A method for rapid tracing of groundwater pollution in chemical parks

Abstract

The present invention relates to the field of groundwater pollution source tracing methods, and in particular to a method for rapid source tracing of groundwater pollution in chemical parks, which solves the problems of high cost and technical threshold, poor operability, and long source tracing time in the prior art for groundwater pollution source tracing. The present invention comprises the following steps: S1, collecting basic data; S2, demarcating the monitoring area; S3, carrying out data matching; obtaining a preliminary judgment result based on the data matching result; S4, carrying out a secondary judgment to determine the source of groundwater pollution in the park. The present invention can collect data through the existing groundwater environmental monitoring wells in the park, which is low in cost and easy to operate; the basic data of the enterprise and the environmental monitoring data are analyzed and matched to obtain a list of enterprises suspected of being pollution sources in the park; on this basis, the pollution source of the park is further locked, and the source tracing of the groundwater pollution in the park is completed.

Description

Technical Field

[0001] The present invention relates to the field of groundwater pollution source tracing methods, and in particular to a method for rapid groundwater pollution source tracing in chemical parks. Background Art

[0002] Groundwater pollution source tracing involves studying the migration patterns of pollutants in groundwater and combining them with modern analytical and detection technologies to identify the source and extent of groundwater contamination. Once contaminated, underground aquifers are difficult and costly to remediate. Source control is key to groundwater pollution remediation and management, so identifying the source of contamination is crucial for developing groundwater pollution prevention and control strategies for chemical parks.

[0003] At present, the commonly used methods for tracing the source of groundwater pollution in chemical parks include mathematical statistics, model optimization and geochemical footprint methods. These methods require sufficient basic data, detailed site information, professional simulation software or specialized isotope detection to achieve the tracing of groundwater pollution in the park. This makes the cost and technical threshold of tracing the source of groundwater pollution in chemical parks high, and the operability poor. It may result in the park being unable to carry out tracing work in a timely manner after discovering groundwater pollution, which is not conducive to the park blocking the pollution source in time and preventing the further spread of groundwater pollution.

[0004] There is an urgent need for a new method of tracing the source of groundwater pollution that can solve the above problems. Summary of the Invention

[0005] The present invention proposes a method for rapid tracing of groundwater pollution in chemical parks, which solves the problems of high cost and technical threshold, poor operability and long tracing time in the prior art of groundwater pollution tracing.

[0006] The technical solution of the present invention is achieved as follows: a method for rapid tracing of groundwater pollution in a chemical park, comprising the following steps:

[0007] S1. Collect basic data: Collect enterprises and pollution-related information related to groundwater pollution in the chemical park and establish a database; the database includes: park enterprise information: enterprise list; enterprise environmental impact report; enterprise completion environmental protection acceptance monitoring report, the enterprise environmental impact report includes the enterprise characteristic exceeding standard factors and the enterprise's groundwater characteristics; park information: park enterprise pollutant emission declaration registration form; park enterprise environmental statistical report; park area hydrogeological survey report; park engineering geological survey report; park and enterprise floor plan; park boundary map; park existing groundwater environment monitoring well information, the monitoring well information includes: coordinates, well depth, water level and monitoring data; park groundwater environment status survey report; park enterprise groundwater environment pollution accident records; the database is set on the server; the server also includes a client with communication connection;

[0008] S2. Delineate the monitoring area: Delineate the monitoring area based on the coordinates of existing groundwater environment monitoring wells. The park and enterprise layout, park boundary map, and groundwater environment monitoring well locations can be gridded using satellite imagery and then delineated.

[0009] S3. Carry out data matching: Regularly monitor groundwater through existing groundwater environmental monitoring wells, upload the test results, and match and save them in combination with the database in S1; obtain a preliminary judgment result based on the data matching results; S31 The test data is processed by the data acquisition module in the server; S32 The test data is compared with the information in the database through the information processing module of the server; When the test data exceeds the set threshold, the information processing module sends a trigger alarm instruction and sends the monitoring well information related to the test data to the park management client; S33 After receiving the alarm instruction, the park management client issues an alarm prompt;

[0010] S4. Based on the determination result in S3, a secondary determination is conducted to determine the source of groundwater pollution in the park. Based on the result of the secondary determination, a self-inspection alert is sent to the enterprise client in the area where the pollution is determined, prompting the enterprise to conduct a self-inspection and report the self-inspection results. At the same time, pipeline water is extracted and sent for inspection. The inspection results can be sent to both the park management client and the enterprise client at the same time.

