous pollutant sensor within the multi-parameter sensor module adopts a dual redundancy configuration of electrochemical and optical sensors. The sulfur dioxide concentration sensor has a detection range of 0 to 500 micrograms per cubic meter and a resolution of 0.1 micrograms per cubic meter; the nitrogen oxide concentration sensor has a detection range of 0 to 200 micrograms per cubic meter and a resolution of 0.05 micrograms per cubic meter; the volatile organic compound concentration sensor uses photoionization detection, with a detection range of 0 to 10,000 micrograms per cubic meter and a resolution of 1 microgram per cubic meter; the ozone concentration sensor has a detection range of 0 to 400 micrograms per cubic meter and a resolution of 0.1 micrograms per cubic meter; and the fine particulate matter concentration sensor uses… Using the principle of laser scattering, the detection range is 0 to 500 micrograms per cubic meter, with a resolution of 0.1 micrograms per cubic meter. In the meteorological parameter sensor module, the temperature sensor has a measurement range of -40 to 60 degrees Celsius with an accuracy of ±0.3 degrees Celsius; the humidity sensor has a measurement range of 0 to 100% with an accuracy of ±2%; the wind speed sensor has a measurement range of 0 to 30 meters per second with an accuracy of ±0.2 meters per second; the wind direction sensor has a measurement range of 0 to 360 degrees with an accuracy of ±3 degrees; and the solar radiation sensor has a measurement range of 0 to 1400 watts per square meter with an accuracy of ±5 watts per square meter. The overall power consumption of the monitoring microstation is no more than 5 watts, the mean time between failures (MTBF) is no less than 8760 hours, and the protection level is no less than IP65.

[0010] Preferably, in the data transmission and aggregation module, the data preprocessing process executed by the edge computing gateway includes four steps: data filtering, data calibration, data completion, and data compression. The data filtering step uses a median filtering algorithm to remove spike noise from the sensor output, with the filtering window width set to 5 sampling points. The data calibration step calibrates and corrects the original concentration data based on calibration factors and a zero-point drift model. The data completion step uses a linear interpolation algorithm to interpolate and supplement data at missing time points. The data compression step uses a compression algorithm based on time series characteristics to compress the original data to 30% to 40% of the original data volume. The module's data transmission protocol supports three protocol formats: Transmission Control Protocol (TCP), User Data Packet Protocol (UDP), and Secure Hypertext Transfer Protocol (STP). Data transmission employs data compression and breakpoint resumption mechanisms to ensure the integrity and reliability of data transmission. A time-series database is deployed within the data aggregation platform to store all sensor data. The time-series database has a write capacity of no less than 100,000 records per second and a query capacity of no less than 50,000 aggregate queries per second.

[0011] Preferably, in the chemical coupling mechanism analysis module, the photochemical reaction path modeling unit integrates a photolysis rate calculation model and a free radical reaction kinetic model. The photolysis rate calculation model calculates the rate constants of various photochemical reactions based on solar radiation intensity, solar zenith angle, and ozone column concentration. The free radical reaction kinetic model establishes a free radical oxidation reaction network based on the reaction rate constants of hydroxyl radicals, nitrate radicals, and ozone with volatile organic compounds. The chemical transformation contribution decomposition algorithm of the coupling analysis engine decomposes the observed concentrations of each pollutant into two parts: direct emission contribution and chemical transformation contribution. The decomposition process adopts a constraint optimization algorithm based on chemical mass balance, with the constraint condition being the material conservation relationship of various chemical reaction paths. The coupling analysis engine is also equipped with an ozone generation sensitivity analysis function, which determines which sensitive range of ozone generation the current atmospheric environment is in by calculating the boundary conditions of the volatile organic compound sensitive area and the nitrogen oxide sensitive area.

[0012] Preferably, in the dynamic threshold and anomaly identification module, the dynamic threshold calculation unit is also configured with a meteorological condition adaptive adjustment function. This function dynamically corrects the dynamic baseline concentration range based on current meteorological parameters, including temperature, humidity, and solar radiation intensity. When the temperature is higher than 30 degrees Celsius, the baseline range of volatile organic compounds is adjusted upward by 15% to 25% of the historical baseline average. When the solar radiation intensity is higher than 800 watts per square meter, the baseline range of ozone is adjusted upward by 20% to 30% of the historical baseline average. The exceedance probability calculation result output by the exceedance determination unit is used to distinguish between normal fluctuations and abnormal events. When the exceedance probability is lower than 50%, it is judged as a normal fluctuation; when the exceedance probability is between 50% and 80%, it is judged as a mild anomaly; when the exceedance probability is between 80% and 95%, it is judged as a moderate anomaly; and when the exceedance probability is higher than 95%, it is judged as a severe anomaly. The exceedance determination unit also calculates the exceedance duration as an auxiliary judgment indicator. When the concentration exceeds the dynamic baseline for more than 30 minutes, an event record is triggered.

[0013] Preferably, in the pollution source tracing and transmission analysis module, the air mass trajectory calculation of the air mass backward trajectory simulation unit adopts a three-dimensional kinematic trajectory model. The model input data includes wind field data, mixing layer height data, and pollutant concentration data. The temporal resolution of the wind field data is 1 hour, and the spatial resolution is not less than 1 km x 1 km. The simulated trajectory time span is 72 hours, and the trajectory output interval is 1 hour. The module is configured with multiple starting point trajectory set generation functions. Each trajectory set contains 50 backward trajectories starting from the target monitoring point at the time of pollution occurrence. The starting heights of the 50 trajectories are set to 10 meters, 50 meters, and 10 meters, respectively. Five height layers—0 meters, 200 meters, and 500 meters—are used, with 10 trajectories generated in a fan-shaped distribution at 3.6-degree intervals. The regional contribution analysis unit uses a source-oriented attenuation model to estimate the regional contribution ratio of various pollution sources. The source-oriented attenuation model calculates the diffusion attenuation coefficient of pollution sources based on the downwind distance and wind speed. The regional contribution analysis unit also divides the monitoring area into three source areas: local contribution area, regional transmission area, and background contribution area. The local contribution area is defined as the area with a downwind distance of less than 20 kilometers, the regional transmission area is defined as the area with a downwind distance of 20 to 100 kilometers, and the background contribution area is defined as the area with a downwind distance of more than 100 kilometers.

[0014] Preferably, in the intelligent early warning decision and release module, the early warning level determination rules of the decision generation engine are as follows: the triggering condition for a blue warning is that the concentration of any pollutant exceeds the upper limit of the dynamic baseline with a probability of exceeding it between 50% and 80%; the triggering condition for a yellow warning is that the concentration of any pollutant exceeds the upper limit of the dynamic baseline with a probability of exceeding it between 80% and 95%, or the ozone concentration exceeds 80% of the environmental quality standard concentration limit; the triggering condition for an orange warning is that the concentration of any pollutant exceeds the upper limit of the dynamic baseline with a probability of exceeding it higher than 95%, or the ozone concentration exceeds the environmental quality standard concentration limit; the triggering condition for a red warning is that the concentration of any pollutant exceeds the upper limit of the dynamic baseline with a probability of exceeding it higher than 95%, and the pollution source tracing analysis results show that there is a significant regional transmission contribution, or the ozone concentration exceeds 150% of the environmental quality standard concentration limit; the strategy corresponding to a blue warning in the response strategy library is to increase the monitoring frequency and initiate data verification; the strategy corresponding to a yellow warning is to send emission reduction recommendation notices to key pollution sources; the strategy corresponding to an orange warning is to initiate an emergency emission reduction response and conduct pollution source investigation; and the strategy corresponding to a red warning is to initiate the highest level emergency response and implement regional joint prevention and control.

