Atmospheric pollution detection early warning method based on artificial intelligence

Abstract

The invention discloses an air pollution detection early warning method based on artificial intelligence. The method comprises the following steps: S1, acquiring environment monitoring data, meteorological data, traffic flow data and health emergency treatment data; s2, constructing a dynamic weight incidence matrix based on the aligned data in the S1; s3, based on the weight matrix output in S2, correcting the basic threshold value to obtain a dynamic threshold value; s4, calculating a health risk comprehensive index according to the data in the S1 and the association weight in the S2; s5, grading pollution levels, and selecting an optimal corresponding strategy from a preset strategy library; and S6, recording the pollutant concentration change after the strategy in the step S4 is executed, and updating the weight matrix in the step S2 through reinforcement learning. Through cross-domain data fusion, dynamic weight analysis and closed-loop feedback optimization, a full-chain intelligent early warning system of'monitoring-evaluation-decision-optimization 'is constructed, and the accuracy, timeliness and health protection initiative of atmospheric pollution prevention and control in a complex urban environment are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and in particular to an air pollution detection and early warning method based on artificial intelligence. Background Art

[0002] With the acceleration of urbanization and the growth of industrial activity, air pollution has become a global challenge. Currently, air pollution early warning systems primarily rely on fixed-threshold monitoring and single-domain data analysis technologies. These systems deploy environmental monitoring stations to collect real-time pollutant concentrations (such as PM2.5 and SO2) and set static early warning thresholds based on national standards. This approach has become a mature system at the basic monitoring level, effectively identifying significant exceedances (such as heavy pollution events) and providing fundamental support for public health protection and pollution control.

[0003] However, the causes of modern urban air pollution are becoming increasingly complex, and the pollution diffusion process is affected by the dynamic coupling of multiple factors, including meteorological conditions (such as inversion layers and wind speed), traffic activities, and industrial emissions. Traditional technologies still face the following challenges in multi-source heterogeneous data fusion, cross-domain correlation analysis, and dynamic decision-making: static thresholds are difficult to adapt to sudden changes in meteorological conditions or sudden pollution events; existing systems typically process environmental, meteorological, and traffic data independently, lacking quantitative analysis of multi-domain interactions, making it difficult to accurately identify the dominant factors of pollution; and relying on a single pollutant concentration may overlook the synergistic effects of low-concentration multiple pollutants or the health exposure risks of sensitive populations, resulting in protective measures lagging behind actual health damage. Summary of the Invention

[0004] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of this application to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides an air pollution detection and early warning method based on artificial intelligence, comprising:

[0007] S1. Collect environmental monitoring data, meteorological data, traffic flow data, and health emergency data, and use spatiotemporal fusion technology to unify multi-source data to the same spatiotemporal benchmark and perform standardized processing;

[0008] S2, construct a dynamic weight association matrix based on the data aligned in S1, and calculate the real-time association weights of each pollution indicator and external features;

[0009] S3, based on the weight matrix output by S2, modifies the basic threshold to obtain a dynamic threshold; when the health emergency data in S1 exceeds the historical average by 20%, the threshold emergency lowering mechanism is triggered;

[0010] S4, calculate the comprehensive health risk index based on the data in S1 and the association weights in S2;

[0011] S5, based on the dynamic threshold generated by S3, combined with the real-time meteorological and traffic data in S1 and the comprehensive health risk index in S4, classifies the pollution level and selects the optimal corresponding strategy from the preset strategy library;

[0012] S6, records the changes in pollutant concentration after the execution of S4 strategy, and updates the weight matrix in S2 through reinforcement learning.

[0013] As a preferred solution of the artificial intelligence-based air pollution detection and early warning method of the present invention, the dynamic weight association matrix in S2 is as follows:

[0014]

[0015] Where: W m，n is the real-time weight of the mth pollution index and the nth external feature; α and β are the adjustment coefficients of the environment and health fields respectively; F n For external features; The pollutant concentration is the characteristic F n sensitivity to change; is the health index feature F n sensitivity to change;

[0016] The dynamic weight association matrix is updated every 10 minutes, and when S1 detects a sudden meteorological event, the matrix is immediately triggered to be recalculated.

