Machine learning based industrial pollution source tracing method and system

Abstract

The present application relates to the technical field of industrial pollution monitoring, in particular to an industrial pollution tracing method and system based on machine learning, which can accurately and efficiently determine the pollution source in an industrial park, meet the high requirements of environmental protection regulations and the public for environmental protection; the method is used in an industrial park, when the industrial pollution monitoring point monitors that the pollution index exceeds the standard, and the emissions of each factory in the park meet the emission standard for pollution, the pollution is traced and analyzed, the pollution source with the largest participation in the pollution exceeding standard event is determined, the method comprises the following steps: obtaining pollution monitoring data information monitored by the industrial pollution monitoring point; extracting the characteristics of the exceeding standard pollutants from the pollution monitoring data information to obtain the park pollutant characteristic portrait; the park pollutant characteristic portrait comprises the composition of the exceeding standard pollutants in the park and the real-time concentration of each exceeding standard pollutant; the park pollutant characteristic portrait is matched in the pre-built factory pollutant emission database.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial pollution monitoring, in particular to an industrial pollution tracing method and system based on machine learning. BACKGROUND

[0002] Under the background of the current rapid industrialization, as one of the main driving forces of economic development, the environmental management of industrial parks, especially pollution prevention and control, is facing unprecedented challenges; with the increasing strictness of environmental protection regulations and the improvement of public environmental awareness, accurately and effectively tracking and controlling industrial pollution sources has become a key link in environmental protection work; however, in the complex environment of industrial parks, it is often difficult to accurately determine the pollution responsibility subject by relying solely on traditional pollution source monitoring methods, especially in the complex situation where all factories are surface compliant emissions, but the pollution monitoring point still records pollution exceeding the standard.

[0003] The existing industrial pollution source tracing method usually relies on the data of fixed monitoring stations and regular inspection of the emission outlets of each factory, which not only has poor timeliness, but also is not capable when facing complex pollution, cross pollution or instantaneous emission events; in addition, due to the influence of meteorological conditions and geographical environmental factors on the diffusion path and concentration distribution of pollutants, it is difficult to accurately locate the responsible party causing environmental pollution only by emission outlet monitoring data. SUMMARY

[0004] To solve the above technical problems, the present application provides an industrial pollution tracing method and system based on machine learning, which can accurately and efficiently determine the pollution source in an industrial park and meet the high requirements of environmental protection regulations and the public for environmental protection.

[0005] In a first aspect, the present application provides an industrial pollution tracing method based on machine learning, which is used to trace and analyze pollution when the pollution index monitored by the industrial pollution monitoring point exceeds the standard, while the emissions of each factory in the industrial park meet the emission standard, to determine the pollution source with the greatest participation in the pollution exceeding event, the method comprising:

[0006] Obtaining pollution monitoring data information monitored by the industrial pollution monitoring point;

[0007] Extracting the characteristics of the exceeding pollutants from the pollution monitoring data information to obtain a park pollutant feature portrait; the park pollutant feature portrait includes the composition of the exceeding pollutants in the park and the real-time concentration of each exceeding pollutant;

[0008] Matching the park pollutant feature portrait in the pre-built factory pollutant emission database to obtain a plurality of first-order suspicious factories participating in pollution emission;

[0009] obtain a set of pollution diffusion environmental impact parameters and a position vector of each first-suspected factory relative to an industrial pollution monitoring point, the set of pollution diffusion environmental impact parameters including wind speed, wind direction, temperature, humidity, and air pressure;

[0010] perform diffusion analysis on each first-suspected factory based on the set of pollution diffusion environmental impact parameters and the position vector of each first-suspected factory, to determine a plurality of second-suspected factories participating in the pollution over-standard event;

[0011] control a UAV carrying a mobile monitoring sensor to monitor pollution data of each second-suspected factory, to obtain a factory pollutant feature portrait of each second-suspected factory;

[0012] compare and analyze the factory pollutant feature portrait of each second-suspected factory with a park pollutant feature portrait respectively, to obtain a pollution event participation degree of each second-suspected factory;

[0013] take the second-suspected factory with the largest pollution event participation degree as a pollution source of the pollution event in the industrial park.

[0014] Further, the method for obtaining the pollution event participation degree of the second-suspected factory includes:

[0015] for each second-suspected factory, calculate a composition similarity of over-standard pollutants in the factory pollutant feature portrait and the park pollutant feature portrait;

[0016] statistically obtain a concentration proportion of each over-standard pollutant in the park pollutant feature portrait, and take the concentration proportion of each over-standard pollutant as a weight coefficient of the over-standard pollutant;

[0017] according to the weight coefficient of each over-standard pollutant, perform weighted calculation on real-time concentrations of the same type of pollutants in the factory pollutant feature portrait and the park pollutant feature portrait, to obtain a pollutant concentration contribution index of the second-suspected factory;

[0018] based on a preset weight, perform weighted calculation on the composition similarity of over-standard pollutants and the pollutant concentration contribution index, to obtain the pollution event participation degree.

[0019] Further, the calculation formula for obtaining the pollution event participation degree of the second-suspected factory is:

[0020] Dj=α·Sj+β·Pj;

[0021] wherein Dj represents the pollution event participation degree of the second-suspected factory, α represents a weight factor of the composition similarity of pollutants, β represents a weight factor of the pollutant concentration contribution, Sj represents the composition similarity, and Pj represents the pollutant concentration contribution index.

[0022] Further, the park pollutant feature portrait acquisition method comprises:

[0023] The acquired industrial pollution monitoring point monitoring data information is pre-processed, including data cleaning, standardization processing and time series analysis;

[0024] Set the pollution exceeding threshold, based on the pollution exceeding threshold, analyze the pollutant concentration of each monitoring point in the monitoring data, and identify the exceeding pollutant whose actual concentration is higher than the pollution exceeding threshold and its concentration;

[0025] Feature extraction is performed on the exceeding pollutant, and the extracted features include component composition and real-time concentration:

[0026] The extracted exceeding pollutant features form a comprehensive park pollutant feature portrait.