[0015] Preferably, the system of the present invention further includes a data quality control module, which is connected to the data transmission and aggregation module. The module is internally configured with a data integrity verification unit, a spatiotemporal consistency verification unit, and a physical rationality verification unit. The data integrity verification unit checks the continuity of the timestamp and the integrity of the numerical field of each data record. When a missing timestamp or an empty numerical field is detected, the record is marked as suspicious data. The spatiotemporal consistency verification unit checks whether there is an obvious and unreasonable spatial gradient in the concentration data of adjacent monitoring micro-stations at the same time. When the spatial gradient exceeds three times the standard deviation of a preset threshold, the relevant data is marked as suspicious data. The physical rationality verification unit checks whether the concentration data of each pollutant conforms to the physical conservation law, including the negative correlation between nitrogen oxides and ozone concentration and the positive correlation between sulfur dioxide and particulate matter sulfate concentration. When an abnormal correlation is detected, the relevant data is marked as suspicious data. The data quality report output by the data quality control module includes the data integrity rate, data availability rate, and data quality score of each monitoring micro-station.

[0016] Preferably, the system of the present invention further includes a three-dimensional visualization display module. The three-dimensional visualization display module is communicatively connected to the data transmission and aggregation module, the chemical coupling mechanism analysis module, the dynamic threshold and anomaly identification module, and the pollution source tracing and transmission analysis module. The module is internally configured with a park three-dimensional scene construction unit, a pollutant concentration field rendering unit, a pollution transmission dynamic demonstration unit, and an early warning information overlay display unit. The park three-dimensional scene construction unit constructs a three-dimensional scene model of the park based on the park's plan layout and building height data. The pollutant concentration field rendering unit generates a three-dimensional color cloud map of pollutant concentration distribution over the park based on real-time concentration data from monitoring micro-stations and spatial interpolation algorithms. The pollution transmission dynamic demonstration unit dynamically displays the transmission path and diffusion range of pollutant air masses in the three-dimensional scene based on the air mass backward trajectory simulation results. The early warning information overlay display unit overlays and displays the early warning status icons of each monitoring micro-station and the contribution ratio heat map of each pollution source area in the three-dimensional scene.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] 1. The present invention discloses an intelligent air quality monitoring system for industrial parks based on multi-sensor fusion. By constructing a chemical coupling mechanism analysis module, a chemical coupling sensing model between pollutants is established. This module integrates three sub-modules: photochemical reaction path modeling, oxidation reaction path modeling, and gas-particle conversion path modeling. It can quantitatively describe the process of ozone generation through photochemical reaction between volatile organic compounds and nitrogen oxides, the process of sulfate generation through oxidation reaction between sulfur dioxide and hydroxyl radicals, and the gas-particle conversion and distribution process of volatile organic compounds. This technical feature fundamentally breaks through the superficial limitation of existing monitoring systems that can only sense pollutant concentrations, realizing a leap from measuring what concentration to understanding why it is this concentration, and providing environmental management departments with the ability to explain the causal causes of pollution.

[0019] 2. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion described in this invention can accurately decompose the observed concentrations of each pollutant into two parts: direct emission contribution and chemical conversion contribution, through a chemical conversion contribution decomposition algorithm of a coupled analysis engine. This decomposition process uses a constraint optimization algorithm based on chemical mass balance, with the material conservation relationship of various chemical reaction paths as the constraint condition. The decomposition result can clearly determine whether the ozone concentration exceeds the standard due to insufficient reaction caused by the direct emission of ozone precursors locally or due to sufficient photochemical reaction of volatile organic compounds and nitrogen oxides transported from upstream. This technical feature enables the monitoring system to have the core ability to distinguish between direct emissions and chemical conversion, solving the fundamental defects of existing technologies in the judgment of anomaly types.

[0020] 3. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion described in this invention changes the traditional static threshold determination mode through a dynamic threshold and anomaly identification module. This module calculates the dynamic baseline concentration range based on the historical data sequence of pollutant concentrations at each monitoring point. This dynamic baseline considers not only changes in the time scale but also the impact of changes in meteorological conditions on the background concentration of pollutants. At the same time, it calculates the probability of the real-time concentration exceeding the dynamic baseline through a probability model based on Bayesian inference. This technical feature transforms the early warning determination from relying on a single concentration threshold to a multi-dimensional determination that comprehensively considers the concentration level, fluctuation characteristics, and duration, effectively avoiding false alarms and missed alarms caused by changes in meteorological conditions, and significantly improving the accuracy and scientific nature of early warning determination.

[0021] 4. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion described in this invention comprehensively utilizes three technical means—transmission trajectory tracking, air mass backward trajectory simulation, and regional contribution analysis—through a pollution source tracing and transmission analysis module to achieve precise location and quantitative attribution of pollution sources. The air mass backward trajectory simulation uses a three-dimensional kinematic trajectory model combined with high spatiotemporal resolution wind field data to accurately reconstruct the transmission path of polluted air masses over the past 72 hours. The regional contribution analysis uses a source-oriented attenuation model to decompose the pollution contribution of the monitoring area into three categories: local contribution, regional transmission contribution, and background contribution. This technical feature enables environmental management departments to clearly identify the primary and secondary responsible areas of pollution incidents, providing quantitative technical support for precise source tracing and differentiated control.

[0022] 5. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion described in this invention establishes an automatic matching mechanism between warning levels and response strategies through an intelligent early warning decision-making and release module. This module predefines four warning levels: blue, yellow, orange, and red, and configures corresponding response strategies for each level. Simultaneously, it automatically adjusts the focus of the response strategy based on the pollution cause type output by the chemical coupling analysis module. When the pollution cause is mainly direct emission, the response strategy focuses on controlling local pollution sources; when the pollution cause is mainly chemical transformation, the response strategy focuses on the synergistic reduction of ozone precursors. This technical feature achieves effective integration of early warning information and control measures, significantly improving the practicality and enforceability of early warning information, and providing closed-loop technical support for precise pollution control decisions by environmental management departments. Attached Figure Description

[0023] The invention will now be further described with reference to the accompanying drawings.

[0024] Figure 1 This is a structural block diagram of an intelligent air quality monitoring system for industrial parks based on multi-sensor fusion, as described in this invention.

[0025] Figure 2 This is a schematic diagram of the entire data flow and module collaborative interaction relationship from the multi-parameter sensor network to intelligent early warning decision-making and release in this invention. Detailed Implementation

[0026] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0027] Example 1:

[0028] This embodiment is applied to an intelligent monitoring and early warning decision-making scenario for atmospheric environmental quality in an industrial park. In this application scenario, a multi-parameter sensor monitoring network covering the entire industrial park is deployed to collect real-time concentration data of pollutants such as sulfur dioxide, nitrogen oxides, volatile organic compounds, ozone, and fine particulate matter at various points in the park, as well as meteorological parameter data such as temperature, humidity, wind speed, wind direction, and solar radiation. The collected multi-source heterogeneous data is preprocessed by an edge computing gateway and then aggregated to a data aggregation platform. The data aggregation platform sends the processed data to a chemical coupling mechanism analysis module, a dynamic threshold and anomaly identification module, and a pollution source tracing and transmission analysis module for in-depth analysis. The chemical coupling mechanism analysis module establishes a coupling analysis model based on the chemical reaction mechanism between pollutants to quantitatively describe the chemical transformation relationship between pollutants. The system comprises three modules: a dynamic threshold and anomaly identification module, which establishes a dynamic baseline based on historical monitoring data and determines the probability and magnitude of exceeding the dynamic baseline by using a probability model; a pollution source tracing and transmission analysis module, which tracks pollution sources and performs quantitative attribution based on air mass trajectory and regional contribution models; and an intelligent early warning decision-making and release module, which integrates multi-dimensional analysis results to generate early warning levels and response strategies and releases information. This system effectively overcomes the limitations of existing monitoring systems that can only perceive the superficial phenomenon of pollutant concentration, achieving a leap from understanding what concentration was measured to understanding why it is that concentration, and providing closed-loop technical support for precise pollution control decisions by environmental management departments.

[0029] like Figure 1 and Figure 2 As shown, this embodiment relates to an intelligent air quality monitoring system for industrial parks based on multi-sensor fusion. The overall architecture of the system includes eight core functional modules: a multi-parameter sensor network module, a data transmission and aggregation module, a chemical coupling mechanism analysis module, a dynamic threshold and anomaly identification module, a pollution source tracing and transmission analysis module, an intelligent early warning decision and release module, a data quality control module, and a three-dimensional visualization display module. The following describes this embodiment in detail from the aspects of system architecture construction and workflow deduction.