[0017] As a preferred solution of the artificial intelligence-based air pollution detection and early warning method of the present invention, the calculation of the dynamic early warning threshold in S3 is as follows:

[0018]

[0019] Where: W m，n is the dynamic association weight calculated in step S2; F n (t) is the external characteristic value collected in real time in S1; is the historical maximum value of the characteristic; H crit is the preset health risk threshold; T base is the basic threshold; γ is the diffusion capacity correction coefficient; at the same time, when health emergency data (such as respiratory disease visit rate) exceeds the historical mean by 20%, the threshold is forced to be lowered, T′ dynamic(t) = T dynamic (t)×(1-δ), δ is the health correction coefficient.

[0020] As a preferred solution of the artificial intelligence-based air pollution detection and early warning method of the present invention, the specific algorithm of the comprehensive health risk index in S4 is as follows:

[0021]

[0022] Among them: H safe，m is the health safety concentration of the pollutant in the mth item (refer to WHO standard); W health，m is the weight related to health data; P m (t) is the real-time concentration of the pollutant in the mth column; H safe，m is the health and safety concentration threshold of the pollutant in the mth item.

[0023] As a preferred solution of the artificial intelligence-based air pollution detection and early warning method of the present invention, the pollution level classification adopts a dual-path parallel decision logic:

[0024] Path A: Single pollutant exceeding the standard judgment: the real-time concentration of each pollutant Pm(t) and its dynamic threshold T dynamic，m (t) comparative judgment;

[0025] The specific rules are as follows:

[0026] Level 1 warning: P m <0.8T dynamic，m ;

[0027] Level 2 warning: 0.8T dynamic，m ≤P m <T dynamic，m ;

[0028] Level 3 warning: P m ≥T dynamic，m ;

[0029] Path B: Comprehensive health risk index determination: through the health risk index H risk (t) Size is used for grade determination:

[0030] The specific rules are as follows:

[0031] Level 1 warning: H risk <1.0;

[0032] Level 2 warning: 1.0≤H risk <1.5;

[0033] Level 3 warning: H risk ≥1.5;

[0034] Final level = max(path A, path B).

[0035] As a preferred solution of the artificial intelligence-based air pollution detection and early warning method of the present invention, the strategy library in S5 contains hierarchical response rules, and its strategy generation satisfies:

[0036] When the warning level is level 3 and the meteorological diffusion coefficient γ is less than 0.3, the industrial production restriction strategy is activated forcibly;

[0037] When the comprehensive health risk index H risk When ≥1.5, priority should be given to implementing protection strategies for sensitive groups.

[0038] As a preferred solution of the artificial intelligence-based air pollution detection and early warning method of the present invention, wherein: the reinforcement learning in S6 includes the calculation and update algorithm of the reward function;

[0039] The reward function is as follows:

[0040] R=ΔP·log(1+ΔH)-λ·NegativeImpact;

[0041] Where: ΔP is the rate of change of pollutant concentration after the implementation of the strategy; ΔH is the rate of change of health emergency data; λ is the negative effect penalty coefficient; NegativeImpact is the normalized negative social impact;

[0042] The update algorithm formula is as follows:

[0043] ΔW m，n =η·(RR predicted );

[0044] Where: η is the learning rate; R predicted The reward value predicted by the model before the policy is executed.

[0045] In a second aspect, the present invention provides an air pollution detection and early warning system based on artificial intelligence, comprising:

[0046] Multi-source data acquisition module: real-time collection of environmental, meteorological, traffic, and health data and pre-processing;

[0047] Dynamic feature coupling analysis module: calculates the dynamic weight matrix;

[0048] Dynamic threshold generation module: calculates the dynamic threshold of each pollutant;

[0049] Health risk assessment module: calculates comprehensive health risk indicators;

[0050] Emergency strategy collaborative decision module: matches relevant decisions of the strategy library according to the preload level;

[0051] Feedback optimization module: records the actual implementation effect of emission reduction measures and updates the weight matrix;

[0052] Visual interaction module: facilitates staff observation and manual intervention.

[0053] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of an artificial intelligence-based atmospheric pollution detection and early warning method as described in the first aspect of the present invention is implemented.

[0054] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored, wherein: when the computer program is executed by a processor, it implements any step of an artificial intelligence-based atmospheric pollution detection and early warning method as described in the first aspect of the present invention.