[0011] S5. Enterprises can quickly detect leaks based on self-inspection alarms and trace the source. If no pollutant leakage is found in the enterprise's self-inspection, the pollutant-emitting enterprise can be quickly identified based on the comparison of the inspection results and the main pollutant detection results to achieve accurate traceability.

[0012] According to a further technical solution, the information processing module in step S32 can also obtain the two monitoring wells with the highest correlation with the detection data of the monitoring well based on the correlation matrix, and redefine the monitoring area based on the positions of the three monitoring wells; and send the newly defined monitoring area to the client at the same time.

[0013] According to the preferred technical solution, in step S32, the newly designated monitoring area is sent to the client at the same time, the database is matched, the list of enterprises in the area is called, and the list of matching enterprises is preliminarily screened; and the Kendall rank correlation coefficient is used for secondary screening, the preliminarily screened list of matching enterprises is combined with the characteristics of groundwater characteristics for correlation calculation, the matching index is sorted in order, and the list of the top three enterprises is sent to the client.

[0014] The preferred technical solution is to set up a correlation calculation model; S41 model establishment: a correlation evaluation model for groundwater pollution source tracing is established based on the correlation matrix and Kendall rank correlation coefficient; S42 model training: the groundwater environmental pollution accident records of park enterprises collected in the database; information such as the decision letter of the park enterprise ordering the rectification of groundwater violations are input into the model, the degree of correlation is sorted by the least squares method, and the correlation weight ranking is set; the correlation between monitoring wells, the characteristics of characteristic objects and the correlation and performance between characteristic object concentrations and enterprises are trained; S43 model evaluation: intelligent training and learning based on neural networks, the data in the inversion model training, and the training results and training process of the correlation evaluation model are evaluated and optimized.

[0015] In a preferred technical solution, the setting of the correlation calculation model includes:

[0016] S41. Model establishment: Establish a correlation evaluation model for groundwater pollution source tracing based on the correlation matrix and Kendall rank correlation coefficient;

[0017] The correlation between monitoring well data was calculated using the Pearson correlation coefficient. and monitoring wells of The monitoring data are and , the Pearson correlation coefficient calculation formula is:

[0018] ,

[0019] in , ; is the number of groundwater monitoring wells, For the number of enterprises, construct a Matrix ; Elements in the matrix Indicates the Data item and The correlation coefficient between the data items;

[0020] Two variables and There are Observations , , …, ; First of all, and Sort them separately and get their ranks and , Kendall rank correlation coefficient The calculation formula is:

[0021] ,

[0022] in For all data pairs and middle, and Subtract the logarithm of the opposite sign from the logarithm of the same sign;

[0023] The Kendall rank correlation coefficient is combined with the correlation matrix to form a comprehensive correlation evaluation index, which is used to screen enterprises and monitoring wells with high correlation with pollution incidents;

[0024] S42. Model training: Input the groundwater pollution accident records of enterprises in the park collected from the database, as well as the decision documents on the rectification of groundwater violations by enterprises in the park, into the model, sort the degree of correlation using the least squares method, and set the correlation weight ranking; train the correlation between monitoring wells, the correlation between characteristic substance characteristics and characteristic substance concentrations and enterprises, and the performance;

[0025] Specifically, by minimizing the loss function and adjusting the model parameters, the model can accurately rank the correlations between monitoring wells, the characteristics of characteristic substances, and the correlations between characteristic substance concentrations and enterprises, and construct a loss function:

[0026] ,

[0027] in, is the sample size, and the correlation prediction value is , the actual dependencies are marked as ;

[0028] S43. Model Evaluation: Conduct intelligent training and learning based on neural networks, invert the data during model training, and evaluate and optimize the training results and training process of the correlation evaluation model. Specifically:

[0029] (1) Neural network intelligent training and learning: A multi-layer perceptron neural network structure is used to intelligently train and learn the correlation evaluation model, and the training data set is divided into a training set, a validation set, and a test set;

[0030] (2) Inversion and optimization:

[0031] Inverse the model training results based on the test set data, that is, compare and analyze the correlation results predicted by the model with the actual situation:

[0032] Accuracy: ;

[0033] Recall: ;

[0034] F1 value: ;

[0035] in, For a real example, For a false positive example, A false counterexample.

[0036] A further technical solution is to conduct on-site inspections to confirm whether there are risks based on the list of the top three enterprises screened in step S32, conduct emission tests on the risky enterprises, and determine that the pollution source is a groundwater pollution source in the park.