[0030] The multi-parameter sensor network module is located at the forefront of the entire monitoring system, undertaking the core task of real-time sensing and data acquisition of multiple atmospheric environmental parameters in the park. The internal structure of this module consists of three parts: a monitoring micro-station array, a multi-parameter sensor module, and a meteorological parameter sensor module. The deployment of the monitoring micro-station array follows the principle of high-density coverage, with a deployment density of 4 to 8 monitoring micro-stations per square kilometer. The horizontal spacing between adjacent monitoring micro-stations is controlled within the range of 300 to 500 meters to ensure that changes in pollutant concentrations at any location within the entire park can be effectively captured. The overall power consumption of the monitoring micro-station is no more than 5 watts, the mean time between failures (MTBF) is no less than 8760 hours, and the protection level is no less than IP65, enabling it to adapt to the complex and harsh atmospheric environmental conditions of the industrial park.

[0031] A multi-parameter sensor module is integrated into each monitoring micro-station to collect air pollutant concentration data in the industrial park. This module contains five gaseous pollutant sensors and one particulate matter concentration sensor. The five gaseous pollutant sensors detect sulfur dioxide, nitrogen oxides, volatile organic compounds (VOCs), ozone, and fine particulate matter (PM2.5). Each sensor employs a dual redundancy configuration of electrochemical and optical sensors to ensure data reliability and self-diagnostic capabilities. The sulfur dioxide concentration sensor has a detection range of 0 to 500 micrograms per cubic meter and a resolution of 0.1 micrograms per cubic meter. It uses ultraviolet fluorescence detection to quantitatively calculate the concentration by measuring the fluorescence intensity emitted by sulfur dioxide molecules after being excited by ultraviolet light. The nitrogen oxide concentration sensor has a detection range of 0 to 200 micrograms per cubic meter and a resolution of 0.05 micrograms per cubic meter. It uses a chemical... The luminescence detection principle quantitatively calculates the concentration value by measuring the chemiluminescence intensity generated by the reaction of nitric oxide and ozone. The volatile organic compound (VOC) concentration sensor adopts the photoionization detection principle, with a detection range of 0 to 10,000 micrograms per cubic meter and a resolution of 1 microgram per cubic meter. It quantitatively calculates the concentration value by measuring the ion current generated by the ionization of organic molecules under ultraviolet light irradiation. The ozone concentration sensor has a detection range of 0 to 400 micrograms per cubic meter and a resolution of 0.1 micrograms per cubic meter. It adopts the ultraviolet absorption detection principle, quantitatively calculating the concentration value by measuring the characteristic absorption intensity of ozone to 254 nanometer ultraviolet light. The fine particulate matter (PM2.5) concentration sensor adopts the laser scattering principle, with a detection range of 0 to 500 micrograms per cubic meter and a resolution of 0.1 micrograms per cubic meter. It inverts the number concentration and volume concentration of particulate matter by measuring the intensity distribution of scattered light from incident laser light.

[0032] Meteorological parameter sensor modules are also integrated into each monitoring micro-station to collect meteorological parameter data for the monitored area. This module is equipped with five meteorological parameter sensors: temperature, humidity, wind speed, wind direction, and solar radiation. The temperature sensor has a measurement range of -40 to 60 degrees Celsius and an accuracy of ±0.3 degrees Celsius. It uses the platinum resistance thermometry principle to accurately calculate the ambient temperature by measuring the change in the resistance of platinum with temperature. The humidity sensor has a measurement range of 0 to 100% and an accuracy of ±2%. It uses the capacitive polymer humidity measurement principle to accurately calculate the relative humidity by measuring the change in the dielectric constant of the humidity-sensitive capacitor. The wind speed sensor has a measurement range of 0 to 30 meters per second and an accuracy of ±0.2 meters per second. It uses the ultrasonic time-of-flight method to accurately calculate wind speed by measuring the time difference of ultrasonic waves under tailwind and headwind conditions. The wind direction sensor has a measurement range of 0 to 360 degrees and an accuracy of ±3 degrees. It uses the magnetoresistive wind direction measurement principle to accurately determine wind direction by measuring the angle of the wind vane relative to the Earth's magnetic north pole. The solar radiation sensor has a measurement range of 0 to 1400 watts per square meter and an accuracy of ±5 watts per square meter. It uses the thermopile detection principle to accurately calculate solar radiation intensity by measuring the thermoelectric potential generated after the thermopile absorbs solar radiation.

[0033] The sampling frequency of the monitoring micro-station is adjustable from 3 to 10 times per minute, with a default sampling frequency of 5 times per minute to ensure effective tracking of rapid changes in pollutant concentration. The sampled data undergoes analog-to-digital conversion within the analog-to-digital conversion module, with a 16-bit conversion accuracy to ensure precise acquisition of small-signal concentration changes. The raw data after analog-to-digital conversion enters the preliminary filtering unit, which uses a moving average filtering algorithm to remove high-frequency random noise. The filtering window width is set to 3 sampling points to balance the conflict between filtering effect and response speed. The filtered sampled data is packaged and encapsulated according to a preset data packet format. Each data packet contains five data fields: a unique identifier for the monitoring micro-station, a timestamp, the concentration values ​​of each pollutant, the values ​​of each meteorological parameter, and a checksum. The packaged data packet is sent to the data transmission and aggregation module through the communication interface.

[0034] The data transmission and aggregation module is located downstream of the multi-parameter sensor network module, undertaking the core tasks of data reception, preprocessing, aggregation, storage, and transmission forwarding. Its internal structure comprises three parts: an edge computing gateway array, a data aggregation platform, and a time-series database. The edge computing gateway array is deployed at a ratio of one edge computing gateway for every 8 to 16 monitoring micro-stations, ensuring balanced access load and controllable response latency for each gateway. The hardware configuration of the edge computing gateway includes two functional components: a data caching unit and a local analysis unit. The data caching unit has a storage capacity of no less than 32 gigabytes, using solid-state drives (SSDs) as the storage medium to meet high write bandwidth requirements. This unit temporarily stores raw sampling data from the connected monitoring micro-stations, providing data buffer protection to prevent data loss during network interruptions or transmission failures. The local analysis unit has an execution capability of 500 million floating-point operations per second and is equipped with an independent digital signal processor chip for performing real-time data quality detection and abnormal data marking tasks.

[0035] The data preprocessing process executed by the edge computing gateway includes four sequential steps: data filtering, data calibration, data completion, and data compression. The data filtering step uses a median filtering algorithm to remove spike noise from the sensor output. The filtering window width is set to 5 sampling points. The median filtering algorithm has a natural ability to suppress occasional extreme measurements and does not introduce filtering delay. The data calibration step calibrates and corrects the original concentration data based on calibration factors and a zero-point drift model. The calibration factors are pre-stored in the calibration parameter table of the edge computing gateway, and the zero-point drift model updates the calibration parameters in real time according to the aging trend of the sensor. The data completion step uses a linear interpolation algorithm to interpolate and supplement data at missing time points. When data at a certain time point is detected to be missing, the line connecting the two valid data points before and after that time point is taken as the interpolation value. The data compression step uses a compression algorithm based on time series characteristics to compress the original data to 30% to 40% of the original data volume. This algorithm utilizes the correlation of the same parameter between consecutive time points, storing only the differential changes to achieve efficient compression.