[0027] Further, the method for building the factory pollutant emission database comprises:

[0028] Collect the basic information and emission data of all factories in the park, and clearly define the types of pollutants allowed to be emitted by each factory, the maximum emission amount and the emission standard;

[0029] The collected data is cleaned and standardized;

[0030] Design the database architecture, including the factory basic information table, the emission record table and the pollutant feature table;

[0031] Based on the geographical position information of the factory, an index and association mechanism is established in the database to support subsequent data query and matching operations.

[0032] Further, the method for acquiring the factory pollutant feature portrait of the second-order suspicious factory comprises:

[0033] According to the position information of the second-order suspicious factory, a drone flight plan is made, including flight path, monitoring height and stay time;

[0034] Before executing the task, all monitoring sensors carried are calibrated;

[0035] The drone automatically takes off according to the predetermined plan, reaches the air above each second-order suspicious factory, and starts real-time sampling of air sample data by starting the monitoring sensor;

[0036] The monitoring data on the drone is transmitted to the ground control center in real time through wireless communication technology, and the data is analyzed in real time to generate a dynamic factory pollutant feature portrait and obtain the factory pollutant feature portrait of each second-order suspicious factory.

[0037] Further, the method for determining the pollution source of the pollution event comprises:

[0038] All second-level suspicious factories are ranked from high to low according to their pollution incident participation degrees;

[0039] The first-ranked second-level suspicious factory, i.e., the factory with the highest pollution incident participation degree, is preliminarily identified as the primary suspect source of the pollution incident;

[0040] The organization personnel conduct on-site verification on the factory, including on-site sampling and analysis of its emissions, comparison with emission permit standards, checking of the operation status of its environmental protection facilities, and rechecking of its daily operation records;

[0041] Combined with the mobile monitoring data of the unmanned aerial vehicle, the fixed monitoring station data and the on-site sampling analysis results, cross verification of multi-source evidence is performed to determine the pollution source;

[0042] An investigation report is prepared to clearly indicate the pollution behavior, pollution degree and specific impact on the environment of the primary suspect factory.

[0043] On the other hand, the application also provides an industrial pollution tracing system based on machine learning, which comprises:

[0044] A data collection module is configured to acquire pollution monitoring data information monitored by an industrial pollution monitoring point;

[0045] A feature extraction module is configured to perform feature extraction on the pollution monitoring data information to obtain a park pollution feature portrait; the park pollution feature portrait comprises a composition of over-standard pollutants and real-time concentrations of the over-standard pollutants in the park;

[0046] An emission data matching module is configured to match the park pollution feature portrait in a pre-built factory pollution emission database to obtain a plurality of first-level suspicious factories participating in pollution emission;

[0047] A diffusion analysis module is configured to acquire a set of pollution diffusion environmental impact parameters and a position vector of each first-level suspicious factory relative to the industrial pollution monitoring point; the set of pollution diffusion environmental impact parameters comprises wind speed, wind direction, temperature, humidity and air pressure; based on the set of pollution diffusion environmental impact parameters and the position vector of each first-level suspicious factory, diffusion analysis is performed on each first-level suspicious factory to determine a plurality of second-level suspicious factories participating in the over-standard pollution event;

[0048] An unmanned aerial vehicle monitoring module is configured to control an unmanned aerial vehicle carrying a mobile monitoring sensor to monitor pollution data of each second-level suspicious factory to obtain a factory pollution feature portrait of each second-level suspicious factory;

[0049] The comparison analysis module is configured to compare and analyze the factory pollutant feature portraits of the respective second-level suspicious factories with the park pollutant feature portrait, to obtain the pollution event participation degrees of the respective second-level suspicious factories.

[0050] The source determination module is configured to determine the second-level suspicious factory with the largest pollution event participation degree as the pollution source of the pollution event in the industrial park.

[0051] In a third aspect, the present application provides an electronic device, comprising a bus, a transceiver, a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory and the processor are connected through the bus, and the computer program is executed by the processor to implement the steps in any of the above methods.

[0052] In a fourth aspect, the present application further provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps in any of the above methods.

[0053] Compared with the prior art, the method has the beneficial effects that: by extracting the park pollutant features and establishing the feature portrait, the method can more finely identify the types and concentrations of pollutants, and can more accurately reflect the pollution situation compared with the traditional method; in combination with the matching of the machine learning model in the factory emission database, the method can efficiently screen out suspicious factories with high matching degree of pollution features, and improves the accuracy of preliminary screening;

[0054] By using real-time monitoring data for analysis, compared with the traditional method relying on fixed monitoring stations and regular inspection, the response speed to pollution events is significantly improved, and the emergency response mechanism can be started faster;

[0055] By incorporating environmental parameters such as wind speed and wind direction for diffusion analysis, the diffusion path and concentration distribution of pollutants can be simulated more closely to the actual situation, so that the pollution source can be located more accurately in a complex environment, and the problem of difficult accurate positioning relying only on emission port data is solved;

[0056] By using unmanned aerial vehicles for mobile monitoring, the monitoring range and flexibility are expanded, and by re-acquiring real-time pollutant data of the second-level suspicious factories and combining machine learning for comparison and analysis, the pollution contribution of each factory can be more scientifically evaluated;

[0057] The method can objectively determine the pollution responsibility subject based on data analysis, reduces the subjectivity and errors of human judgment, helps environmental protection law enforcement to be more fair and efficient, and also promotes enterprises to pay more attention to their environmental compliance;

[0058] Through continuous accumulation of monitoring data and optimization of the model, the pollution pattern of the industrial park can be promoted to be understood in depth, and scientific basis can be provided for formulating more effective prevention measures and environmental management policies;

[0059] In conclusion, the method can accurately and efficiently determine the pollution source in the industrial park by collecting pollution data in real time, constructing the pollution characteristic portrait of the park, and combining the factory emission database and pollution diffusion analysis, so as to meet the high requirements of environmental protection regulations and the public for environmental protection. BRIEF DESCRIPTION OF DRAWINGS

[0060] Fig. 1 is a flowchart of the present application;

[0061] Fig. 2 is a flowchart of the pollution event participation degree acquisition method of the second suspicious factory;

[0062] Fig. 3 is a structural diagram of the industrial pollution tracing system based on machine learning. DETAILED DESCRIPTION

[0063] In the description of the present application, those skilled in the art should know that the present application can be implemented as a method, an apparatus, an electronic device and a computer readable storage medium. Therefore, the present application can be specifically implemented as the following forms: complete hardware, complete software (including firmware, resident software, microcode, etc.), hardware and software combined form. In addition, in some embodiments, the present application can also be implemented as a computer program product in one or more computer readable storage media, which contains computer program code.

