A river pollutant tracing method and system thereof

Abstract

The application provides a river pollutant tracing method and system. The method comprises the following steps: constructing a potential pollution source feature fingerprint library of a fusion LSTM time series feature extraction model, identifying an abnormal water quality fingerprint of a monitored river section, collecting an upstream water sample and detecting a water quality fingerprint if an abnormality is identified, comparing the water quality fingerprint with the abnormal water quality fingerprint, determining a target river section according to a comparison result, collecting samples from all enterprises of the target river section in real time, comparing detection result data with detection result data of the abnormal water sample, determining a potential pollution source according to a comparison result, determining a high-matching candidate source from the potential pollution source by using the LSTM time series feature extraction model, performing spatial migration verification on the high-matching candidate source, and confirming a pollution source according to a verification result. The application can effectively improve the accuracy, efficiency and result reliability of river pollutant tracing, and effectively cope with complex river channel multi-source pollution, intermittent discharge and other scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of river pollutant tracing, in particular to a river pollutant tracing method and system. BACKGROUND

[0002] River pollutant tracing is one of the core technologies of water environment management. Various methods have been developed in the prior art: for example, some studies identify pollutant sources by collecting stable isotope ratios in water bodies, and use the "fingerprint" characteristics of isotopes to distinguish different emission sources. However, such methods focus on a single dimension of characteristics and are difficult to cope with mixed pollution scenarios from multiple sources. Some technologies store the types and concentration ratios of pollutants emitted by enterprises by constructing a pollution source feature database, and use similarity algorithms to compare abnormal water samples with database features to lock in pollution sources. However, these technologies ignore the time sequence variation of enterprise emissions (such as periodic emissions caused by production cycles and trend fluctuations caused by changes in equipment operating status), making it difficult to capture dynamic characteristics such as intermittent emissions from factories, and leading to misjudgment or missed judgment due to static feature matching bias. In addition, for pollution range positioning, the prior art often uses multi-point sampling along the river and analyzes the concentration gradient to trace the source. However, this method requires sequential sampling and comparison, which is time-consuming and labor-intensive, especially in complex river sections with tributaries, resulting in low efficiency and difficulty in accurately defining the boundaries of pollution diffusion. At the same time, the verification methods of the prior art are relatively single, relying mainly on static data feature matching, lacking spatial verification of the dynamic migration process of pollutants, and making it difficult to confirm the direct causal relationship between pollution sources and pollution events.

[0003] In summary, the prior art has problems such as insufficient multi-dimensional feature fusion, lack of time sequence rule mining, inefficient pollution range positioning, and single verification method, making it difficult to meet the demand for rapid and accurate tracing of pollutants in complex river channels. Therefore, there is an urgent need for an efficient tracing method that integrates time sequence feature extraction, intelligent hierarchical investigation, and spatial trajectory verification. SUMMARY

[0004] The purpose of the present application is to provide a river pollutant tracing method and system that can solve the problems of insufficient multi-dimensional feature fusion, lack of time sequence rule mining, inefficient pollution range positioning, and single verification method in the prior art, forming a complete technical chain of "feature storage- abnormality identification-range locking-source screening-verification confirmation", significantly improving the accuracy, efficiency, and reliability of river pollutant tracing results by integrating time sequence feature extraction, intelligent hierarchical investigation, and spatial trajectory verification, and effectively dealing with complex river multi-source pollution and intermittent emission scenarios.

[0005] The present application also provides a river pollutant tracing method, comprising the following steps:

[0006] Constructing a potential pollution source feature fingerprint library that integrates an LSTM time sequence feature extraction model;

[0007] abnormal water quality fingerprint identification is performed on the monitoring river section;

[0008] If an abnormality is identified, upstream water samples are collected and water quality fingerprints are obtained through detection, and the target river section is determined according to the comparison result;

[0009] All enterprises in the target river section are sampled in real time, and the detection result data is compared with the detection result data of the abnormal water sample, and the potential pollution source is determined according to the comparison result;

[0010] The LSTM time series feature extraction model is used to determine a high-matching candidate source from the potential pollution source;

[0011] The high-matching candidate source is verified by spatial migration, and the pollution source is confirmed according to the verification result.