[0036] The edge computing gateway employs a dual-channel parallel transmission architecture of Time Division Duplex (TDD) Long Term Evolution (LTE) and Low Power Wide Area Network (LPWAN) to send data to the upper-level data aggregation platform. The LTE channel serves as the primary transmission channel, providing high-bandwidth, low-latency data transmission services with a latency of no more than 2 seconds and a reliability of no less than 99.5%. The LPWAN channel serves as a backup transmission channel, automatically switching to and activating when the LTE channel fails, ensuring redundant reliability of the data transmission link. The edge computing gateway supports three protocol formats: Transmission Control Protocol (TCP), User Datagram Protocol (UDP), and Secure Hypertext Transfer Protocol (STP). These can be flexibly configured according to the interface requirements of the upper-layer system. During data transmission, data compression and breakpoint resumption mechanisms are enabled to ensure the integrity and reliability of data transmission. When a transmission interruption is detected, the edge computing gateway automatically retains the uncompleted data in its local buffer and continues transmission after transmission is restored.

[0037] The data aggregation platform deploys a time-series database to store all sensor data. The time-series database adopts a distributed architecture with three data replica nodes to ensure high availability of data storage. The time-series database has a write capacity of no less than 100,000 records per second and a query capacity of no less than 50,000 aggregate queries per second, which can meet the storage performance requirements of continuous writing of high-frequency sampling data from large-scale monitoring micro-station arrays. The data aggregation platform also undertakes the task of distributing the aggregated data to downstream analysis modules. Data distribution adopts a message queue mechanism to achieve asynchronous and decoupled transmission. Each downstream analysis module obtains the required data from its subscribed message topics.

[0038] The chemical coupling mechanism analysis module is located downstream of the data transmission and aggregation module. It undertakes the core task of establishing a model of the chemical coupling relationship between pollutants and quantitatively calculating the coupling strength and chemical transformation contribution. The internal structure of this module includes four functional components: coupling analysis engine, photochemical reaction path modeling unit, oxidation reaction path modeling unit, and gas-particle conversion path modeling unit. The coupling analysis engine is the core scheduling unit of the entire module. It is responsible for coordinating the model calculation process of the three modeling sub-units and integrating the output results of each sub-unit to generate the final coupling analysis report.

[0039] The photochemical reaction pathway modeling unit constructs a nonlinear response model between changes in volatile organic compound (VOC) and nitrogen oxide (NOx) concentrations and the ozone generation rate based on the standard EKMA curve for ozone formation. This unit integrates two sub-models: a photolysis rate calculation model and a free radical reaction kinetic model. The photolysis rate calculation model calculates the rate constants of various photochemical reactions based on solar radiation intensity, solar zenith angle, and ozone column concentration. The solar zenith angle is calculated using astronomical algorithms based on the geographical coordinates of the monitoring microstation and the current time. The free radical reaction kinetic model establishes a free radical oxidation reaction network based on the reaction rate constants of hydroxyl radicals, nitrate radicals, and ozone with VOCs. This network includes three typical oxidation pathways for VOCs: alkane oxidation, olefin oxidation, and aromatic hydrocarbon oxidation. The photochemical reaction pathway modeling unit outputs a sensitivity matrix between the ozone generation rate and precursor concentration. The elements of the sensitivity matrix represent the change in ozone generation rate caused by a unit change in concentration.

[0040] The oxidation reaction pathway modeling unit establishes a calculation model for sulfate formation rate based on the oxidation reaction mechanism of sulfur dioxide and hydroxyl radicals. This unit uses a pseudo-first-order reaction kinetic equation to describe the oxidation process of sulfur dioxide. The oxidation reaction rate constant is affected by three factors: hydroxyl radical concentration, temperature, and relative humidity. The hydroxyl radical concentration is estimated using empirical formulas for solar radiation intensity and nitrogen oxide concentration. Temperature and relative humidity data are acquired in real time from the meteorological parameter sensor module. The oxidation reaction pathway modeling unit outputs time series predictions of sulfate formation rate and sulfate concentration.

[0041] The gas-particle conversion pathway modeling unit establishes a calculation model for the distribution ratio of volatile organic compounds (VOCs) between the gas and particulate phases based on the gas-particle distribution equilibrium model. This unit uses an eight-phase parallel adsorption isotherm model to describe the dynamic distribution process of VOCs between the gas and particulate phases. The model input parameters include the saturated vapor pressure of VOCs, particulate concentration, temperature, and organic matter content. The gas-particle conversion pathway modeling unit outputs the distribution ratio of various VOCs between the gas and particulate phases and the predicted concentration of VOCs in the particulate phase.

[0042] The chemical transformation contribution decomposition algorithm of the coupling analysis engine decomposes the observed concentrations of each pollutant into two parts: direct emission contribution and chemical transformation contribution. The decomposition process adopts a constrained optimization algorithm based on chemical mass balance. The material conservation relationship of various chemical reaction paths is used as the constraint to construct the optimization objective function. The optimization objective function is defined as minimizing the sum of squared residuals between the observed concentration and the decomposed concentration. The constraints include the material conservation equations of each chemical transformation path and the non-negativity constraint of direct emission contribution. The coupling analysis engine uses a sequential quadratic programming algorithm to solve this constrained optimization problem to obtain the optimal decomposition results of direct emission contribution and chemical transformation contribution. The output results of the coupling analysis engine include the coupling strength coefficient between various pollutants and the proportion of chemical transformation contribution. The coupling strength coefficient quantitatively represents the degree of correlation between the concentration changes of the two pollutants, and the proportion of chemical transformation contribution represents the proportion of the observed concentration generated by the chemical transformation process.

[0043] The coupled analysis engine is also equipped with an ozone formation sensitivity analysis function. This function determines which ozone formation sensitivity zone the current atmospheric environment is in by calculating the boundary conditions of the volatile organic compound (VOC) sensitive zone and the nitrogen oxide (NOx) sensitive zone. The sensitivity analysis is based on the EKMA curve model to calculate the partial derivatives of the ozone formation rate with respect to the concentrations of VOCs and NOxes. When the partial derivative of the ozone formation rate with respect to the concentration of VOCs is greater than the partial derivative with respect to the concentration of NOxes, it is determined that the current environment is in the VOC control zone; otherwise, it is determined to be in the NOx control zone.

[0044] The dynamic threshold and anomaly identification module is located downstream of the data transmission and aggregation module. It undertakes the core task of establishing a dynamic baseline concentration range and determining real-time concentration anomalies. The internal structure of this module includes two functional components: a dynamic threshold calculation unit and an exceedance judgment unit. The dynamic threshold calculation unit calculates the dynamic baseline concentration range based on the historical data sequence of pollutant concentrations at each monitoring point. The dynamic baseline concentration range includes two time scales: hourly baseline range and daily baseline range. The hourly baseline range is calculated based on historical data from the same period within the last 30 days. The calculation process includes calculating the mean and standard deviation of the historical data, setting the upper limit of the baseline range to the mean plus twice the standard deviation, and setting the lower limit of the baseline range to the mean minus twice the standard deviation. The daily baseline range is calculated based on historical data from the last 90 days. The calculation method is the same as that of the hourly baseline range, but the statistical time window is longer. The hourly baseline range is used to identify short-term concentration fluctuation anomalies, and the daily baseline range is used to identify persistent concentration rise anomalies.

[0045] The dynamic threshold calculation unit is also equipped with a meteorological condition adaptive adjustment function. This function dynamically corrects the dynamic baseline concentration range according to the current meteorological parameters to eliminate the interference of meteorological condition changes on the baseline. When the temperature is higher than 30 degrees Celsius, the baseline range of volatile organic compounds is adjusted upward by 15% to 25% of the historical baseline average, because high temperature conditions promote the volatilization and release of volatile organic compounds, resulting in a natural increase in background concentration. When the solar radiation intensity is higher than 800 watts per square meter, the baseline range of ozone is adjusted upward by 20% to 30% of the historical baseline average, because strong radiation conditions promote photochemical reactions to generate ozone, resulting in a natural increase in ozone background concentration.