[0064] The above computer readable storage medium can adopt any combination of one or more computer readable storage media. The computer readable storage medium includes: an electric, magnetic, optical, electromagnetic, infrared or semiconductor system, device or instrument, or any combination thereof. More specific examples of computer readable storage medium include: portable computer diskette, hard disk, random access memory, read-only memory, erasable programmable read-only memory, flash memory, optical fiber, compact disk read-only memory, optical storage device, magnetic storage device or any combination thereof. In the present application, the computer readable storage medium can be any tangible medium containing or storing programs, which can be used or combined with instruction execution system, device or instrument.

[0065] The acquisition, storage, use, processing and the like of data in the technical solution of the present application comply with the relevant provisions of national laws.

[0066] The method, apparatus and electronic device provided by the present application are described by flowchart and / or block diagram.

[0067] It should be understood that each block of the flowchart and / or block diagram, and combinations of blocks in the flowchart and / or block diagram, can be implemented by computer readable program instructions. These computer readable program instructions can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0068] These computer readable program instructions can also be stored in a computer readable storage medium that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable storage medium produce an article of manufacture including instructions which implement the function / act specified in the flowchart and / or block diagram block or blocks.

[0069] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable data processing apparatus implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0070] The application is described below with reference to the accompanying drawings.

[0071] Embodiment one: as shown in the figure, the machine learning-based industrial pollution tracing method of the application specifically comprises the following steps: Figs. 1-2

[0072] S1, obtaining pollution monitoring data information monitored by an industrial pollution monitoring point;

[0073] The S1 step is to use advanced sensing technology and network communication technology to collect environmental quality data of each key position in the industrial park in real time; the following is a detailed introduction to the S1 step:

[0074] The monitoring stations are scientifically and reasonably arranged in the industrial park; the monitoring stations should cover the key areas of the entire park, including the vicinity of industrial discharge outlets, the surroundings of environmentally sensitive areas, and the diffusion path of pollutants; when arranging, factors such as terrain, wind direction, population density, etc. are considered to ensure that the monitoring network can fully reflect the pollution situation of the park;

[0075] Each monitoring station is equipped with multiple high-sensitivity sensors for detecting common air and water pollutants, including sulfur dioxide, nitrogen oxides, particulate matter, and volatile organic compounds;

[0076] ​The monitored data is transmitted in real time to a central data processing system through wired and wireless networks; the data transmission protocol needs to ensure the security and integrity of the data to avoid distortion or loss during transmission;

[0077] The collected data is pre-processed, including data cleaning, format unification and preliminary analysis, to facilitate subsequent feature extraction and analysis.

[0078] In this step, by scientifically arranging monitoring stations in key areas of the industrial park, all-weather and all-around monitoring of various pollutants in the air and water can be achieved, and pollution abnormalities can be captured in time to provide early warning signals for environmental protection, which helps to respond quickly and reduce the impact of environmental pollution incidents; the application of high-sensitivity sensors ensures the accuracy of monitoring data, combined with the coverage monitoring of industrial discharge outlets, environmental sensitive areas and pollutant diffusion paths, the potential pollution sources can be more accurately located, laying a foundation for subsequent pollution tracing work; real-time transmission of monitoring data through wired and wireless networks and the use of secure data transmission protocols ensure the integrity and security of data transmission during the process, so that the information received by the central data processing system is accurate and reliable, enhancing the effectiveness of data analysis and the reliability of decision-making; the data preprocessing process improves the quality and usability of the data, simplifies the complexity of subsequent feature extraction and in-depth analysis, accelerates the conversion process from data to insight, and improves the overall work efficiency.

[0079] S2, feature extraction of pollutants exceeding the standard, obtaining the park pollution feature portrait; the park pollution feature portrait includes the composition of pollutants exceeding the standard in the park and the real-time concentration of each pollutant exceeding the standard;

[0080] Step S2 is to analyze the large amount of pollution data collected from the monitoring points in depth to identify the specific types of pollutants and their concentration distribution, forming an accurate park pollution feature portrait; the following is a detailed introduction to this step:

[0081] The collected monitoring data is pre-processed, including data cleaning, standardization and time series analysis;

[0082] Set the pollution exceeding standard threshold, based on the pollution exceeding standard threshold, analyze the pollutant concentration of each monitoring point in the monitoring data, and identify the pollutants exceeding the standard and their concentrations whose actual concentrations are higher than the pollution exceeding standard threshold;

[0083] Feature extraction of pollutants exceeding the standard, the extracted features include composition and real-time concentration:

[0084] The extracted over-standard pollutants are characterized to form a comprehensive park pollutant characteristic profile; the park pollutant characteristic profile reflects the real-time situation of various over-standard pollutants at different monitoring points in the industrial park, and provides a comprehensive perspective to understand the distribution and source of the pollutants.

[0085] In this step, through detailed analysis of the monitoring data, it can be accurately identified which pollutant concentration exceeds the specified threshold, and then the specific over-standard pollutant type is locked, and the pollution source is accurately traced; the pre-processing step removes invalid or erroneous data, ensuring the accuracy and efficiency of subsequent analysis, so that in the face of sudden pollution incidents, the problem can be quickly located and timely response measures can be initiated; the constructed park pollutant characteristic profile not only reveals the real-time concentration of each over-standard pollutant, but also presents the characteristics of their spatial distribution in the park, which helps the regulatory authorities to make precise strategies and take effective control measures for key areas and pollutants; the comprehensive information provided by this step provides a scientific basis for environmental protection decision-making; decision-makers can develop or adjust environmental protection policies based on the detailed data of the pollutant characteristic profile, and optimize resource allocation; the long-term accumulation of pollutant characteristic profile data helps to analyze pollution trends and evaluate the effectiveness of existing prevention and control measures, providing data support for the environmental governance and sustainable development strategy of the park, and realizing the forward-looking planning of pollution prevention and control; Step S2 not only promotes the rapid response and effective handling of pollution incidents, but also lays a solid foundation for scientific decision-making and long-term planning of park pollution management.