[0037] According to a further technical solution, a real-time monitoring device may be provided in the monitoring well, and the real-time monitoring device includes a pH sensor; a heavy metal concentration sensor: an electrochemical method: determining the concentration of heavy metal ions in the solution by measuring the potential difference between the electrode and the heavy metal ions; a characteristic organic pollutant monitoring probe: collecting images in the detection well in real time, transmitting the images to the server for image processing, and monitoring turbidity and other characteristic pollutants in real time.

[0038] In the preferred technical solution, the real-time monitoring device communicates with the data acquisition module of the server to process the data; the data collected by the real-time monitoring device is evaluated through a correlation evaluation model, and the correlation degree and weight of the collected data are optimized; and the correlation information is regularly sent to the client.

[0039] The present invention discloses a method for rapid source tracing of groundwater pollution in a chemical park. The method can collect data through existing groundwater environmental monitoring wells in the park, with low cost and simple operation. The basic data of enterprises and environmental monitoring data are analyzed and matched to obtain a list of enterprises suspected of being pollution sources in the park. On this basis, the pollution sources in the park can be further locked through analysis and judgment of the monitoring data of suspected pollution sources and on-site inspections, and the source tracing of groundwater pollution in the park can be completed. The use of this method makes the source tracing of groundwater pollution in the chemical park more efficient and less costly. By setting up a correlation evaluation system, the correlation between monitoring wells and collected data and pollution source enterprises can be further improved, the accuracy of matching enterprises can be improved, the scope of secondary inspections can be narrowed, and the efficiency of source tracing can be improved. In the source tracing process, the area can be quickly locked through correlation judgment, and multi-channel synchronous parallel processing can be carried out through manual inspection of the park, self-inspection of enterprises and comparison of test results to improve the efficiency and accuracy of source tracing and narrow the scope of pollution impact. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 : Schematic diagram of the present invention. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0043] Example 1

[0044] A method for rapid source tracing of groundwater pollution in a chemical park, which collects data through existing groundwater environmental monitoring wells, and enters the data and pollution-related data collected from enterprises and parks into a database through a server; the divided monitoring areas containing monitoring well information are also stored and put into the database; a data acquisition module is set to process the collected data; a comparison is performed through a data processing module; a threshold is set, and when the detection data exceeds the set threshold, the information processing module sends a trigger alarm instruction, and the monitoring well information related to the detection data is sent to the client; after the client receives the alarm instruction, an alarm prompt and the monitoring area are sent to the client as a preliminary judgment result; the staff uses the information collected by the client to conduct secondary monitoring data analysis and on-site inspections to further identify the pollution source in the park and fully trace the source. The specific steps are as follows:

[0045] S1. Collect basic data: collect enterprises and pollution-related information related to groundwater pollution in the chemical park, and establish a database; park enterprise situation: enterprise list; enterprise environmental impact report form; enterprise completion environmental protection acceptance monitoring report, the enterprise environmental impact report form includes the enterprise characteristic exceeding standard factors and the enterprise's groundwater characteristics; park situation: park enterprise pollutant emission declaration registration form; park enterprise environmental statistical report; park area hydrogeological survey report; park engineering geological survey report; park and enterprise floor plan; park boundary range map; park existing groundwater environment monitoring well information, the monitoring well information includes: coordinates, well depth, water level and monitoring data; park groundwater environment status survey report; park enterprise groundwater environment pollution accident record; the database is set on the server; the server also includes a client with communication connection.

[0046] S2. Delineate monitoring areas: Combined with satellite images, the park and enterprise layout, park boundary map, and groundwater environment monitoring well locations are gridded and delineated. Each monitoring area contains at least one existing groundwater environment monitoring well.

[0047] S3 carries out data matching: regular monitoring of groundwater is carried out through existing groundwater environmental monitoring wells, and the test results are uploaded and matched and saved in combination with the database in S1; a preliminary judgment result is obtained based on the data matching results; S31 the test data is processed by the data acquisition module in the server; S32 the test data is compared with the information in the database through the information processing module of the server; when the test data exceeds the set threshold, the information processing module sends a trigger alarm instruction and sends the monitoring well information related to the test data to the client; S33 after the client receives the alarm instruction, it issues an alarm prompt.

[0048] S4. Based on the determination result in S3, a secondary determination is conducted, and an on-site survey is conducted to confirm whether there is a risk. The discharge of the risk enterprise is tested, and the pollution source is determined to be a groundwater pollution source in the park. Based on the result of the secondary determination, a self-inspection alert is sent to the enterprise client in the area determined to be polluted, prompting the enterprise to conduct a self-inspection and report the self-inspection results. At the same time, pipeline water is extracted and sent for inspection. The inspection results can be sent to both the park management client and the enterprise client.