[0046] The exceedance determination unit calculates the exceedance probability and exceedance magnitude of the real-time concentration at each monitoring point relative to the dynamic baseline. The exceedance probability is calculated using a Bayesian inference-based probability model. This model uses the historical concentration distribution as the prior distribution and the real-time concentration observation value as the likelihood function input. The posterior distribution is calculated using the Bayesian formula, and the probability value of the real-time concentration exceeding the upper limit of the baseline is extracted from it. The exceedance magnitude is calculated using the standardized deviation value. The calculation formula is to divide the difference between the real-time concentration and the mean of the dynamic baseline by the standard deviation of the dynamic baseline to obtain the dimensionless deviation multiple. The exceedance probability calculation result output by the exceedance determination unit is used to distinguish between normal fluctuations and abnormal events. When the exceedance probability is less than 50%, it is judged as normal fluctuation. When the exceedance probability is between 50% and 80%, it is judged as mild abnormality. When the exceedance probability is between 80% and 95%, it is judged as moderate abnormality. When the exceedance probability is greater than 95%, it is judged as severe abnormality. The exceedance determination unit also calculates the exceedance duration as an auxiliary judgment indicator. When the concentration exceeds the dynamic baseline for more than 30 minutes, event recording is triggered and the relevant data is saved to the event database.

[0047] The pollution source tracing and transmission analysis module is located downstream of the data transmission and aggregation module. It undertakes the core task of tracing pollution sources and quantitatively attributing them. The internal structure of this module includes three functional components: a transmission trajectory tracking unit, an air mass backward trajectory simulation unit, and a regional contribution analysis unit.

[0048] The transmission trajectory tracking unit calculates the propagation direction and speed of pollutants based on the spatial topology of the monitoring network and the concentration gradient changes at each monitoring point. The unit maintains a spatial topology map of the park's monitoring network, where nodes represent each monitoring micro-station and edges represent the spatial adjacency between adjacent monitoring micro-stations. The unit calculates the concentration gradient between adjacent monitoring micro-stations, with the concentration gradient pointing in the direction of increasing concentration, i.e., the upwind direction of pollutant propagation. The unit combines real-time wind speed and direction data to correct the concentration gradient direction and obtain a comprehensive propagation direction estimate. The comprehensive propagation speed is obtained by dividing the magnitude of the concentration gradient by the spatial distance.

[0049] The air mass backward trajectory simulation unit uses a Lagrange air mass trajectory model combined with wind field data and mixing layer height data to simulate the transport trajectory of pollutant air masses over the past 72 hours. The unit is equipped with a three-dimensional kinematic trajectory model. The model input data includes wind field data, mixing layer height data, and pollutant concentration data. The wind field data has a temporal resolution of 1 hour and a spatial resolution of at least 1 km x 1 km, sourced from numerical weather prediction products provided by meteorological departments. The mixing layer height data also comes from numerical weather prediction products, reflecting the vertical mixing range of the atmospheric boundary layer. The air mass trajectory calculation uses a three-dimensional kinematic model. The tic method obtains the displacement trajectory of the air mass by numerically integrating the flow field. The integration time step is set to 10 minutes to balance computational accuracy and efficiency. The simulated trajectory time span is 72 hours, and the trajectory output interval is 1 hour. The unit is configured with multiple starting point trajectory set generation functions. Each trajectory set contains 50 backward trajectories starting from the target monitoring point at the time of pollution occurrence. The starting heights of the 50 trajectories are set to five height layers: 10 meters, 50 meters, 100 meters, 200 meters, and 500 meters. Ten trajectories are generated in a fan-shaped distribution with an angle interval of 3.6 degrees to ensure the transmission path covers different air layers.

[0050] The regional contribution analysis unit uses a source-oriented attenuation model to estimate the regional contribution ratio of various pollution sources. The source-oriented attenuation model calculates the diffusion attenuation coefficient of the pollution source based on the downwind distance and wind speed. The diffusion attenuation coefficient is inversely proportional to the square of the downwind distance and directly proportional to the wind speed. The unit divides the monitoring area into three source areas: local contribution area, regional transmission area, and background contribution area. The local contribution area is defined as the area with a downwind distance of less than 20 kilometers, the regional transmission area is defined as the area with a downwind distance of 20 to 100 kilometers, and the background contribution area is defined as the area with a downwind distance of more than 100 kilometers. The unit integrates the propagation direction estimation output by the transmission trajectory tracking unit and the trajectory landing point distribution output by the air mass backward trajectory simulation unit to calculate the contribution ratio of each source area to the pollution concentration in this monitoring area.

[0051] The intelligent early warning decision and release module is located downstream of the dynamic threshold and anomaly identification module and the pollution source tracing and transmission analysis module. It undertakes the core task of generating early warning decisions and releasing information based on the comprehensive multi-dimensional analysis results. The internal structure of this module includes two functional components: a decision generation engine and a response strategy library.

[0052] The decision generation engine performs multi-level early warning assessments, with four levels: blue, yellow, orange, and red. A blue warning is triggered when the concentration of any pollutant exceeds the dynamic baseline upper limit with a probability between 50% and 80%. A yellow warning is triggered when the concentration of any pollutant exceeds the dynamic baseline upper limit with a probability between 80% and 95%, or when ozone concentration exceeds 80% of the environmental quality standard concentration limit. An orange warning is triggered when the concentration of any pollutant exceeds the dynamic baseline upper limit with a probability greater than 95%, or when ozone concentration exceeds the environmental quality standard concentration limit. A red warning is triggered when the concentration of any pollutant exceeds the dynamic baseline upper limit with a probability greater than 95%, and pollution source analysis indicates a significant regional transport contribution, or when ozone concentration exceeds 150% of the environmental quality standard concentration limit. The decision generation engine comprehensively considers the pollutant concentration exceeding the limit, the type of pollution cause, pollution transport characteristics, and the contribution ratio of each pollution source to generate the warning level and response strategy.

[0053] The response strategy library predefines response strategies corresponding to different warning levels and different pollution causes. The strategy for a blue warning is to increase monitoring frequency and initiate data verification; the strategy for a yellow warning is to send emission reduction recommendations to key pollution sources; the strategy for an orange warning is to initiate emergency emission reduction response and conduct pollution source investigation; and the strategy for a red warning is to initiate the highest level emergency response and implement regional joint prevention and control. The decision generation engine automatically adjusts the focus of the response strategy based on the pollution cause type output by the chemical coupling mechanism analysis module. When the pollution cause is mainly direct emission, the response strategy focuses on local pollution source control; when the pollution cause is mainly chemical transformation, the response strategy focuses on the coordinated emission reduction of ozone precursors. The intelligent warning decision and release module is connected to the data transmission and aggregation module through a communication interface, sending the generated warning information and response strategies to the data aggregation platform and then forwarding them to the business system of the environmental management department.

[0054] The data quality control module connects to the data transmission and aggregation module and undertakes the core task of multi-dimensional quality verification and evaluation of the raw sampling data. Its internal structure includes three functional components: a data integrity verification unit, a spatiotemporal consistency verification unit, and a physical rationality verification unit. The data integrity verification unit checks the continuity of the timestamps and the integrity of the numerical fields in each data record. When a missing timestamp or an empty numerical field is detected, the record is marked as suspicious data. The spatiotemporal consistency verification unit checks whether there is a significant and unreasonable spatial gradient in the concentration data of adjacent monitoring micro-stations at the same time. When the spatial gradient exceeds three times the standard deviation of a preset threshold, the relevant data is marked as suspicious data. The physical rationality verification unit checks whether the concentration data of each pollutant conforms to the physical conservation law, including the negative correlation between nitrogen oxides and ozone concentrations and the positive correlation between sulfur dioxide and particulate matter sulfate concentrations. When an abnormal correlation is detected, the relevant data is marked as suspicious data. The data quality report output by the data quality control module includes the data integrity rate, data availability rate, and data quality score for each monitoring micro-station.