[0012] Optionally, in the river pollution source tracing method described in the present application, the potential pollution source feature fingerprint library constructed by the LSTM time series feature extraction model comprises:

[0013] Pollution samples of each enterprise in the monitoring river section at different pollution periods are collected to obtain pollution feature data, including pollution types, pollution concentration ratios, pollution concentration time series data, and stable isotope ratios;

[0014] The pollution concentration time series data in the sample is trained by using a preset LSTM algorithm to extract time series feature vectors, including periodic feature data and trend feature data, and an LSTM time series feature extraction model is obtained;

[0015] The pollution feature data, the time series feature vectors, and the LSTM time series feature extraction model are stored in a database to generate a potential pollution source feature fingerprint library.

[0016] Optionally, in the river pollution source tracing method described in the present application, the abnormal water quality fingerprint identification on the monitoring river section comprises:

[0017] The water quality fingerprints of the monitoring river section are collected in real time, and the fluorescence peak position offset, the new peak intensity ratio, and the abnormal duration are obtained by combining the water quality fingerprints of the normal water quality;

[0018] If the fluorescence peak position offset is greater than a preset offset threshold, the new peak intensity ratio is greater than a preset intensity ratio threshold, and the abnormal duration is greater than a preset time threshold, it is determined that the water quality is abnormal.

[0019] Optionally, in the river pollution source tracing method described in the present application, if an abnormality is identified, upstream water samples are collected and water quality fingerprints are obtained through detection, and the target river section is determined according to the comparison result, comprising:

[0020] If an anomaly is identified, the water quality fingerprint of the upstream section is collected and compared with the water quality fingerprint of the section where the water quality anomaly is determined;

[0021] If the comparison is successful, sampling continues upstream;

[0022] If the comparison is unsuccessful, water samples are re-collected at the midpoint between the upstream section and the section where the water quality anomaly is determined and compared with the water quality fingerprint of the section where the water quality anomaly is determined;

[0023] According to the water quality fingerprint comparison result, the same cycle of pollution tracing is carried out until the similarity change rate of the water quality fingerprints of two adjacent sampling points exceeds the preset change rate threshold;

[0024] The river section between the two adjacent sampling points is taken as the target river section.

[0025] Optionally, in the river pollution source tracing method described in the present application, the detection result data of the target river section is compared with the detection result data of the abnormal water sample, and the potential pollution source is determined according to the comparison result, including:

[0026] The detection result data includes the type of pollutants, the concentration ratio of pollutants, and the stable isotope ratio;

[0027] The type of pollutants, the concentration ratio of pollutants, and the stable isotope ratio of the enterprise water sample and the abnormal water sample are compared one by one;

[0028] The enterprise whose type of pollutants, concentration ratio of pollutants, and stable isotope ratio are all successfully compared is selected as the potential pollution source.

[0029] Optionally, in the river pollution source tracing method described in the present application, the high-matching candidate source is determined from the potential pollution source using the LSTM time series feature extraction model, including:

[0030] Obtain the time series data of the concentration of pollutants of the potential pollution source water sample and the abnormal water sample, and input it into the LSTM time series feature extraction model to obtain the periodicity feature data and trend feature data of the discharge;

[0031] The periodicity feature data and trend feature data of the discharge of the potential pollution source water sample and the abnormal water sample are compared, and the potential pollution source that successfully compares is taken as the high-matching candidate source.

[0032] Optionally, in the river pollution source tracing method described in the present application, the high-matching candidate source is verified by spatial displacement, and the pollution source is confirmed according to the verification result, including:

[0033] Starting from the location of the high-matching candidate source, the concentration time series data of each section downstream along the river is collected;

[0034] inputting the concentration time series data of pollutants of potential pollution sources, the pollutant species, the pollutant concentration ratio and the stable isotope ratio into a pre-constructed hydrodynamic-water quality coupling model to generate concentration time series data of pollutants at different spatial points, and comparing the concentration time series data with the actually monitored concentration time series data of each section;

[0035] if the comparison passes, the high-matching candidate source is determined as the pollution source.

[0036] In a second aspect, the present application provides a river pollutant source tracing system, which comprises a memory and a processor, wherein the memory stores a program of a river pollutant source tracing method, and the program of the river pollutant source tracing method is executed by the processor to realize the following steps:

[0037] constructing a potential pollution source feature fingerprint library by using an LSTM time series feature extraction model;

[0038] identifying an abnormal water quality fingerprint of a monitored river section;

[0039] if the identification is abnormal, collecting an upstream water sample and detecting a water quality fingerprint, comparing the water quality fingerprint with the abnormal water quality fingerprint, and determining a target river section according to a comparison result;

[0040] collecting real-time samples of all enterprises in the target river section, comparing the detection result data with the detection result data of the abnormal water sample, and determining a potential pollution source according to a comparison result;

[0041] determining a high-matching candidate source from the potential pollution source by using the LSTM time series feature extraction model;

[0042] verifying the high-matching candidate source by using spatial migration, and confirming the pollution source according to a verification result.