[0086] S3, matching the park pollutant characteristic profile in the pre-built factory pollutant emission database to obtain a plurality of first suspicious factories participating in pollution emission;

[0087] Step S3 is to preliminarily screen out the first suspicious factories that may contribute to the current pollution event by matching the park pollutant characteristic profile with the existing factory pollutant emission database;

[0088] The method for building the factory pollutant emission database comprises:

[0089] Collect the basic information of all factories in the park, including the factory name, address, business type and emission type; at the same time, collect the emission-related data, including the emission concentration, emission amount, emission period and emission port position of various pollutants, and clearly define the types of pollutants allowed to be emitted by each factory, the maximum emission amount and the emission standard;

[0090] Clean and standardize the collected data to ensure the accuracy and consistency of the data; check whether the data is complete, process possible abnormal values or error data, and unify the data format and unit for subsequent data analysis and matching;

[0091] A well-designed database architecture is established, including factory basic information table, emission record table, pollutant characteristics table and multiple associated tables, to facilitate data management and query;

[0092] Suitable indexes and association mechanisms are established in the database to support subsequent data query and matching operations; the indexes and association mechanisms are constructed based on the unique identifier or geographical location of the factory, etc., to quickly locate and identify potential pollution sources;

[0093] The factory pollutant emission database includes:

[0094] Factory basic information, including factory name, geographical location coordinates, industry category, main production products or processes, environmental protection rating, pollution discharge permit number and validity period, etc.; these basic information helps to quickly locate and analyze the background of potential pollution sources;

[0095] Emission port information, recording the location, number, emission substance category and corresponding emission standard limit value of each factory's emission port;

[0096] Pollutant emission record, detailed record of historical and real-time emission data, including emission concentration, emission rate, emission total amount and emission period of various pollutants; these data should be updated regularly, and contain abnormal emission record and processing;

[0097] Pollutant composition analysis, for emission substances, the database should contain detailed chemical composition analysis report, including main pollutants and potential trace pollutants, which helps to build more accurate pollutant characteristics portrait;

[0098] Environmental impact assessment report, factory submitted environmental impact assessment report abstract, covering the potential impact assessment of its production activities on the surrounding environment, and the prevention and mitigation measures taken;

[0099] Monitoring and detection report, factory self-monitoring report, third-party detection report and environmental protection department's spot check results, which verify whether the factory's actual emission conforms to the permission provisions, and provide basis for evaluating the factory's compliance;

[0100] Emergency response record, recording the response measures, accident cause analysis, post-processing and rectification of past emergency emission events, which helps to evaluate the factory's emergency management level and prevention strategies for similar future events.

[0101] In this step, by quickly comparing the pollution characteristics monitored in the park with the factory emission information in the database, the factories that match the characteristics of the pollution event can be preliminarily screened out, i.e. the first-order suspicious factories, which significantly improves the efficiency and pertinence of pollution source identification, making the investigation work targeted; the detailed factory emission data, composition analysis and environmental assessment data in the database provide a solid foundation for pollution source analysis, making the pollution source tracing based on scientific and comprehensive data, improving the accuracy and reliability of judgment, and avoiding the limitations of relying solely on field observation or experience judgment; this method not only traces pollution events, but also through comprehensive analysis of historical emission records, monitoring and detection reports and emergency response records, it enhances the supervision of the daily environmental compliance of factories, promotes enterprises to strengthen self-management, comply with emission standards and reduce illegal emission behavior; the integrated factory information and emission data provide rich decision support information for environmental management departments, helping decision makers quickly grasp the pollution status in the park, timely adjust the supervision strategy, develop or optimize pollution prevention and control measures, and effectively respond to sudden environmental events; through this method, the responsible subject causing pollution can be more accurately located, providing direct basis for subsequent law enforcement actions, environmental restoration and compensation, thereby speeding up the pollution control process, protecting the ecological environment, and also maintaining the seriousness and fairness of the law; the implementation of step S3 not only improves the timeliness and accuracy of pollution event response, but also promotes the systematization and scientization of environmental management in the entire industrial park through strengthening supervision and data analysis, laying a foundation for realizing green and sustainable development.

[0102] S4, obtain a set of pollution diffusion environmental impact parameters and a position vector of each first-order suspicious factory relative to the industrial pollution monitoring point, the set of pollution diffusion environmental impact parameters including wind speed, wind direction, temperature, humidity and air pressure;

[0103] The environmental impact parameters and position vectors obtained through S4 stage can effectively combine machine learning algorithms for pollution diffusion simulation and analysis, thereby accurately determining the pollution impact range of each first-order suspicious factory, providing basic data support for subsequent monitoring and analysis of second-order suspicious factories; the following is a detailed introduction to this step:

[0104] Deploy wind speed and wind direction sensors within the industrial park to obtain data in real time or periodically; the sensors should be distributed throughout the key positions of the industrial park, covering various wind field conditions that affect the diffusion of pollutants;

[0105] Install temperature, humidity and air pressure sensors together with the wind speed and wind direction sensors to comprehensively record environmental parameters; regularly collect and integrate the data of these environmental parameters to reflect the propagation and diffusion of pollutants under different meteorological conditions;

[0106] The accurate position information of each first-suspected factory relative to the industrial pollution monitoring point is obtained through GPS positioning or map data to ensure the accuracy and real-time performance of the position vector.

[0107] In this step, by deploying a multi-dimensional sensor network in the industrial park, the key parameters such as wind speed, wind direction, temperature, humidity and air pressure are monitored in real time, which can accurately capture the small environmental changes affecting the diffusion of pollutants and significantly improve the accuracy of the simulation of the pollutant diffusion model; real-time monitoring of environmental parameters and position information obtained by GPS technology ensures rapid response to pollution incidents; once the monitoring point finds that the pollution exceeds the standard, it can immediately analyze according to the latest environmental data and quickly lock the possible pollution source, providing valuable time for timely control measures; the diffusion simulation considering meteorological conditions and geographical location information provides scientific and quantitative decision-making basis for environmental managers; S4 step greatly enhances the rapid response capability and traceability accuracy of industrial pollution incidents through high-tech monitoring means and data integration analysis.