[0055] The 3D visualization module communicates with the data transmission and aggregation module, the chemical coupling mechanism analysis module, the dynamic threshold and anomaly identification module, and the pollution source tracing and transmission analysis module. Its core task is to present multi-dimensional monitoring data and analysis results in an intuitive 3D visualization manner. The module's internal structure includes four functional components: a park 3D scene construction unit, a pollutant concentration field rendering unit, a pollution transmission dynamic demonstration unit, and an early warning information overlay display unit. The park 3D scene construction unit constructs a 3D scene model of the park based on the park's floor plan and building height data. The spatial reference coordinate system of the 3D scene model is consistent with the geographic coordinate system of the monitoring network. The pollutant concentration field rendering unit... The pollution unit generates a three-dimensional color cloud map of pollutant concentration distribution over the park based on real-time concentration data from monitoring micro-stations and spatial interpolation algorithms. The spatial interpolation algorithm uses the Kriging interpolation method to generate a continuous and smooth concentration isosurface. The pollution transport dynamic demonstration unit dynamically displays the transport path and diffusion range of polluted air masses in a three-dimensional scene based on the back-trajectory simulation results of the air mass. The transport path extends from the starting point to the end point in the form of a strip animation, and the diffusion range gradually unfolds in the form of a semi-transparent cone. The early warning information overlay display unit overlays and displays the early warning status icons of each monitoring micro-station and the contribution ratio heat map of each pollution source area in a three-dimensional scene. The early warning status icons display different colors and shapes according to the early warning level.

[0056] The workflow of the intelligent air quality monitoring system for industrial parks based on multi-sensor fusion in this embodiment is as follows:

[0057] Multi-parameter synchronous sampling and environmental data acquisition: The monitoring micro-station array in the multi-parameter sensor network module synchronously starts the sampling task. The multi-parameter sensor modules and meteorological parameter sensor modules of each monitoring micro-station perform parallel sampling according to the set frequency. During the sampling process, each sensor converts the physical quantity of the concentration of target pollutants in the atmosphere into the corresponding analog electrical signal. After amplification and filtering, the electrical signal enters the analog-to-digital conversion module to be converted into a digital signal. After preliminary filtering to remove spike noise, the digital signal is encapsulated according to the data packet format. The packaged data packet is sent to the corresponding edge computing gateway through the communication interface via the Time Division Duplex Long Term Evolution Network or Low Power Wide Area Network. The edge computing gateway receives the data packet from the connected monitoring micro-station and performs verification processing. The data packet that passes the verification enters the data preprocessing process and sequentially performs median filtering to remove spike noise, calibration factor correction, linear interpolation to complete missing data points, and compression processing based on time series features.

[0058] Data aggregation and quality control verification: The edge computing gateway sends preprocessed data to the data aggregation platform via a dual-channel parallel transmission architecture. The time-series database of the data aggregation platform receives and writes all sensor data. During the writing process, the quality verification process of the data quality control module is triggered. The data integrity verification unit traverses each data record to check the continuity of timestamps and the integrity of numerical fields. When a discontinuous timestamp or an empty numerical field is detected, the record is marked as suspicious data and the anomaly type is recorded. The spatiotemporal consistency verification unit reads the concentration data of adjacent monitoring micro-stations at the same time and performs spatial gradient calculation. When the spatial gradient exceeds three times the standard deviation of the preset threshold, the records of the relevant micro-stations are marked as suspicious data. The physical rationality verification unit extracts the concentration time series data of each pollutant and calculates the correlation coefficient between nitrogen oxides and ozone and the correlation coefficient between sulfur dioxide and sulfate. When the correlation coefficient deviates from the preset reasonable range, the records of the corresponding time period are marked as suspicious data. The data quality control module summarizes the suspicious data marking of each monitoring micro-station, calculates the data integrity rate, data availability rate, and data quality score, and generates a data quality report.

[0059] Chemical coupling mechanism analysis and contribution decomposition: The data aggregation platform distributes the aggregated data to the chemical coupling mechanism analysis module to initiate the chemical coupling analysis process. The coupling analysis engine first calls the photochemical reaction path modeling unit to read the current concentrations of volatile organic compounds (VOCs), nitrogen oxides (NOx), and solar radiation intensity. The photolysis rate calculation sub-model of the photochemical reaction path modeling unit calculates the rate constants of various photochemical reactions based on solar radiation intensity and solar zenith angle. The solar zenith angle is calculated using astronomical algorithms based on the geographical coordinates of the monitoring micro-station and the current time. The free radical reaction kinetics sub-model establishes a free radical oxidation reaction network based on the reaction rate constants of hydroxyl radicals and VOCs and calculates the ozone generation rate. The photochemical reaction path modeling unit outputs a sensitivity matrix between the ozone generation rate and precursor concentration. The chemical reaction path modeling unit reads the current sulfur dioxide concentration, temperature, and relative humidity data, and calculates the sulfate formation rate based on the empirical estimate of the hydroxyl radical concentration. The gas-particle conversion path modeling unit reads the current volatile organic compound concentration, particulate matter concentration, and temperature data, and calculates the distribution ratio of various volatile organic compounds between the gas and particulate phases based on the eight-phase parallel adsorption isotherm model. The coupling analysis engine summarizes the output results of the three modeling units and executes the chemical conversion contribution decomposition algorithm. The algorithm uses a sequential quadratic programming algorithm to solve an optimization problem constrained by material conservation to obtain the decomposition results of direct emission contribution and chemical conversion contribution. The decomposition results include a coupling strength coefficient matrix and a chemical conversion contribution ratio vector. The coupling analysis engine assembles the decomposition results into a coupling analysis report and sends it to the intelligent early warning decision and release module.

[0060] The dynamic baseline calculation and anomaly detection process involves the data aggregation platform simultaneously distributing data to the dynamic threshold and anomaly identification modules to initiate the dynamic baseline calculation and anomaly detection procedures. The dynamic threshold calculation unit queries historical concentration data sequences for each monitoring point from the time-series database and calculates the hourly and daily baseline ranges respectively. For the hourly baseline range, it queries historical data from the same time period within the last 30 days, calculates the mean and standard deviation, and sets the upper limit of the baseline range to the mean plus twice the standard deviation and the lower limit to the mean minus twice the standard deviation. The daily baseline range queries historical data from the last 90 days and calculates it using the same method. Simultaneously, the dynamic threshold calculation unit reads current meteorological parameter data and adjusts the baseline ranges of relevant pollutants upwards according to the adaptive adjustment rules based on meteorological conditions. After receiving the dynamic baseline calculation results, the exceedance determination unit reads the real-time concentration data of each monitoring point to calculate the exceedance probability and exceedance magnitude relative to the dynamic baseline. The exceedance probability calculation adopts the Bayesian inference method, using the historical concentration distribution as the prior distribution and the real-time observation value as the likelihood function input to calculate the posterior distribution and extract the probability value of exceeding the upper limit of the baseline. The exceedance magnitude calculation adopts the standardized deviation formula, dividing the difference between the real-time concentration and the mean of the dynamic baseline by the standard deviation of the dynamic baseline. The exceedance determination unit classifies and determines the current anomaly level according to the exceedance probability threshold and calculates the exceedance duration. When the duration exceeds 30 minutes, an event record is triggered. The dynamic threshold and anomaly identification module assembles the exceedance determination results into an anomaly analysis report and sends it to the intelligent early warning decision and release module.

[0061] The pollution transmission trajectory tracking and regional contribution analysis are performed simultaneously. The data aggregation platform distributes data to the pollution source tracing and transmission analysis module to initiate the pollution source tracing and analysis process. The transmission trajectory tracking unit reads the spatial topology map of the monitoring network and the concentration gradient data of adjacent monitoring micro-stations to calculate the comprehensive propagation direction and speed of pollutants. The comprehensive propagation direction is corrected by combining real-time wind direction data. The air mass backward trajectory simulation unit reads wind field data, mixing layer height data, and pollutant concentration data to perform three-dimensional kinematic trajectory calculation. The trajectory calculation process integrates along the reverse direction of time for 72 hours with the target monitoring point as the starting point, generating 50 backward trajectories. Each trajectory records the position coordinates of each hour. The regional contribution analysis unit reads the transmission trajectory tracking results and the air mass backward trajectory simulation results and uses a source-oriented attenuation model to estimate the regional contribution ratio of various pollution sources. The model calculates the diffusion attenuation coefficient based on the downwind distance and wind speed of the pollution source, and then calculates the contribution ratio of the three source areas—local contribution area, regional transmission area, and background contribution area—based on the comprehensive distribution of the landing points of the transmission trajectories. The pollution source tracing and transmission analysis module assembles the source tracing and analysis results into a source tracing and analysis report and sends it to the intelligent early warning decision and release module.