[0055] Beneficial effects of the present invention:

[0056] By integrating real-time data such as meteorology and traffic, the warning threshold is dynamically revised to solve the warning lag problem of traditional fixed thresholds in scenarios such as calm weather and sudden pollution. The impact weights of multi-field factors (such as traffic flow and industrial emissions) on pollution are quantified based on a dynamic weight matrix to support the accurate identification of dominant pollution sources. The comprehensive health risk index is used to evaluate the synergistic effects of multiple pollutants and the exposure risks of sensitive populations, breaking through the limitations of single pollutant concentration determination. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0058] Figure 1 This is a flow chart of the artificial intelligence-based air pollution detection and early warning method proposed by the present invention;

[0059] Figure 2 This is a flow chart of an artificial intelligence-based air pollution detection and early warning method proposed by the present invention;

[0060] Figure 3 This is a system architecture diagram of an artificial intelligence-based air pollution detection and early warning system proposed in this invention. DETAILED DESCRIPTION

[0061] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0063] Reference Figure 1-3 The present invention provides an air pollution detection and early warning method based on artificial intelligence, comprising:

[0064] S1. Collect environmental monitoring data, meteorological data, traffic flow data, and health emergency data, and use spatiotemporal fusion technology to unify multi-source data to the same spatiotemporal benchmark and perform standardized processing;

[0065] S2, construct a dynamic weight association matrix based on the data aligned in S1, and calculate the real-time association weights of each pollution indicator and external features;

[0066] The dynamic weight association matrix in S2 is as follows:

[0067]

[0068] Where: W m，n is the real-time weight of the mth pollution index and the nth external feature (such as traffic volume, wind speed); α and β are the adjustment coefficients in the environment and health fields (obtained through historical data training); F n are external characteristics (such as traffic and meteorological parameters); The pollutant concentration is the characteristic F n Sensitivity to changes, such as increased traffic volume during peak hours in the morning and evening Significantly increased weight Then it increases; is the health index feature F n Sensitivity to changes, such as when an inversion layer descends, Increase, diffusion conditions deteriorate, leading to increased health risks; identify key driving factors through this matrix (such as the contribution of morning and evening peak traffic flow to NO2);

[0069] The dynamic weight association matrix is updated every 10 minutes, and when S1 detects a sudden meteorological event, the matrix is immediately recalculated. The 10-minute update can improve the timeliness of environmental response. The following scenario example: a sudden emission in an industrial area causes the PM2.5 concentration to drop from 60μg / m 3 Sudden rise to 90 μg / m 3(exceeding the threshold of 75 μg / m 3 ), and the wind speed dropped from 3m / s to 1m / s. The traditional one-hour update may delay the discovery of worsening diffusion conditions, resulting in an emergency response. This solution uses a 10-minute update. Within 6 update cycles, the downward trend of wind speed can be captured and an early warning can be issued. When health data is abnormal (such as a surge in emergency volume), the system automatically increases the β value, making health sensitivity the dominant weight calculation. When health emergency data exceeds the historical average by 20%, the β value is triggered to increase. For example: base β = 0.4, the number of pediatric emergency visits increased by 30% on that day, β new =0.4×1.3=0.52, the weight formula becomes

[0070] S3, based on the weight matrix output by S2, modifies the basic threshold to obtain a dynamic threshold, generating independent dynamic thresholds for each pollutant; when the health emergency data in S1 exceeds the historical average by 20%, the threshold emergency lowering mechanism is triggered;

[0071] The calculation of dynamic warning thresholds in S3 is as follows:

[0072]

[0073] Where: W m，n is the dynamic association weight calculated in step S2; F n (t) is the external characteristic value collected in real time in S1; is the historical maximum value of the characteristic; H crit is the preset health risk threshold; T base is the basic threshold (set based on national standards); γ is the diffusion capacity correction coefficient, which reflects the real-time correction coefficient of meteorological conditions on the diffusion capacity of pollutants. Its value is negatively correlated with the height of the inversion layer. H inversion (t) is the real-time inversion layer height, is the historical average inversion layer height; Quantify the comprehensive impact of external factors on pollution. If the current external conditions exacerbate pollution, the value of this item will increase, resulting in a lower threshold. At the same time, when health emergency data (such as the rate of visits for respiratory diseases) exceeds the historical average by 20%, the threshold will be forced to be lowered. T' dynamic (t) = T dynamic (t)×(1-δ), δ is the health correction coefficient, the default value is 0.2, that is, the threshold is reduced by 20%. The example is as follows: the original dynamic threshold T dynamic (t) = 68 μg / m 3 , after the health abnormality is triggered, T' dynamic (t) = 68 × 0.8 = 54.4 μg / m 3;