[0108] S5, based on the set of pollution diffusion environmental impact parameters and the position vector of each first-suspected factory, performing diffusion analysis on each first-suspected factory to determine a plurality of second-suspected factories participating in the pollution exceeding standard event;

[0109] S5 step is to perform diffusion analysis on the identified first-suspected factory to further accurately locate the second-suspected factory that may cause the pollution exceeding standard event. The following is a detailed introduction to this step:

[0110] According to the geographical features and environmental conditions of the industrial park, select a suitable pollutant diffusion model; the pollutant diffusion model includes Gaussian diffusion model, computational fluid dynamics model and diffusion model based on statistics;

[0111] Set the environmental parameters required by the model, including wind speed, wind direction, temperature, humidity and air pressure;

[0112] Input the real-time pollution data of the monitoring point, the emission data of the first-suspected factory and the environmental parameters into the model for preliminary simulation;

[0113] Use historical data and actual monitoring results to correct the model to ensure that the simulation results are as consistent as possible with the actual situation;

[0114] Based on the known concentration of the pollution monitoring point, the reverse tracking technology is used to simulate the source direction and path of the pollutants from the monitoring point; it helps to narrow down the range of suspected pollution sources and determine which first-suspected factory is most likely to cause the pollution of the monitoring point to exceed the standard;

[0115] Simulate the emission scenarios of each first-order suspected factory to evaluate its contribution to the pollution concentration of the monitoring point; including changing the emission rate, emission time, etc. of a single or multiple factories, observing the changes in the simulation results to determine which factors have the greatest impact on pollution incidents;

[0116] Based on the results of the diffusion simulation, identify the few first-order suspected factories whose emissions have the highest contribution to the pollution of the monitoring point under various simulation scenarios, and upgrade them to second-order suspected factories.

[0117] In this step, through diffusion analysis of first-order suspected factories, we can go beyond single emission data comparison and consider the actual impact of meteorological conditions and geographical factors on pollutant dispersion, thus more accurately locating the pollution source; we can effectively distinguish between direct responsible parties and indirect influencers in complex environments, improving the accuracy of pollution incident tracing; simulation analysis provides quantitative data support, enabling environmental management to make decisions based on scientific evidence and determine which factories are most likely to have participated in the pollution incident, providing a solid foundation for taking further regulatory measures or legal action; by accurately locating second-order suspected factories, we can deploy limited monitoring and law enforcement resources in a targeted manner, avoiding blind patrols and improving work efficiency and the effectiveness of interventions; reverse tracking and emission scenario simulation help identify which combinations of emission characteristics and environmental conditions are most likely to cause pollution to exceed standards, providing specific guidance for factories to improve emission management and prevent future pollution incidents, promoting a shift from passive response to proactive prevention; S5 not only enhances the ability to trace specific pollution incidents, but also provides strong support for long-term environmental protection strategies and efficient use of resources in industrial parks.

[0118] S6, control the unmanned aerial vehicle carrying mobile monitoring sensors to monitor the pollution data of each second-order suspected factory to obtain the factory pollutant characteristic portrait of each second-order suspected factory;

[0119] S6 involves using unmanned aerial vehicles carrying mobile monitoring sensors to monitor the pollution data of identified second-order suspected factories, where professional-grade unmanned aerial vehicles with long endurance and stable flight performance should be selected to ensure safe task execution in complex and variable industrial park environments;

[0120] Mobile monitoring sensors are loaded under the unmanned aerial vehicle or integrated into the unmanned aerial vehicle structure. The sensors need to have high sensitivity and fast response capability to monitor specific pollutant species and their concentrations in the atmosphere in real time. The sensors also need to adapt to various weather conditions to ensure the accuracy of data collection;

[0121] Integrate GPS and inertial navigation systems to ensure that the unmanned aerial vehicle can accurately reach the predetermined monitoring position above each second-order suspected factory, while cooperating with the pre-set flight route to achieve automatic inspection;

[0122] The method for obtaining the factory pollutant feature portrait of the second suspicious factory comprises:

[0123] According to the location information of the second suspicious factory, a detailed unmanned aerial vehicle flight plan is formulated by using professional software, including a flight path, a monitoring height, and a stay time, so as to cover the key emission areas of each factory;

[0124] Before performing the task, all the monitoring sensors carried are calibrated to ensure the reliability and consistency of the monitoring data; the working mode and parameter setting of the sensor are adjusted according to the known pollutant types in the park to achieve the best monitoring effect;

[0125] The unmanned aerial vehicle automatically takes off according to the predetermined plan, starts the monitoring sensor after reaching the air above each second suspicious factory, and begins to collect air sample data in real time; this process includes vertical sampling at different height layers so as to better understand the vertical diffusion of the pollutants;

[0126] The monitoring data on the unmanned aerial vehicle is transmitted to the ground control center in real time through wireless communication technology, the data is analyzed in real time, a dynamic factory pollutant feature portrait is generated, and the factory pollutant feature portrait of each second suspicious factory is obtained;

[0127] In this step, the mobility and flexibility of the unmanned aerial vehicle can cover the key emission areas above multiple second suspicious factories in a short time, significantly improving the monitoring efficiency and spatial coverage compared with the traditional ground monitoring method, which helps to quickly lock the potential pollution source; the high-sensitivity sensors carried by the unmanned aerial vehicle can monitor the concentration of multiple specific pollutants in real time, and the calibration program ensures the accuracy and consistency of the monitoring data, even in complex weather conditions, the effectiveness of the data is guaranteed, and the scientificity and authority of the pollution monitoring are enhanced; the vertical sampling strategy at different height layers makes the monitoring not limited to the ground level, but covers the vertical diffusion characteristics of the pollutants, this three-dimensional monitoring method provides more comprehensive information for analyzing the pollution diffusion path and concentration distribution, which helps to more accurately evaluate the pollution contribution rate of each factory; the data monitored by the unmanned aerial vehicle is transmitted back to the ground control center in real time through wireless communication, realizing the real-time analysis of the pollution situation, providing a basis for the environmental protection department to respond to pollution incidents quickly, accelerating the deployment of pollution control and emergency treatment measures, and effectively reducing the duration and impact range of environmental pollution; the integrated GPS and inertial navigation system ensures that the unmanned aerial vehicle can accurately perform the preset flight task, reduces the error and risk of manual operation, ensures the safe performance of the monitoring activity, and also reduces the labor intensity of personnel, improves the intelligent level of the overall monitoring task; the implementation of S6 step not only greatly improves the efficiency and precision of industrial pollution monitoring, but also enhances the timeliness and effectiveness of environmental supervision through high-tech means.