[0049] S5. Enterprises can quickly detect leaks based on self-inspection alarms and trace the source. If no pollutant leakage is found in the enterprise's self-inspection, the pollutant-emitting enterprise can be quickly identified based on the comparison of the inspection results and the main pollutant detection results to achieve accurate traceability.

[0050] This embodiment collects complete information on enterprises and parks related to pollutants and uses existing monitoring wells for data collection. It does not require complex detection methods, is simple to operate, and is low-cost.

[0051] Example 2

[0052] On the basis of Example 1, by setting a correlation matrix in the information processing module, the two monitoring wells with the highest correlation with the detection data of the monitoring well are obtained, and the monitoring area is redefined according to the positions of the three monitoring wells; and the newly defined monitoring area is sent to the client at the same time; while the newly defined monitoring area is sent to the client, the database is matched, the list of enterprises in the area is called, and the list of matching enterprises is preliminarily screened; and the Kendall rank correlation coefficient is used for secondary screening, the preliminarily screened list of matching enterprises is combined with the characteristics of groundwater characteristics for correlation calculation, the matching index is sorted in order, and the top three enterprise lists are sent to the client.

[0053] Take the case where a monitoring well in a certain industrial park detected an abnormally high concentration of benzene series, reaching 0.08 mg / L, as an example:

[0054] S1. Collect basic data

[0055] Collect detailed basic information about enterprises within the park: including company name, unified social credit code, registered address, actual production address, legal representative, contact information, establishment date, and production scale. Enterprise Environmental Impact Report: Obtain each enterprise's environmental impact report and extract the enterprise's characteristic exceeding-standard factors and groundwater characteristics. Park Information: Pollutant Emission Registration Form: Collect pollutant emission registration forms from enterprises in the park to obtain information on the types, quantities, and destinations of pollutants reported to the ecological and environmental authorities. Environmental Statistics: Obtain environmental statistics from enterprises in the park to analyze the overall pollutant emission trends and structure of the park, including total emissions of conventional pollutants such as chemical oxygen demand, ammonia nitrogen, sulfur dioxide, and nitrogen oxides, as well as characteristic pollutants such as benzene and heavy metals. Hydrogeological Survey Report: A hydrogeological survey report for the park's area is provided. Engineering Geological Survey Report: Document the park's engineering geological conditions. Floor Plan and Boundary Map: Draw detailed floor plans of the park and enterprises, noting the locations of enterprise workshops, warehouses, wastewater treatment facilities, and solid waste storage areas. Define a boundary map of the park to define the boundary between the park and the surrounding environment. Groundwater environmental monitoring well information: and monitoring data (including benzene concentrations and other pollutant concentrations). Environmental status survey report: Compile the park's groundwater environmental status survey report to understand the park's groundwater environmental baseline values, historical pollution status, and other information. Pollution accident records and rectification decisions: Record groundwater pollution accidents at park enterprises.

[0056] All the above data are stored in a server equipped with a high-performance processor, and a Web-based client application is developed for park management departments, enterprises and other users to access through the network to realize data query, upload and download functions.

[0057] S2. Define monitoring areas

[0058] Using satellite imagery, we gridded the layout of the park and its enterprises, the park boundaries, and the locations of groundwater environmental monitoring wells. Centered on the monitoring well where excessive benzene concentrations were detected, and taking into account groundwater flow patterns, we initially defined an initial circular monitoring area. Furthermore, taking into account the distribution of enterprises within the park and the topographical characteristics, we adjusted this area appropriately to align its boundaries with prominent geographical landmarks such as enterprise boundaries and roads. Ultimately, we determined the number of enterprises covered by the monitoring area.

[0059] S3. Carry out data matching

[0060] S31. Data processing: The data acquisition module in the server is written in Python and uses the pandas library to process the regularly monitored groundwater data.

[0061] S32. Data comparison and analysis: Compare the processed detection data with the information in the database. When the benzene concentration in the monitoring well exceeds the set threshold, the information processing module of the server immediately sends a trigger alarm instruction.

[0062] Correlation matrix calculation: Establish a correlation matrix between groundwater monitoring well data and enterprise emission data; the matrix elements represent the degree of correlation between the monitoring wells and the enterprises. The Pearson correlation coefficient is used to calculate the degree of correlation, and the formula is: ,in and They are monitoring well data and enterprise emission data, and are their respective means, is the number of data samples. Through calculation, find the two monitoring wells with the highest correlation with the monitoring well detection data.