[0043] Optionally, in the river pollutant source tracing system described in the present application, the construction of the potential pollution source feature fingerprint library by using the LSTM time series feature extraction model comprises:

[0044] collecting pollutant samples of each enterprise in the monitored river section at different pollution periods to obtain pollutant feature data, including pollutant species, pollutant concentration ratio, pollutant concentration time series data and stable isotope ratio;

[0045] training the pollutant concentration time series data in the samples by using a preset LSTM algorithm to extract time series feature vectors, including periodic feature data and trend feature data, and obtaining an LSTM time series feature extraction model;

[0046] storing the pollutant feature data, the time series feature vectors and the LSTM time series feature extraction model in a database to generate the potential pollution source feature fingerprint library.

[0047] Optionally, in the river pollutant tracing system provided in the present application, the abnormal water quality fingerprint identification of the monitoring river section comprises:

[0048] The water quality fingerprint of the monitoring river section is collected in real time, and the fluorescence peak position offset, the new peak intensity ratio and the abnormal duration are obtained by processing the water quality fingerprint of the normal water quality;

[0049] If the fluorescence peak position offset is greater than the preset offset threshold, the new peak intensity ratio is greater than the preset intensity ratio threshold, and the abnormal duration is greater than the preset time threshold, it is determined that the water quality is abnormal.

[0050] As can be seen from the above, the river pollutant tracing method and system provided in the present application systematically construct a potential pollution source fingerprint library by fusing static characteristics (pollutant types, concentration ratios, stable isotope ratios) and dynamic time sequence rules (discharge periodicity characteristics, trend characteristics), and in the pollution event period, comprehensive comparison is carried out by relying on multi-dimensional fingerprint information (covering pollutant composition, quantitative proportion, isotope identification and time sequence characteristic vector), and the pollutant migration trajectory is analyzed and deduced by combining the water dynamics-water quality coupling model. Even in the face of complex scenes such as coexistence of multiple suspected enterprises and intermittent pollution, the present application can effectively exclude interference sources and accurately lock the real pollution source, solving the problems of low tracing efficiency and insufficient matching accuracy of traditional methods in complex pollution situations.

[0051] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and obtained by the structure particularly pointed out in the written description and the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0053] Figure 1 The flowchart of the river pollutant tracing method provided in the embodiments of the present application;

[0054] Figure 2 The flowchart of the river pollutant tracing method provided in the embodiments of the present application;

[0055] Figure 3 The flowchart of the river pollutant tracing method provided in the embodiments of the present application; Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0057] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0058] Please refer to Figure 1 , Figure 1 This is a flowchart of a river pollutant tracing method according to some embodiments of this application. This river pollutant tracing method is used in terminal devices, such as computers and mobile terminals. The river pollutant tracing method includes the following steps:

[0059] S11. Construct a potential pollution source feature fingerprint database that integrates the LSTM temporal feature extraction model;

[0060] S12. Conduct abnormal water quality fingerprint identification on the monitored river section;

[0061] S13. If an anomaly is identified, upstream water samples are collected and water quality fingerprints are obtained. These fingerprints are then compared with the abnormal water quality fingerprints, and the target river section is determined based on the comparison results.

[0062] S14. Conduct real-time sampling of all enterprises in the target river section, compare the detection results with the detection results of abnormal water samples, and determine potential pollution sources based on the comparison results;

[0063] S15. Use the LSTM temporal feature extraction model to determine high-matching candidate sources from potential pollution sources;

[0064] S16. Spatial migration verification is performed on highly matched candidate sources, and the pollution source is confirmed based on the verification results.