[0074] Calculation process example: Pollutant: PM2.5; Basic threshold: T base =75 μg / m 3 ; Real-time data: traffic flow F traffic =3000 vehicles / hour, weight W PM2.5，traffic =0.8; Inversion layer height: H inversion (t) = 200 meters, historical average m; Health emergency visits increase by 25% (triggering correction); Calculation steps: Synthetic Dynamic Threshold: Health Modifier: Threshold reduced by 20% → T' dynamic (t) = 102 × 0.8 = 81.6 μg / m 3 ; If the real-time PM2.5 concentration P current =85 μg / m 3 , then trigger the third level warning;

[0075] S4. Calculate the comprehensive health risk index based on the data in S1 and the correlation weights in S2; address the monitoring blind spot of "cooperative health hazards of multiple low-concentration pollutants";

[0076] The specific algorithm for the comprehensive health risk index in S4 is as follows:

[0077]

[0078] Among them: H safe，m is the health safety concentration of the pollutant in the mth item (refer to WHO standard); W health，m is the health data related weight, only W is used m，n Medium health sensitivity P m (t) is the real-time concentration of the pollutant in the mth column; H safe，m is the health and safety concentration threshold of the mth pollutant; it breaks through the limitation of traditional environmental early warning that only focuses on pollutant concentration, directly quantifies health losses, and forms a dual-track early warning system with the dynamic threshold system, solving the monitoring blind spot of "multiple low-concentration pollutants synergistically harming health";

[0079] Calculation process example: Pollutant data: PM2.5 concentration P1 = 60 μg / m 3 (H safe，1 =35 μg / m 3 ), O3 concentration P2=120μg / m 3 (H safe，2 =100 μg / m 3 ); Health weight: W health，1 =0.8,W health，2=0.3; Risk value of single pollutant: PM2.5: (60 / 35) × 0.8 = 1.371, O3: (120 / 100) × 0.3 = 0.36; H risk =1.371+0.36=1.731;

[0080] S5, based on the dynamic threshold generated by S3, combined with the real-time meteorological and traffic data in S1 and the comprehensive health risk index in S4, classifies the pollution level and selects the optimal corresponding strategy from the preset strategy library;

[0081] Pollution level classification adopts dual-path parallel decision logic:

[0082] Path A: Judgment of single pollutant exceeding the standard: Real-time concentration of each pollutant P m (t) and its dynamic threshold T dynamic，m (t) comparative judgment;

[0083] The specific rules are as follows:

[0084] Level 1 warning: P m <0.8T dynamic，m ; No action is required, continue monitoring;

[0085] Level 2 warning: 0.8T dynamic，m ≤P m <T dynamic，m ; Issue an early warning and notify staff;

[0086] Level 3 warning: P m ≥T dynamic，m ; Issue an early warning and notify staff;

[0087] Path B: Comprehensive health risk index determination: through the health risk index H risk (t) Size is used for grade determination:

[0088] The specific rules are as follows:

[0089] Level 1 warning: H risk <1.0; no action required, continue monitoring;

[0090] Level 2 warning: 1.0≤H risk <1.5; issue an early warning and notify staff;

[0091] Level 3 warning: H risk ≥1.5; issue an early warning and notify staff;

[0092] final_level = max(pathA,pathB);

[0093] The example is as follows: If PM2.5 triggers the second level warning, H riskIf a Level 3 alert is triggered, the final level 3 alert will be determined. The single pollutant pathway can accurately identify sudden exceedances of specific pollutants, ensuring targeted emergency measures. The health risk pathway can capture the synergistic effects of low-concentration concentrations of multiple pollutants to address monitoring blind spots where "health damage has already occurred even though a single item has not exceeded the standard." When the comprehensive health risk indicator pathway determines a higher level, health protection strategies will be prioritized even if pollutant concentrations have not exceeded the standard.

[0094] The policy library in S5 contains hierarchical response rules, and its policy generation satisfies:

[0095] When the warning level is level 3 and the meteorological diffusion coefficient γ is less than 0.3, the industrial production restriction strategy is activated forcibly;

[0096] When the comprehensive health risk index H risk When the rate is ≥1.5, priority should be given to the protection strategy for sensitive populations;

[0097] The example is as follows: When the system detects that the PM2.5 concentration reaches the dynamic threshold and the comprehensive health risk index H risk =1.6, the strategy library executes the following process:

[0098] Match the weather conditions (calm and stable weather) → activate "Odd-even license plate restriction for motor vehicles";

[0099] Detection of excessive concentration in the school area → superimpose "School Suspension Notice";

[0100] If the industrial source contribution weight is greater than 0.7, the "50% emission reduction for key enterprises" policy will be triggered simultaneously.