[0128] S7, compare the factory pollutant feature portrait of each second suspicious factory with the park pollutant feature portrait respectively to obtain the pollution event participation degree of each second suspicious factory;

[0129] S7 step is based on the data and analysis collected in advance, more detailed comparison and analysis is carried out on the second suspicious factory to obtain the participation degree evaluation of each second suspicious factory in the pollution event;

[0130] The method for obtaining the pollution event participation degree of the second suspicious factory comprises

[0131] The park pollutant feature portrait and the pollutant feature portrait of each second suspicious factory are converted into feature vectors; each pollutant component is regarded as a dimension, and the concentration value or existence is regarded as the numerical value of the dimension;

[0132] The cosine similarity algorithm is applied to measure the similarity between the pollutant component composition of the second suspicious factory and the park pollutant feature portrait; the cosine similarity can effectively reflect the similarity in the direction of two vectors, and can accurately evaluate the matching degree of the component structure even in the case of large absolute concentration difference;

[0133] Based on the park pollutant feature portrait, the proportion of the concentration of each over-standard pollutant in the total over-standard pollutant concentration is calculated to serve as the weight coefficient of the pollutant; the high concentration pollutant has greater influence on the overall pollution level, and should be given higher weight;

[0134] For each second suspicious factory, the concentration of the same pollutant in the pollutant feature portrait of the factory and the park pollutant feature portrait is multiplied by the corresponding weight coefficient, and then the weighted concentrations of all pollutants are summed to obtain the pollutant concentration contribution index of the factory; it ensures that more consideration is given to the pollutants that have greater impact on the environment;

[0135] Two preset weight factors are set, one is used to consider the similarity of the pollutant component composition, and the other is used to consider the pollutant concentration contribution; the two weight factors need to be adjusted according to the actual situation to balance the importance of component matching and concentration influence;

[0136] According to the preset weight factors, the pollutant component composition similarity and the pollutant concentration contribution index of each second suspicious factory are weighted and summed; if the component similarity is considered more important, a higher weight is given; otherwise; the final score is the pollution event participation degree of the factory;

[0137] The calculation formula for obtaining the pollution event participation degree of the second suspicious factory is:

[0138] Dj=α·Sj+β·Pj

[0139] wherein Dj represents the pollution event participation degree of the second-order suspected factory, a represents a weight factor of the similarity of the component composition of the pollutant, β represents a weight factor of the concentration contribution of the pollutant, Sj represents the component composition similarity, and Pj represents the pollutant concentration contribution index.

[0140] In this step, through in-depth comparison and analysis of the second-order suspected factories, it can more accurately identify which factories played a key role in a specific pollution event; it helps to go beyond the preliminary screening and directly reach the core of the pollution problem, providing accurate targets for subsequent supervision and management; by converting complex environmental monitoring data into feature vectors and applying the cosine similarity algorithm and weighted concentration contribution calculation, a method for quantitatively evaluating the pollution contribution of each factory is provided; this quantitative method makes the attribution of pollution events more objective and scientific, reducing the error of subjective judgment; this step combines component similarity and concentration contribution information, greatly improving the efficiency and accuracy of pollution tracing; it can lock the most likely pollution source in a short time and accelerate the implementation of response measures; by setting adjustable weight factors, this method can flexibly adjust the evaluation focus according to different industrial park characteristics, seasonal changes, pollutant types, and other factors, ensuring the flexibility and adaptability of the method; accurately identifying the main participants in pollution events allows environmental protection departments and park managers to more reasonably allocate limited regulatory resources, focusing on solving the most critical problems and improving the overall effectiveness of environmental protection and management.

[0141] S8, the second-order suspected factory with the largest pollution event participation degree is determined as the pollution source of the pollution event in the industrial park;

[0142] Each second-order suspected factory has obtained its own pollution event participation degree score through detailed comparison and analysis; these scores comprehensively reflect the similarity of the pollutant emission characteristics of each factory to the pollutant characteristics of the park, the relative importance of the emission concentration, and the influence degree predicted by the diffusion model;

[0143] The method for determining the pollution source of a pollution event comprises:

[0144] All second-order suspected factories are ranked from high to low according to their pollution event participation degrees;

[0145] The second-order suspected factory ranked first, i.e., the factory with the highest pollution event participation degree score, is preliminarily identified as the primary suspect source of the pollution event;

[0146] In order to confirm it finally, a professional team needs to be organized to conduct on-site verification of the factory; this includes sampling and analyzing its emissions in the field, comparing with the emission permit standards, checking the operation status of its environmental protection facilities, and reviewing its daily operation records, etc.;

[0147] By combining mobile monitoring data from drones, data from fixed monitoring stations, and on-site sampling analysis results, cross-validation of multi-source evidence is conducted to eliminate the possibility of misjudgment and ensure that the determination of pollution sources is scientific and rigorous.

[0148] After completing all verification procedures, a detailed investigation report will be prepared, clearly identifying the pollution behavior, degree of pollution, and specific environmental impact of the primary suspected factory. This report will serve as the basis for official announcements and will also provide a basis for subsequent legal accountability and rectification measures.

[0149] In this step, by identifying the second-order suspected factory with the highest participation in the pollution incident as the primary suspected source, the most likely factory to cause pollution can be located efficiently and accurately, significantly improving the efficiency and accuracy of pollution source tracing compared to traditional methods. By comprehensively considering factors such as the similarity of pollutant emission characteristics, concentration contribution, and diffusion model predictions, participation scores are assigned to each second-order suspected factory and they are ranked, ensuring the objectivity and comprehensiveness of the assessment. The factory ranked first, as the primary suspected source, provides a clear target for subsequent on-site verification, enhancing the focus of the investigation and reducing unnecessary resource consumption. A professional team is organized to conduct on-site verification, sampling analysis, facility inspection, and operation. Multi-dimensional verification, including record review, not only enhanced the scientific rigor of the judgment but also effectively avoided misjudgments and ensured the reliability of the investigation results. Cross-verification of multi-source evidence, combining drone monitoring, fixed-site data, and on-site sampling results, promoted information sharing and collaboration among environmental protection departments, monitoring agencies, and the companies involved, increasing the transparency of the handling process. The final investigation report detailed the pollution behavior, extent, and environmental impact of the primary suspected factory, providing a solid factual basis and legal grounds for official government announcements, initiating legal accountability procedures, and formulating specific rectification measures. This facilitates the effective resolution of pollution issues and provides guidance for the prevention and management of similar incidents in the future.