[0062] The multi-level early warning decision generation and information dissemination module's decision generation engine receives coupling analysis reports from the chemical coupling mechanism analysis module, anomaly analysis reports from the dynamic threshold and anomaly identification module, and source tracing analysis reports from the pollution source tracing and transmission analysis module, initiating a multi-level early warning assessment process. The decision generation engine first reads the exceedance probability and exceedance magnitude data from the anomaly analysis reports and determines the warning level based on the triggering conditions for blue, yellow, orange, and red warnings. Simultaneously, the decision generation engine reads the chemical transformation contribution ratio from the coupling analysis reports and the regional contribution ratio from the source tracing analysis reports as supplementary information for early warning decision-making. As a reference, when the warning level is determined to be yellow or above, the decision generation engine retrieves the corresponding warning level strategy content from the response strategy library and automatically adjusts the strategy focus according to the pollution cause type. The decision generation engine assembles the warning level, strategy content, and auxiliary reference data into a warning decision message and sends it to the data aggregation platform through the communication interface. The data aggregation platform forwards the warning decision message to the business system of the environmental management department and the 3D visualization display module. The warning information overlay display unit of the 3D visualization display module overlays and displays the warning status icon and the source area contribution heat map in the 3D scene of the park, so that monitoring personnel can intuitively understand the warning information and the distribution of pollution sources.

[0063] Example 2:

[0064] This embodiment provides a variant design based on Embodiment 1, mainly focusing on improving the model accuracy and expanding the coupling analysis function of the chemical coupling mechanism analysis module. In this variant design, a gas-phase free radical chemical mechanism model is added to the photochemical reaction path modeling unit. This model is based on complete aqueous and gas-phase chemical mechanisms to construct an atmospheric chemical mechanism model containing more than 200 chemical reaction paths. The atmospheric chemical mechanism model classifies volatile organic compounds into five categories according to carbon bond structure type: alkanes, alkenes, alkynes, aromatics, and oxygen-containing organic compounds, and establishes oxidation reaction paths for each category. The ozone generation EKMA curve calculation of this model uses an orthogonal grid method to search for the ozone generation peak position in the two-dimensional parameter space composed of volatile organic compound concentration and nitrogen oxide concentration, and calculates the offset direction and offset amount of the current concentration point relative to the peak.

[0065] Regarding the oxidation reaction pathway modeling unit, this variant design expands the model coverage of the unit. In addition to sulfate, it also adds nitrate and ammonium salt generation pathway modeling. The nitrate generation pathway modeling is based on two pathways: the oxidation reaction of nitrogen dioxide with hydroxyl radicals and the reaction of ammonia with nitric acid. The ammonium salt generation pathway modeling is based on the gaseous reaction of ammonia with sulfuric acid and nitric acid. The output of the oxidation reaction pathway modeling unit is expanded to include a complete report containing the generation rates of three types of secondary inorganic aerosols: sulfate, nitrate, and ammonium salt.

[0066] In terms of the gas-particle conversion pathway modeling unit, this variant design adds a secondary organic aerosol generation model. The secondary organic aerosol generation model calculates the generation potential of secondary organic aerosols based on the atmospheric aging process of volatile organic compound oxidation products. The model uses the product volatility distribution function to describe the volatility characteristics of oxidation products and determines the proportion of oxidation products entering the particulate phase based on the volatility threshold. The output results of the secondary organic aerosol generation model include the secondary organic aerosol generation rate, generation amount, and volatility distribution histogram.

[0067] Regarding the coupling analysis engine, this variant design adds a multi-pollutant synergistic sensitivity analysis function. This function calculates the joint sensitivity of ozone generation rate to volatile organic compound and nitrogen oxide concentrations by constructing an ozone isoconcentration line grid in a two-dimensional precursor concentration space. The results of the joint sensitivity analysis are used to identify the synergistic control zone and unidirectional control zone of ozone generation control, providing more refined technical support for the formulation of precise control strategies.

[0068] Example 3:

[0069] This embodiment provides another variant design based on Embodiment 1, mainly focusing on expanding the coverage of the monitoring network and improving the accuracy of source tracing analysis. In this variant design, the multi-parameter sensor network module expands the functional configuration of the monitoring micro-station by adding two types of mobile monitoring terminals: UAV-borne mobile monitoring units and vehicle-mounted mobile monitoring units. The UAV-borne mobile monitoring unit is integrated into a multi-rotor UAV platform and transmits monitoring data back in real time via a 4G network, enabling it to perform rapid mobile monitoring tasks at factory boundaries and pollution accident sites that are difficult for humans to reach. The vehicle-mounted mobile monitoring unit is integrated into an environmental monitoring vehicle and is powered by the vehicle's power supply, performing air quality monitoring along roadsides while the vehicle is in motion, effectively expanding the spatial coverage of the monitoring network.

[0070] In terms of pollution source tracing and transmission analysis modules, this variant design expands the model configuration of the air mass backward trajectory simulation unit and adds an ensemble trajectory simulation function. The ensemble trajectory simulation function introduces the uncertainty of weather forecasts on the basis of traditional deterministic trajectory simulation. By running multiple weather model members, it generates multiple sets of trajectory ensembles to statistically evaluate the probability distribution of the air mass source direction. The time span of the ensemble trajectory simulation is extended to 120 hours to meet the tracking needs of long-distance cross-regional transmission events, and the trajectory output interval is shortened to 15 minutes to meet the fine tracking needs of short-term sudden pollution events.

[0071] Regarding the regional contribution analysis unit, this variant design adds a source apportionment forward model function. The source apportionment forward model inverts the emission intensity of each pollution source based on the observed pollutant concentration at the receptor point. This model uses the Bayesian source apportionment method, with the prior emission inventory of the pollution source as prior information and the observed concentration at the receptor point as the likelihood function input. It calculates the posterior probability distribution of the emission intensity of each pollution source through the Markov chain Monte Carlo sampling method. The output of the source apportionment forward model includes point estimates of the contribution ratio of each pollution source and confidence intervals, which can quantify the uncertainty level of the source apportionment results.

[0072] Comparative example:

[0073] To objectively evaluate the technical effects of the various embodiments of the present invention, a comparative example of the prior art is provided below for comparison and analysis. This comparative example adopts the technical architecture of a traditional industrial park air quality monitoring system. In this architecture, the sensor network only collects the concentration data of each pollutant without performing multi-parameter collaborative sampling and synchronous collection of meteorological parameters. The data transmission module adopts a single-channel transmission architecture and does not have edge computing and data preprocessing capabilities. The monitoring system does not configure a chemical coupling mechanism analysis module, a dynamic threshold and anomaly identification module, and a pollution source tracing and transmission analysis module. It only configures a simple threshold comparison alarm function. The early warning decision adopts a static threshold judgment mode. When the pollutant concentration exceeds a preset fixed threshold, the corresponding level of early warning is triggered.

[0074] Regarding the ability to analyze the causes of pollution, the comparative model, lacking a chemical coupling mechanism analysis module, cannot establish a model of the chemical coupling relationship between pollutants and can only provide superficial data on pollution concentration. When ozone concentration is abnormal, the comparative model cannot determine whether the exceedance is due to the direct emission of ozone precursors locally or to the full photochemical reaction of volatile organic compounds and nitrogen oxides transported from upstream. Experimental data shows that in the comparative model's monitoring data, about 35% of ozone exceedance events are due to insufficiently reacted exceedances caused by direct local emissions, about 45% are due to fully reacted exceedances caused by regional transport, and about 20% are mixed exceedances. The comparative model cannot distinguish between these three types of exceedance causes, resulting in a lack of targeted early warning and response strategies. Embodiment 1 of this invention, through the chemical transformation contribution decomposition algorithm of the chemical coupling mechanism analysis module, can accurately distinguish between direct emission contributions and chemical transformation contributions, improving the accuracy of exceedance cause judgment to over 90%.