[0065] It needs to be explained that the present application realizes the all-round storage of enterprise pollution characteristics by constructing a potential pollution source feature fingerprint library of a fusion LSTM (Long Short Term Memory Network) time sequence feature extraction model, not only including static features such as pollutant types, concentration ratios, stable isotope ratios, but also providing multi-dimensional benchmarks for tracing by extracting dynamic time sequence vectors such as periodicity and trend features of emissions through the LSTM algorithm; the precise identification of water quality anomalies is realized through the joint determination of the fluorescence peak position offset, the new peak intensity ratio and the abnormal duration, avoiding the misjudgment of a single indicator; the intelligent strategy of "upstream sampling promotion + dichotomy supplementary sampling" is used to locate the target river section, combined with the adjacent sampling point water quality fingerprint similarity rate threshold, to quickly narrow down the pollution range and greatly improve the investigation efficiency; after screening potential pollution sources based on the triple comparison of pollutant types, concentration ratios and stable isotope ratios, the high-matching candidate sources are further determined through the LSTM time sequence feature vector comparison, improving the screening accuracy from static features and dynamic rules; finally, the river fine water dynamics-water quality coupling model is used to simulate the migration trajectory of pollutants from the candidate source to the monitoring section, and the simulated concentration time sequence data is compared and verified with the actual monitoring data to establish the direct causal relationship between the pollution source and the pollution event. Form a complete traceability chain of "feature storage- anomaly identification- range locking- source screening- verification confirmation", solve the problems of time sequence feature missing, low matching accuracy and long tracing period of traditional methods, realize the accurate and efficient tracing of river pollutants, and is especially suitable for rivers in industrial clusters, which can efficiently trace complex scenes such as intermittent illegal discharge and multi-source composite pollution.

[0066] Please refer to Figure 2 , Figure 2 is a flow chart of constructing a potential pollution source feature fingerprint library of a river pollutant tracing method in some embodiments of the present application. According to the embodiment of the present application, the potential pollution source feature fingerprint library of the fusion LSTM time sequence feature extraction model comprises:

[0067] S21, collecting the pollution samples of each enterprise in the monitoring river section at different pollution periods to obtain pollutant feature data, including pollutant types, pollutant concentration ratios, pollutant concentration time sequence data and stable isotope ratios;

[0068] S22, training the pollutant concentration time sequence data in the sample by using a preset LSTM algorithm to extract time sequence feature vectors, including periodicity feature data and trend feature data, and obtaining an LSTM time sequence feature extraction model;

[0069] S23, storing the pollutant feature data, time sequence feature vectors and LSTM time sequence feature extraction model into a database to generate a potential pollution source feature fingerprint library.

[0070] It should be noted that by collecting pollution samples of enterprises at different time periods, static characteristics such as types of pollutants, concentration ratios and stable isotope ratios are fused with time sequence characteristics (periodicity, trend) extracted by the LSTM to generate a potential pollution source characteristic fingerprint library, thereby providing a multi-dimensional benchmark for subsequent comparison. Compared with a traditional database that only stores static characteristics, the fingerprint library can capture the time sequence law of enterprise pollution, lay a data foundation for accurate matching of pollution sources, reduce misjudgments caused by ignoring time sequence characteristics, and solve the problem of distinguishing similar pollutants emitted by similar enterprises. The time sequence of pollutant concentration time sequence data is composed of pollutant concentration values collected at fixed time intervals.

[0071] The emission periodicity characteristic data includes peak frequency, peak interval, phase and peak base ratio. The peak frequency refers to the number of emission concentration peak values appearing per unit time, directly reflects the activity degree of periodic emission, and can distinguish continuous pollution and intermittent pollution. The peak interval refers to the time difference between adjacent two emission peaks (for example, a peak appears every 8 hours), which is a core quantitative index of periodicity and can accurately identify fixed interval laws such as daily cycle and weekly cycle. The phase refers to the relative time position of the peak value in the cycle (for example, the peak value appears at 12 o'clock every day), which can distinguish the “time imprint” of different pollution sources (for example, the peak phase of similar factories may differ by 2 hours due to different production shifts). The peak base ratio refers to the ratio of the peak concentration to the baseline concentration (the average concentration in the non-peak period) in the cycle (for example, the peak value is 5 times the baseline value), which can reflect the intensity fluctuation range of periodic emission and enhance the feature recognition degree.

[0072] The trend characteristic data includes peak slope, change rate, trend duration and mutation node. The peak slope refers to the rate of emission concentration rising from the baseline value to the peak value (for example, rising from 1 mg / L to 5 mg / L in 1 hour, the slope is 4 mg / (L·h)), which can reflect the “start-up characteristics” of the pollution process (for example, the slope difference between instantaneous emission and slow accumulation emission). The change rate refers to the rate of concentration change with time in the overall trend (for example, increasing by 0.2 mg / L per day), which quantifies the “steepness” of the trend and distinguishes between slow deterioration and sharp over-standard. The trend duration refers to the duration of the same change trend (increasing / decreasing) (for example, a decreasing trend for 10 consecutive days), which can exclude short-term interference (such as instantaneous leakage) and focus on long-term stable emission characteristics. The mutation node refers to the time point at which the concentration trend changes significantly (for example, the concentration suddenly jumps from 2 mg / L to 8 mg / L on May 5), which can be associated with abnormal events of the pollution source (such as equipment failure, illegal discharge), and enhance the timeliness of the traceability.