[0063] Initial screening and secondary screening: Based on the correlation matrix results, a preliminary list of companies that may be related to pollution is screened. Then, a secondary screening is performed using the Kendall rank correlation coefficient, calculated as follows: , where S is the sum of the rank differences of the data pairs, is the number of data pairs. Correlation is calculated based on the characteristics of groundwater features (benzene series), and the matching index is ranked from high to low. The top three companies are ultimately determined and sent to the client.

[0064] S33. Client Alarm Prompt: After receiving the alarm command, the park management department client and the relevant enterprise client will issue multiple alarm prompts. The client application will pop up a red alarm window, display the alarm information, and emit a beeping sound. In addition, an alarm text message will be sent to the relevant person in charge via SMS.

[0065] S4. Secondary determination

[0066] The park’s environmental monitoring department organized professional and technical personnel to conduct a secondary assessment of the top three companies.

[0067] On-site investigation: Conduct detailed on-site surveys of the three companies, inspect key areas such as their production workshops, storage warehouses, and wastewater treatment facilities to check for leakage, and verify the companies' production records, pollutant emission records, and other information.

[0068] Pipeline water extraction and testing: Water samples are collected from pipelines at companies involved in BTEX emissions. Water samples are collected using specialized sampling bottles and protected with protective agents such as copper sulfate to prevent microbial decomposition. The samples are then sent to a third-party testing agency for testing, including the concentrations of BTEX compounds such as benzene, toluene, and xylene.

[0069] Analysis of results: Test results showed that the benzene concentration in the pipe water samples of one enterprise was much higher than the average level of the other two enterprises and the park. Combined with the on-site investigation, this enterprise was preliminarily determined to be the source of pollution. The test results were sent to both the park management client and the client of the enterprise.

[0070] S5. Enterprise self-inspection and traceability

[0071] After receiving the self-inspection alert, the company immediately launched the internal investigation procedure.

[0072] Equipment inspection: Organize professional maintenance personnel to conduct a comprehensive inspection of production equipment, pipelines, valves, etc., and use ultrasonic leak detectors, infrared thermal imagers and other equipment to detect whether there are any leaks.

[0073] Ledger verification: Verify production ledgers, raw material usage records, wastewater discharge records and other information to check whether there are any data anomalies or illegal discharge behaviors.

[0074] Emergency treatment: If a leak is found, take emergency measures immediately to stop the leak and clean and disinfect the leak area.

[0075] If the company's self-inspection reveals no pollutant leaks, the park's environmental monitoring department will conduct further analysis based on the submitted inspection results and the main pollutant detection results. By comparing the company's production process and raw material composition with the composition and concentration of pollutants in the monitoring wells, and combining them with groundwater flow models, they ultimately determined that benzene leaked from one of the company's reactors due to aging seals, which then seeped through the soil and entered the groundwater, accurately tracing the source.

[0076] Through correlation determination, a new monitoring area is obtained, and the determination scope is narrowed by preliminary screening; through correlation calculation of groundwater characteristic properties, the preliminary determination scope is further narrowed, the scope of secondary determination is reduced, and the cost and manpower input cost are further reduced.

[0077] Example 3

[0078] Based on the above two embodiments, a correlation calculation model is preferably set up to illustrate the establishment, training and evaluation process of the correlation evaluation model for groundwater pollution source tracing:

[0079] 1. Correlation Matrix Construction

[0080] The correlation matrix is ​​used to quantify the degree of correlation between different monitoring well data and between monitoring well data and enterprise emission data. Groundwater monitoring wells and Enterprises, build a Matrix The elements in the matrix Indicates the The first data item (which can be a monitoring well or an enterprise) is associated with the The correlation coefficient between the data items.

[0081] The correlation between monitoring well data is calculated using the Pearson correlation coefficient. and monitoring wells of The monitoring data are and , the Pearson correlation coefficient calculation formula is: ,in , .

[0082] For the correlation between monitoring well data and enterprise emission data, the Pearson correlation coefficient is also used, and the enterprise emission data is regarded as another set of time series data for calculation. The emission data of a certain pollutant is , monitoring wells The corresponding pollutant monitoring data is , then the correlation coefficient between them is calculated in the same way as above.