[0150] Example 2: Fig. 3 As shown, the machine learning-based industrial pollution tracing system of the present invention specifically includes the following modules;

[0151] The data acquisition module is used to acquire pollution monitoring data from industrial pollution monitoring points.

[0152] The feature extraction module is used to extract the features of pollutants exceeding the standards from pollution monitoring data to obtain a pollutant feature profile of the park; the pollutant feature profile of the park includes the composition of pollutants exceeding the standards in the park and the real-time concentration of each pollutant exceeding the standards.

[0153] The emission data matching module is used to match the pollutant feature profiles in the park with a pre-built factory pollutant emission database to obtain multiple first-order suspected factories involved in pollution emissions.

[0154] a diffusion analysis module configured to obtain a set of pollution diffusion environmental impact parameters and a position vector of each first-suspected factory relative to an industrial pollution monitoring point, the set of pollution diffusion environmental impact parameters including wind speed, wind direction, temperature, humidity, and air pressure; based on the set of pollution diffusion environmental impact parameters and the position vector of each first-suspected factory, performing diffusion analysis on each first-suspected factory to determine a plurality of second-suspected factories participating in the pollution over-standard event;

[0155] a UAV monitoring module configured to control a UAV carrying a mobile monitoring sensor to monitor pollution data of each second-suspected factory to obtain a factory pollutant feature portrait of each second-suspected factory;

[0156] a comparison and analysis module configured to compare and analyze the factory pollutant feature portrait of each second-suspected factory with the park pollutant feature portrait respectively to obtain a pollution event participation degree of each second-suspected factory;

[0157] a source determination module configured to determine a second-suspected factory with the largest pollution event participation degree as a pollution source of the pollution event in the industrial park.

[0158] The system uses real-time monitoring data and a feature extraction module to quickly form a park pollutant feature portrait, can quickly identify over-standard pollutants and their real-time concentrations, and helps to timely locate potential pollution sources.

[0159] By establishing a factory pollutant emission database and using a machine learning algorithm, the park pollutant feature portrait and factory emission data can be intelligently matched to identify a plurality of first-suspected factories, thereby improving the accuracy and efficiency of the traceability.

[0160] The system not only considers static monitoring data, but also processes complex pollution, cross-pollution, or instantaneous emission events through a diffusion analysis module and a real-time monitoring UAV module, thereby greatly improving the adaptability of the system under changing environmental conditions.

[0161] The comparison and analysis module uses detailed factory pollutant feature portraits to accurately assess the participation degree of each second-suspected factory in the pollution event, can not only qualitatively determine the pollution source, but also quantitatively analyze the contribution of different factories to the pollution event, and helps to accurately determine the main pollution source.

[0162] The source determination module of the system can quickly identify the largest participant in the pollution event, provides scientific basis and decision support for environmental management departments, helps to quickly develop and implement effective pollution prevention and control measures, and protects the environment and public health.

[0163] In summary, the machine learning-based industrial pollution tracing system effectively addresses the pollution tracing problem in the complex environment of industrial parks by integrating real-time monitoring, big data analysis, and machine learning technology, thereby improving the efficiency and accuracy of environmental protection work.

[0164] The various variations and specific embodiments of the machine learning-based industrial pollution tracing method in the foregoing embodiment one are also applicable to the machine learning-based industrial pollution tracing system of the present embodiment. Through the foregoing detailed description of the machine learning-based industrial pollution tracing method, those skilled in the art can clearly understand the implementation method of the machine learning-based industrial pollution tracing system of the present embodiment. Therefore, in the interest of brevity, the implementation method of the machine learning-based industrial pollution tracing system of the present embodiment will not be described in detail here.

[0165] In addition, the present application also provides an electronic device, comprising a bus, a transceiver, a memory, a processor and a computer program stored in the memory and executable on the processor, the transceiver, the memory and the processor are connected through the bus respectively, the computer program is executed by the processor to realize each process of the above-mentioned method for controlling output data, and the same technical effect can be achieved, in order to avoid repetition, here will not repeat.

[0166] The above only describes the preferred embodiments of the present application. It should be noted that those skilled in the art can make several improvements and modifications without departing from the technical principles of the present application, and these improvements and modifications should also be considered as within the protection scope of the present application.

Claims

1. A machine learning-based method for tracing industrial pollution sources, characterized in that, The method is used in industrial parks where pollution monitoring points detect pollution levels exceeding standards, but all factories within the park are discharging pollutants in accordance with emission standards. The method involves source tracing analysis to identify the pollution source most significantly involved in the pollution exceeding event. The method includes: Obtain pollution monitoring data from industrial pollution monitoring sites; The characteristics of pollutants exceeding the standards are extracted from the pollution monitoring data to obtain a pollutant characteristic profile of the park; the pollutant characteristic profile of the park includes the composition of pollutants exceeding the standards in the park and the real-time concentration of each pollutant exceeding the standards. The pollutant characteristic profiles in the park are matched with a pre-built database of factory pollutant emissions to identify multiple suspected first-order factories involved in pollution emissions. Obtain a set of environmental impact parameters for pollution diffusion and the location vectors of each first-order suspected factory relative to the industrial pollution monitoring points. The set of environmental impact parameters for pollution diffusion includes wind speed, wind direction, temperature, humidity, and air pressure. Based on the set of environmental impact parameters for pollution diffusion and the location vectors of each first-order suspected factory, diffusion analysis was conducted on each first-order suspected factory to identify multiple second-order suspected factories involved in this pollution exceeding the standard event. Control drones equipped with mobile monitoring sensors to monitor pollution data of each suspected second-order factory and obtain a profile of the pollutants of each suspected second-order factory. The pollutant feature profiles of each second-order suspected factory are compared and analyzed with the pollutant feature profiles of the industrial park to obtain the pollution event participation degree of each second-order suspected factory. The second-order suspected factory with the highest involvement in the pollution incident was identified as the source of pollution in the industrial park. The formula for calculating the participation degree of the pollution event of the second-order suspected factory is as follows: ； in This indicates the level of involvement in pollution events by a second-order suspected factory. Weighting factors representing the similarity of pollutant composition. Weighting factors representing the contribution of pollutant concentration. Indicates the similarity of component composition. This indicates the contribution index of pollutant concentration.