[0101] S6, records the changes in pollutant concentration after the execution of S4 strategy, and updates the weight matrix in S2 through reinforcement learning;

[0102] Reinforcement learning in S6 includes the calculation and update algorithm of the reward function;

[0103] The reward function is as follows:

[0104] R=ΔP·log(1+ΔH)-λ·NegativeImpact;

[0105] Where: ΔP is the rate of change of pollutant concentration after the policy is implemented; ΔH is the rate of change of health emergency data; λ is the negative effect penalty coefficient (the default value is 0.3, obtained through historical data training); NegativeImpact is the normalized negative social impact (such as the increase in the congestion index);

[0106] The update algorithm formula is as follows:

[0107] ΔW m，n =η·(RR predicted );

[0108] Where: η is the learning rate (default is 0.01 to prevent overfitting); R predicted The reward value predicted by the model before the strategy is executed, the predicted reward value R predicted The calculation adopts the standard temporal difference method in the field of reinforcement learning, based on the current dynamic correlation matrix W m，n (t) Based on the real-time environmental status, estimate the expected environmental improvement rate and health risk reduction rate after the implementation of the strategy;

[0109] An example is as follows: After an industrial zone implements the "production restriction" policy: PM2.5 decreases by 12% → ΔP = 0.12; the number of respiratory emergency visits decreases by 5% → ΔH = 0.05; the congestion index increases by 15%; R = 0.12·log(1.05)-0.3·0.15≈0.12·0.0488-0.045=-0.039; the model predicts a reward value of 0.02; ΔW = 0.01·(-0.039-0.02) = -0.00059, and the relevant weight W PM2.5，industry A slight downward adjustment will lower the priority of industrial production restriction strategies in similar scenarios in the future; by real-time evaluation of the actual effects of emergency strategies, model parameters will be dynamically adjusted to ensure continuous optimization of the system as the environment changes.

[0110] This embodiment also provides an artificial intelligence-based air pollution detection and early warning system, including:

[0111] Multi-source data acquisition module: real-time collection of environmental, meteorological, traffic, and health data and pre-processing;

[0112] Dynamic feature coupling analysis module: calculates the dynamic weight matrix;

[0113] Dynamic threshold generation module: calculates the dynamic threshold of each pollutant;

[0114] Health risk assessment module: calculates comprehensive health risk indicators;

[0115] Emergency strategy collaborative decision module: matches relevant decisions of the strategy library according to the preload level;

[0116] Feedback optimization module: records the actual implementation effect of emission reduction measures and updates the weight matrix;

[0117] Visual interaction module: facilitates staff observation and manual intervention.

[0118] This embodiment also provides a computer device suitable for an artificial intelligence-based atmospheric pollution detection and early warning method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement an artificial intelligence-based atmospheric pollution detection and early warning method proposed in the above embodiment.

[0119] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0120] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements an artificial intelligence-based air pollution detection and early warning method as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0121] In summary, the present invention has constructed a full-chain intelligent early warning system of "monitoring-assessment-decision-making-optimization" through cross-domain data fusion, dynamic weight analysis and closed-loop feedback optimization, which significantly improves the accuracy, timeliness and health protection initiative of air pollution prevention and control in complex urban environments.

[0122] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based air pollution detection and early warning method, characterized by: include: S1. Collect environmental monitoring data, meteorological data, traffic flow data, and health emergency data, and use spatiotemporal fusion technology to unify multi-source data to the same spatiotemporal benchmark and perform standardized processing; S2, construct a dynamic weight association matrix based on the data aligned in S1, and calculate the real-time association weights of each pollution indicator and external features; S3, based on the weight matrix output by S2, correct the basic threshold to obtain the dynamic threshold; When the health emergency data in S1 exceeds the historical average by 20%, the threshold emergency lowering mechanism is triggered; S4, calculate the comprehensive health risk index based on the data in S1 and the association weights in S2; S5, based on the dynamic threshold generated by S3, combined with the real-time meteorological and traffic data in S1 and the comprehensive health risk index in S4, classifies the pollution level and selects the optimal corresponding strategy from the preset strategy library; S6, records the changes in pollutant concentration after the execution of S4 strategy, and updates the weight matrix in S2 through reinforcement learning.