[0083] 2. Introduction of Kendall rank correlation coefficient

[0084] The Kendall rank correlation coefficient is used to measure the degree of rank correlation between two variables. In groundwater pollution tracing, it is particularly suitable for processing non-numeric data or data with many repeated values, such as the association between corporate violation records and monitoring well pollution incidents.

[0085] Suppose two variables and There are Observations , , …, First, and Sort them separately and get their ranks and Kendall rank correlation coefficient The calculation formula is:

[0086] ,in For all data pairs and middle, and Subtract the logarithm of the opposite sign from the logarithm of the same sign.

[0087] The Kendall rank correlation coefficient is combined with the correlation matrix to form a comprehensive correlation evaluation index, which is used to screen enterprises and monitoring wells with high correlation with pollution incidents.

[0088] 3. Model training

[0089] S41. Data input

[0090] The information collected in the database, such as the records of groundwater pollution accidents of enterprises in the industrial park and the decisions ordering enterprises in the industrial park to correct groundwater violations, is sorted and pre-processed and converted into a format suitable for model training.

[0091] For pollution accident records, we extract information such as the time of the accident, the type of pollutants involved, the scope of pollution impact, and treatment measures; for illegal behavior decisions, we extract information such as the name of the illegal enterprise, the facts of the illegality, and the content of the punishment. At the same time, we associate the corresponding monitoring well data with the enterprise emission data to form a training data set.

[0092] S42, least squares sorting

[0093] The least square method is used to sort the degree of correlation. Assume that the correlation prediction value output by the model is The actual relevance mark (such as whether it is a pollution source enterprise) is , construct the loss function ( is the sample size).

[0094] By minimizing the loss function and adjusting the model parameters, the model can accurately rank the correlations between monitoring wells, the characteristics of characteristic substances, and the correlations between characteristic substance concentrations and enterprises. For example, by optimizing the correlation scores between different enterprises and pollution monitoring wells using the least squares method, the actual pollution source enterprises are ranked higher in terms of correlation.

[0095] S43. Correlation weight setting

[0096] Based on the actual situation and historical data of the chemical park, weights are set for different types of correlations (such as spatial correlation between monitoring wells, concentration correlation between enterprise emissions and monitoring well data, and the correlation between enterprise violation records and pollution incidents).

[0097] For example, if historical data indicates that corporate violations are important indicators for pollution source tracing, a higher weight will be assigned to the correlation between corporate violations and pollution incidents, calculated using the Kendall rank correlation coefficient. If the spatial proximity of monitoring wells significantly influences pollution spread, a corresponding weight will be assigned to the spatial correlation of monitoring wells. By continuously adjusting weights, the model's accuracy in tracing actual pollution sources can be optimized.

[0098] 4. Model Evaluation

[0099] (1) Neural network intelligent training and learning

[0100] Use neural network structures such as multi-layer perceptron (MLP) to intelligently train and learn the relevance evaluation model. Divide the training data set into a training set, a validation set, and a test set (e.g., a ratio of 7:1:2).

[0101] During training, the neural network calculates predictions through forward propagation and updates network parameters through backpropagation algorithms (such as gradient descent), continuously adjusting the model's ability to predict correlations. Simultaneously, the model is evaluated in real time using a validation set to avoid overfitting.

[0102] (2) Inversion and optimization

[0103] Inverse the model training results based on the test set data, that is, compare and analyze the correlation results predicted by the model with the actual situation. Calculate the model's accuracy, recall rate, F1 value and other evaluation indicators, for example:

[0104] Accuracy:

[0105] Recall:

[0106] F1 value:

[0107] in, are true positives (the number of correctly predicted pollution sources), are false positives (number of incorrectly predicted pollution sources), is the number of false negative examples (the number of sources incorrectly predicted as non-polluting sources).

[0108] Based on the evaluation results, the training process and parameters of the correlation evaluation model are optimized, such as adjusting the number of layers and nodes of the neural network, resetting the correlation weights, improving the data preprocessing method, etc., to continuously improve the accuracy and reliability of the model in groundwater pollution tracing.

[0109] By establishing a correlation evaluation model for groundwater pollution source tracing, the correlation between monitoring wells, the correlation between characteristic substance properties and characteristic substance concentrations and enterprises can be effectively determined, greatly improving the accuracy of tracing the source of pollution enterprises through monitoring well data; further narrowing the scope of preliminary determination, reducing the scope of secondary determination, further reducing costs, and reducing manpower input costs.