[0075] Regarding the accuracy of early warning judgment, the comparative model, due to its use of a static threshold judgment mode, is prone to false alarms when changes in meteorological conditions cause a natural increase in the background concentration of pollutants, and prone to missed alarms when pollutants slowly accumulate and have not yet reached a fixed threshold. Experimental data shows that the false alarm rate of the comparative model is as high as 25% to 30% during the high-temperature period in summer, and the missed alarm rate is as high as 15% to 20% during the rainy weather period. Embodiment 1 of the present invention, by combining dynamic threshold and dynamic baseline calculation of the anomaly identification module with Bayesian inference probability model, comprehensively considers the impact of changes in time scale and meteorological conditions on baseline correction, and controls the false alarm rate to within 5% and the missed alarm rate to within 3%.

[0076] Regarding the accuracy of pollution source tracing, the comparative model, lacking transmission trajectory tracking and air mass backward trajectory simulation functions, can only roughly determine the direction of pollution source based on the spatial distribution of the monitoring network and cannot provide a quantitative regional contribution ratio. Experimental data shows that the accuracy rate of the comparative model in determining the direction of pollution source is only 60% to 70%, and it cannot provide a quantitative estimate of the contribution ratio. Embodiment 1 of this invention improves the accuracy rate of determining the direction of pollution source to over 92% by using three technical means in the pollution source tracing and transmission analysis module: transmission trajectory tracking, air mass backward trajectory simulation, and regional contribution analysis, and controls the quantitative estimation error of the regional contribution ratio to within 10%.

[0077] Regarding the effectiveness of early warning decisions, comparative studies show that the lack of correlation between early warning strategies and pollution causes leads to a lack of targeted early warning response measures. Experimental data shows that after the early warning information reaches the environmental management department, the management department needs to spend an additional 2 to 4 hours to analyze the pollution causes before it can formulate targeted control measures. Embodiment 1 of this invention, through the strategy library pre-definition and cause correlation adjustment mechanism of the intelligent early warning decision and release module, provides targeted response strategy content while generating early warning information, thus shortening the pollution control response time to within 30 minutes.

[0078] Table 1: Comparison of Embodiments of the Invention with Comparative Examples

[0079] Comparison indicators Comparative Example Example 1 Example 2 Example 3 Accuracy of determining the cause of exceeding the standard 0% 90%-93% 93%-96% 91%-94% Chemical transformation contributes to decomposition error Unable to decompose <8% <6% <7% False alarm rate during hot summer periods 25%-30% <5% <4% <4.5% underreporting rate during rainy weather 15%-20% <3% <2.5% <2.8% Accuracy of pollution source direction identification 60%-70% 92%-95% 93%-96% 94%-97% Quantitative estimation error of regional contribution ratio Unable to quantify <10% <8% <9% Early warning response strategy response time 2-4 hours <30 minutes <25 minutes <28 minutes Data quality score 65-75 points 90-95 points 92-96 points 91-95 points Monitoring network space coverage 70% 95% 95% 98%

[0080] Analysis and summary of the above comparative data show that the embodiments of the present invention are significantly superior to the comparative examples of existing technologies in core performance indicators such as pollution cause analysis capability, early warning judgment accuracy, pollution source tracing accuracy, and early warning decision effectiveness. Embodiment 1 of the present invention establishes a chemical coupling sensing model between pollutants through a chemical coupling mechanism analysis module, fundamentally breaking through the limitation of traditional monitoring systems that can only sense the superficial concentration, and realizing a leap from what concentration was measured to why it is this concentration. Embodiment 2 further improves the model accuracy and functional integrity of chemical coupling analysis by expanding the coverage of the atmospheric chemical mechanism model and adding a secondary organic aerosol generation model. Embodiment 3 significantly expands the spatial coverage of the monitoring network and the temporal accuracy of source tracing analysis by introducing a mobile monitoring terminal and a ensemble trajectory simulation function. The coordinated operation of all functional modules throughout the process ensures that each embodiment can maintain stable and reliable monitoring performance and early warning decision quality in different industrial park monitoring scenarios.

[0081] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent air quality monitoring system for industrial parks based on multi-sensor fusion, characterized in that, It includes a multi-parameter sensor network module, a data transmission and aggregation module, a chemical coupling mechanism analysis module, a dynamic threshold and anomaly identification module, a pollution source tracing and transmission analysis module, and an intelligent early warning decision and release module; The multi-parameter sensor network module is connected to the data transmission and aggregation module via a communication network; The data transmission and aggregation module is connected to the chemical coupling mechanism analysis module; The chemical coupling mechanism analysis module is communicatively connected to the dynamic threshold and anomaly identification module and the pollution source tracing and transmission analysis module, respectively. The dynamic threshold and anomaly identification module is connected to the intelligent early warning decision and release module; The pollution source tracing and transmission analysis module is connected to the intelligent early warning decision and release module; The intelligent early warning decision-making and release module is communicatively connected to the data transmission and aggregation module.

2. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion according to claim 1, characterized in that, The multi-parameter sensor network module is internally configured with a multi-parameter sensor module and a meteorological parameter sensor module. The multi-parameter sensor module is used to collect atmospheric pollutant concentration data in the industrial park, and the meteorological parameter sensor module is used to collect meteorological parameter data in the monitoring area.

3. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion according to claim 1, characterized in that, The data transmission and aggregation module is internally configured with an edge computing gateway, a data caching unit, and a local analysis unit. The input end of the edge computing gateway is connected to the output end of the multi-parameter sensor network module. The data caching unit is used to temporarily store raw data from the monitoring micro-stations, and the local analysis unit is used to perform data preprocessing and preliminary analysis tasks.

4. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion according to claim 1, characterized in that, The chemical coupling mechanism analysis module is internally configured with a photochemical reaction path modeling unit, an oxidation reaction path modeling unit, and a gas-particle conversion path modeling unit. The output results of the coupling analysis engine include the coupling strength coefficient and chemical conversion contribution ratio between various pollutants.

5. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion according to claim 1, characterized in that, The dynamic threshold and anomaly identification module is internally configured with a dynamic threshold calculation unit and an exceedance determination unit. The dynamic threshold calculation unit calculates the dynamic baseline concentration range based on the historical data sequence of pollutant concentrations at each monitoring point, and the exceedance determination unit calculates the exceedance probability and exceedance magnitude of the real-time concentration at each monitoring point relative to the dynamic baseline.

6. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion according to claim 1, characterized in that, The pollution source tracing and transmission analysis module is internally configured with a transmission trajectory tracking unit, an air mass backward trajectory simulation unit, and a regional contribution analysis unit. The transmission trajectory tracking unit calculates the direction and speed of pollutant propagation based on the spatial topology of the monitoring network and the concentration gradient changes at each monitoring point.

7. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion according to claim 1, characterized in that, The intelligent early warning decision and release module is internally configured with a decision generation engine and a response strategy library. The decision generation engine performs multi-level early warning assessments, and the early warning levels are divided into four levels. The response strategy library predefines response strategies corresponding to different early warning levels and different pollution causes.

8. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion according to claim 1, characterized in that, It also includes a data quality control module, which is connected to the data transmission and aggregation module. The module is internally configured with a data integrity verification unit, a spatiotemporal consistency verification unit, and a physical rationality verification unit.

9. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion according to claim 8, characterized in that, The data quality control module outputs a data quality report that includes the data integrity rate, data availability rate, and data quality score for each monitoring micro-station.

10. The intelligent air quality monitoring system for industrial parks based on multi-sensor fusion according to claim 1, characterized in that, It also includes a 3D visualization module, which is communicatively connected to the data transmission and aggregation module, the chemical coupling mechanism analysis module, the dynamic threshold and anomaly identification module, and the pollution source tracing and transmission analysis module. The module is internally configured with a park 3D scene construction unit, a pollutant concentration field rendering unit, and a pollution transmission dynamic demonstration unit.

Images