2. The method for detecting and warning air pollution based on artificial intelligence according to claim 1, characterized in that: The dynamic weight association matrix in S2 is as follows: Where: W m，n is the real-time weight of the mth pollution index and the nth external feature; α and β are the adjustment coefficients of the environment and health fields respectively; F n For external features; The pollutant concentration is the characteristic F n sensitivity to change; is the health index feature F n sensitivity to change; The dynamic weight association matrix is updated every 10 minutes, and when S1 detects a sudden meteorological event, the matrix is immediately triggered to be recalculated.

3. The method for detecting and warning air pollution based on artificial intelligence according to claim 2, characterized in that: The calculation of the dynamic warning threshold in S3 is as follows: Where: W m，n is the dynamic association weight calculated in step S2; F n (t) is the external characteristic value collected in real time in S1; is the historical maximum value of the characteristic; H crit is the preset health risk threshold; T base is the basic threshold; γ is the diffusion capacity correction coefficient; at the same time, when health emergency data (such as respiratory disease visit rate) exceeds the historical mean by 20%, the threshold is forced to be lowered, T′ dynamic (t) = T dynamic (t)×(1-δ), δ is the health correction coefficient.

4. The method for detecting and warning air pollution based on artificial intelligence according to claim 3, characterized in that: The specific algorithm of the comprehensive health risk index in S4 is as follows: Among them: H safe，m is the health safety concentration of the pollutant in the mth item (refer to WHO standard); W health，m is the weight related to health data; P m (t) is the real-time concentration of the pollutant in the mth column; H safe，m is the health and safety concentration threshold of the pollutant in the mth item.

5. The method for detecting and warning air pollution based on artificial intelligence according to claim 1, characterized in that: The pollution level classification adopts dual-path parallel decision logic: Path A: Judgment of single pollutant exceeding the standard: Real-time concentration of each pollutant P m (t) and its dynamic threshold T dynamic，m (t) comparative judgment; The specific rules are as follows: Level 1 warning: P m <0.8T dynamic，m ; Level 2 warning: 0.8T dynamic，m ≤P m <T dynamic，m ; Level 3 warning: P m ≥T dynamic，m ; Path B: Comprehensive health risk index determination: through the health risk index H risk (t) Size is used for grade determination: The specific rules are as follows: Level 1 warning: H risk <1.0; Level 2 warning: 1.0≤H risk <1.5; Level 3 warning: H risk ≥1.5; Final level = max(path A, path B).

6. The method for detecting and warning air pollution based on artificial intelligence according to claim 5, characterized in that: The policy library in S5 contains hierarchical response rules, and its policy generation satisfies: When the warning level is level 3 and the meteorological diffusion coefficient γ is less than 0.3, the industrial production restriction strategy is activated forcibly; When the comprehensive health risk index H risk When ≥1.5, priority should be given to implementing protection strategies for sensitive groups.

7. The method for detecting and warning air pollution based on artificial intelligence according to claim 6, characterized in that: The reinforcement learning in S6 includes the calculation and update algorithm of the reward function; The reward function is as follows: R=ΔP·log(1+ΔH)-λ·NegativeImpact; Where: ΔP is the rate of change of pollutant concentration after the implementation of the strategy; ΔH is the rate of change of health emergency data; λ is the negative effect penalty coefficient; NegativeImpact is the normalized negative social impact; The update algorithm formula is as follows: ΔW m，n =η·(RR predicted ); Where: η is the learning rate; R predicted The reward value predicted by the model before the policy is executed.

8. An artificial intelligence-based air pollution detection and early warning system, based on the artificial intelligence-based air pollution detection and early warning method according to any one of claims 1 to 7, characterized in that: include: Multi-source data acquisition module: real-time collection of environmental, meteorological, traffic, and health data and pre-processing; Dynamic feature coupling analysis module: calculates the dynamic weight matrix; Dynamic threshold generation module: calculates the dynamic threshold of each pollutant; Health risk assessment module: calculates comprehensive health risk indicators; Emergency strategy collaborative decision module: matches relevant decisions of the strategy library according to the preload level; Feedback optimization module: records the actual implementation effect of emission reduction measures and updates the weight matrix; Visual interaction module: facilitates staff observation and manual intervention.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the artificial intelligence-based air pollution detection and early warning method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of an artificial intelligence-based air pollution detection and early warning method according to any one of claims 1 to 7 are implemented.

Images