[0110] Example 4

[0111] The concentration of heavy metals is determined by measuring the potential difference between the heavy metal ions in the solution and the electrode using a heavy metal concentration sensor in a real-time monitoring device;

[0112] The characteristic organic pollutant monitoring probe is used to collect real-time images of the test well, which are then transmitted to the server for image processing to monitor turbidity and other characteristic pollutants in real time.

[0113] Building on the three aforementioned embodiments, real-time monitoring devices are installed in monitoring wells and in key monitoring wells throughout the park. A pH sensor uses a high-precision, industrial-grade pH sensor, installed 2 meters below the water surface in the monitoring well. It automatically collects pH data every 10 minutes and transmits the data to a server via a 4G wireless communication module. When the pH value exceeds the normal range (6.5-8.5), the server immediately issues an alarm. A heavy metal concentration sensor uses an electrochemical method to simultaneously measure the concentrations of multiple heavy metal ions. The sensor is immersed in the monitoring well and measures data every 30 minutes. The data is initially processed by the sensor's built-in data processing module and then transmitted to the server via a wireless network. A characteristic organic pollutant monitoring probe uses a high-definition camera to capture real-time images of the well. The camera is fixed above the wellhead with a bracket, facing the wellhead, and captures images every 5 minutes. The images are transmitted to the server via a 5G network. The server processes the images using a deep learning image recognition algorithm to monitor turbidity and the presence of characteristic pollutants such as oil film and floating debris in real time. When an obvious oil film is detected in the image, the server immediately issues an alarm.

[0114] The real-time monitoring device communicates with the server's data collection module, processes the data, and evaluates the data collected using a correlation evaluation model. The correlation level and weight of the collected data are optimized to improve the accuracy of the correlation evaluation. This correlation data is then regularly sent to the client. When the collected data exceeds a set threshold, a real-time notification is provided, slowing the spread of pollutants and quickly identifying the pollution area, achieving rapid source tracing.

[0115] Example 5

[0116] Based on the above four embodiments, the pollution area is redefined through the detection results and correlation judgment, and the pollutant-related enterprises are quickly screened through correlation; through manual verification of the park, self-inspection of enterprises and comparison of detection results, multi-channel synchronous parallel processing is carried out to improve the efficiency and accuracy of tracing the source and narrow the scope of pollution impact.

[0117] Of course, without departing from the spirit and essence of the present invention, technicians familiar with the field should be able to make various corresponding changes and modifications based on the present invention, but these corresponding changes and modifications should all fall within the scope of protection of the claims attached to the present invention.

Claims

1. A method for rapid tracing of groundwater pollution in a chemical park, characterized by: The following steps are involved: S1. Collect basic data: Collect information on enterprises and pollution related to groundwater pollution in the chemical park and establish a database; S2. Delineate the monitoring area: Delineate the monitoring area based on the coordinates of existing groundwater environment monitoring wells; grid the park and enterprise layout, park boundary map, and groundwater environment monitoring well locations using satellite imagery for delineation; S3. Conduct data matching: Regularly monitor groundwater through existing groundwater environmental monitoring wells, upload the test results, and match and save them in combination with the database in S1; based on the data matching results, draw a preliminary judgment result; S31, the detection data is processed by the data acquisition module in the server; S32. The information processing module of the server compares the detection data with the information in the database. When the detection data exceeds a set threshold, the information processing module sends a trigger alarm instruction and sends the monitoring well information related to the detection data to the park management client. The information processing module obtains the two monitoring wells with the highest correlation with the detection data of the monitoring wells based on the correlation matrix, and re-defines the monitoring area based on the locations of the three monitoring wells; and simultaneously sends the newly demarcated monitoring area to the park management client. S33, after receiving the alarm instruction, the park management client issues an alarm prompt; S4. Based on the determination results in S3, conduct a secondary determination to determine the source of groundwater pollution in the park; Based on the results of the secondary assessment, a self-inspection alert is sent to the enterprise client in the area where the pollution is determined, prompting the enterprise to conduct a self-inspection and report the results. At the same time, water samples from the pipeline are taken for inspection. The inspection results can be sent to both the park management client and the enterprise client at the same time. When the newly designated monitoring area is sent to the park management client, the database is matched, the list of enterprises in the area is called, and the list of matching enterprises is preliminarily screened; A secondary screening is performed using the Kendall rank correlation coefficient. The list of initially screened matching companies is combined with the characteristics of groundwater features for correlation calculation. The matching index is sorted in order, and the list of the top three companies is sent to the park management client and the top three company clients. S5. Enterprises can quickly detect leaks based on self-inspection alarms and trace the source. If no pollutant leakage is found in the enterprise's self-inspection, the pollutant-emitting enterprise can be quickly identified based on the comparison of the inspection results and the main pollutant detection results to achieve accurate traceability.