2. The industrial pollution source tracing method based on machine learning as described in claim 1, characterized in that, Methods for obtaining the involvement of second-order suspected factories in pollution events include: For each second-order suspected factory, calculate the similarity of the composition of pollutants exceeding the standard in the factory's pollutant feature profile and the pollutant feature profile of the industrial park; The concentration percentage of each pollutant exceeding the standard in the pollutant characteristic profile of the park is statistically analyzed, and the concentration percentage of each pollutant exceeding the standard is used as the weighting coefficient of each pollutant exceeding the standard. Based on the weight coefficients of each pollutant exceeding the standard, the real-time concentrations of pollutants of the same type in the pollutant feature profile of the factory and the pollutant feature profile of the industrial park are weighted and calculated to obtain the pollutant concentration contribution index of the second-order suspected factory. Based on preset weights, the similarity of the composition of pollutants exceeding the standard and the contribution index of pollutant concentration are weighted and calculated to obtain the participation degree of pollution events.

3. The industrial pollution source tracing method based on machine learning as described in claim 1, characterized in that, The method for obtaining the pollutant feature profile of the park includes: The pollution monitoring data obtained from industrial pollution monitoring points are preprocessed, including data cleaning, standardization, and time series analysis. Set pollution exceedance thresholds, and based on these thresholds, analyze the pollutant concentrations at each monitoring point in the monitoring data to identify pollutants whose actual concentrations exceed the pollution exceedance thresholds and their concentrations. Feature extraction was performed on pollutants exceeding the standards. The extracted features included component composition and real-time concentration. The extracted characteristics of pollutants exceeding the standards will be combined to create a comprehensive profile of pollutants in the park.

4. The industrial pollution source tracing method based on machine learning as described in claim 1, characterized in that, The method for building the factory pollutant emission database includes: Collect basic information and emission data of all factories in the park, and clarify the types of pollutants that each factory is allowed to emit, the maximum emission amount, and the emission standards. The collected data is cleaned and standardized. Design the database architecture, including a basic factory information table, an emissions record table, and a pollutant characteristic table; Based on the factory's geographical location information, an index and association mechanism are established in the database to support subsequent data query and matching operations.

5. The industrial pollution source tracing method based on machine learning as described in claim 1, characterized in that, Methods for obtaining the pollutant characteristic profile of a second-order suspected factory include: Based on the location information of the suspected second-order factory, formulate a drone flight plan, including flight path, monitoring altitude and dwell time; Before carrying out the mission, all onboard monitoring sensors were calibrated; The drones took off automatically according to the pre-determined plan. After arriving over each of the suspected second-level factories, they activated their monitoring sensors to start collecting air sample data in real time. Monitoring data from the drone is transmitted to the ground control center in real time via wireless communication technology. The data is analyzed instantly to generate dynamic profiles of pollutants from the factory, thus obtaining a profile of pollutants from each suspected second-order factory.

6. The industrial pollution source tracing method based on machine learning as described in claim 1, characterized in that, Methods for determining the source of pollution in pollution incidents include: All second-order suspected factories are sorted from highest to lowest according to their involvement in pollution events; The top-ranked Tier 2 suspected factory, which has the highest score for involvement in the pollution incident, has been preliminarily identified as the primary suspected source of this pollution incident. The organization arranged for personnel to conduct an on-site inspection of the factory, including sampling and analyzing its emissions, comparing them with emission permit standards, checking the operation of its environmental protection facilities, and reviewing its daily operation records. By combining mobile monitoring data from drones, data from fixed monitoring stations, and on-site sampling analysis results, cross-validation of multi-source evidence is conducted to determine the source of pollution. Prepare an investigation report that clearly identifies the pollution behavior, degree of pollution, and specific environmental impact of the primary suspect factory.

7. A machine learning-based industrial pollution source tracing system, wherein the system is applied to the machine learning-based industrial pollution source tracing method as described in claim 1, characterized in that, The system includes: The data acquisition module is used to acquire pollution monitoring data from industrial pollution monitoring points. The feature extraction module is used to extract the features of pollutants exceeding the standards from pollution monitoring data to obtain a pollutant feature profile of the park; the pollutant feature profile of the park includes the composition of pollutants exceeding the standards in the park and the real-time concentration of each pollutant exceeding the standards. The emission data matching module is used to match the pollutant feature profiles in the park with a pre-built factory pollutant emission database to obtain multiple first-order suspected factories involved in pollution emissions. The diffusion analysis module is used to obtain the set of environmental impact parameters for pollution diffusion and the position vectors of each first-order suspected factory relative to the industrial pollution monitoring point. The set of environmental impact parameters for pollution diffusion includes wind speed, wind direction, temperature, humidity and air pressure. Based on the set of environmental impact parameters for pollution diffusion and the position vectors of each first-order suspected factory, diffusion analysis is performed on each first-order suspected factory to identify multiple second-order suspected factories involved in this pollution exceedance event. The drone monitoring module is used to control drones equipped with mobile monitoring sensors to monitor pollution data of each second-order suspected factory and obtain a characteristic profile of the pollutants of each second-order suspected factory. The comparison and analysis module is used to compare and analyze the pollutant feature profiles of each second-order suspected factory with the pollutant feature profiles of the industrial park to obtain the pollution event participation degree of each second-order suspected factory. The source identification module identifies the second-order suspected factory with the highest involvement in the pollution incident as the source of pollution in the industrial park.

8. An electronic device for tracing industrial pollution sources based on machine learning, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, characterized in that, When the computer program is executed by the processor, it implements the steps of the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

Images