2. The method according to claim 1, characterized in that The database in S1 includes: Enterprise information in the park: enterprise list; enterprise environmental impact report; enterprise completion environmental protection acceptance monitoring report, the enterprise environmental impact report includes the enterprise's characteristic exceeding standard factors and the enterprise's groundwater characteristics; Park information: Registration form for pollutant emissions from enterprises in the park; Environmental statistics report for enterprises in the park; Hydrogeological survey report for the park area; Engineering geological survey report for the park; Layout plan of the park and enterprises; Boundary map of the park; Information on existing groundwater environment monitoring wells in the park. The monitoring well information includes: coordinates, well depth, water level and monitoring data; park groundwater environmental status investigation report; park enterprise groundwater environmental pollution accident records; the database is set on the server; the server also includes a client with communication connection.

3. The method according to claim 1 or 2, characterized in that: Set up the correlation calculation model; The setting of the correlation calculation model includes: S3A, Model establishment: Establish a correlation evaluation model for groundwater pollution source tracing based on the correlation matrix and Kendall rank correlation coefficient; The correlation between monitoring well data is calculated using the Pearson correlation coefficient. The k-time monitoring data of monitoring well i and monitoring well j are {x i1 , x i2 ,…,x ik } and {x j1 , x j2 ,…,x jk }, the Pearson correlation coefficient calculation formula is: in r ij Represents the correlation coefficient between the i-th data item and the j-th data item; The calculation formula of Kendall's rank correlation coefficient τ is: Among them, r i and s i (i=1,2,…,n) is the rank of two variables; S is the rank of all data pairs (r i ,s i ) and (r j , s j )(i<j), (r i -r j ) and (s i -s j ) is the logarithm of the same sign minus the logarithm of the opposite sign; The Kendall rank correlation coefficient is combined with the correlation matrix to form a comprehensive correlation evaluation index, which is used to screen enterprises and monitoring wells with high correlation with pollution incidents; S3B, Model Training: The groundwater pollution accident records of enterprises in the park collected from the database and the decision documents ordering enterprises in the park to correct groundwater violations are input into the model. The correlation degree is ranked using the least squares method and the correlation weight ranking is set. The correlation between monitoring wells, the correlation between characteristic substance characteristics and characteristic substance concentrations and enterprises, and performance training are carried out. Specifically, by minimizing the loss function and adjusting the model parameters, the model can accurately rank the correlations between monitoring wells, the characteristics of characteristic substances, and the correlations between characteristic substance concentrations and enterprises, and construct a loss function: Among them, N is the number of samples, and the correlation prediction value is The actual correlation is labeled y; S3C, Model Evaluation: Conduct intelligent training and learning based on neural networks, invert data during model training, and evaluate and optimize the training results and training process of the correlation evaluation model. Specifically: (1) Neural network intelligent training and learning uses a multi-layer perceptron neural network structure to intelligently train and learn the correlation evaluation model, and divides the training data set into a training set, a validation set, and a test set; (2) Inversion and optimization: Inverse the model training results based on the test set data, that is, compare and analyze the correlation results predicted by the model with the actual situation: Accuracy: Recall: F1 value: Among them, TP is a true positive example, FP is a false positive example, and FN is a false negative example.

4. The method according to claim 3, wherein: Based on the list of the top three enterprises selected in step S3B, an on-site survey is conducted to confirm whether there are risks, and discharge tests are conducted on the risky enterprises to determine that the pollution source is a groundwater pollution source in the industrial park.

5. The method according to claim 4, characterized in that: A real-time monitoring device is provided in the monitoring well, and the real-time monitoring device includes a pH sensor, a heavy metal concentration sensor and a characteristic organic pollutant monitoring probe; The concentration of heavy metals is determined by measuring the potential difference between the heavy metal ions in the solution and the electrode using a heavy metal concentration sensor in a real-time monitoring device; The characteristic organic pollutant monitoring probe is used to collect real-time images of the test well, which are then transmitted to the server for image processing to monitor turbidity and other characteristic pollutants in real time.

6. The method according to claim 5, characterized in that: The real-time monitoring device communicates with the data collection module of the server to process the data; the data collected by the real-time monitoring device is evaluated through a correlation evaluation model, and the correlation degree and weight of the collected data are optimized; and the correlation information is regularly sent to the park management client.

Images