[0073] According to the embodiment of the present application, the abnormal water quality fingerprint identification of the monitoring river section comprises:

[0074] Real-time acquisition of the water quality fingerprint of the river section, and processing of the water quality fingerprint of the normal water quality to obtain a fluorescence peak position offset, a new peak intensity ratio and an abnormal duration;

[0075] If the fluorescence peak position offset is greater than a preset offset threshold, the new peak intensity ratio is greater than a preset intensity ratio threshold, and the abnormal duration is greater than a preset time threshold, the water quality is determined to be abnormal.

[0076] It should be noted that the multi-parameter determination of the abnormal water quality of the fluorescence peak position offset, the new peak intensity ratio and the abnormal duration can more sensitively and accurately capture the abnormal water quality than the traditional single index (such as exceeding the concentration of a certain pollutant), reduce the risk of misjudgment caused by background fluctuation or accidental interference, and realize rapid locking of the pollution event.

[0077] According to the embodiment of the present application, if the abnormality is identified, the upstream water sample is collected and the water quality fingerprint is detected, and the water quality fingerprint is compared with the abnormal water quality fingerprint, and the target river section is determined according to the comparison result, comprising:

[0078] If the abnormality is identified, the water quality fingerprint of the upstream section is collected and compared with the water quality fingerprint of the water quality abnormal section;

[0079] If the comparison is successful, the sampling is continued to the upstream;

[0080] If the comparison is not successful, the water sample is re-collected at the two-position between the upstream section and the water quality abnormal section, and compared with the water quality fingerprint of the water quality abnormal section;

[0081] According to the water quality fingerprint comparison result, the same cycle of pollution tracing and investigation is carried out until the similarity change rate of the water quality fingerprints of two adjacent sampling points exceeds a preset change rate threshold;

[0082] The river section between the two adjacent sampling points is taken as the target river section.

[0083] It should be noted that the target river section is located by the strategy of "upstream sampling + two-position supplementary sampling", and the similarity change rate threshold of the adjacent sampling points, which can quickly narrow the pollution source range, and compared with the traditional whole river section sampling without target, the method greatly reduces the sampling amount and analysis cost, and improves the accuracy and efficiency of the target river section positioning.

[0084] According to the embodiment of the present application, the all enterprises of the target river section are sampled in real time, the detection result data is compared with the detection result data of the abnormal water sample, and the potential pollution source is determined according to the comparison result, comprising:

[0085] The detection result data includes the type of the pollutant, the concentration ratio of the pollutant and the stable isotope ratio;

[0086] The enterprise water sample is compared with the pollutant type, pollutant concentration ratio and stable isotope ratio of the abnormal water sample one by one;

[0087] The enterprise with the compared successful pollutant type, pollutant concentration ratio and stable isotope ratio is selected as a potential pollution source.

[0088] It should be noted that, compared with single index comparison, the three comparisons of the pollutant type, concentration ratio and stable isotope ratio can significantly improve the specificity of the pollution source screening (for example, the stable isotope ratio has a "fingerprint" characteristic and is difficult to be tampered with by people), and reduce the misjudgment caused by similar pollutants emitted by different enterprises.

[0089] According to an embodiment of the present application, the high matching candidate source is determined from the potential pollution source by using the LSTM time sequence feature extraction model, comprising:

[0090] The pollutant concentration time sequence data of the potential pollution source water sample and the abnormal water sample are acquired and input into the LSTM time sequence feature extraction model to obtain emission periodicity feature data and trend feature data;

[0091] The emission periodicity feature data and the trend feature data of the potential pollution source water sample and the abnormal water sample are compared, and the potential pollution source with the compared successful is taken as a high matching candidate source.

[0092] It should be noted that, the LSTM model is used to extract the emission periodicity, trend and other time sequence features of the potential pollution source and the abnormal water sample and compare them, so that the high matching candidate source is further screened from the "time law" aspect. Compared with the screening only relying on static features, the time sequence features can capture the behavior law of enterprise pollution (such as periodic emission caused by production shift), exclude interference sources that do not match the time sequence law of abnormal events, and improve the reliability of the candidate source.

[0093] Please refer to Figure 3 , Figure 3 It is a flowchart for confirming a pollution source in a river pollution source tracing method in some embodiments of the present application. According to an embodiment of the present application, the spatial migration verification is performed on the high matching candidate source, and the pollution source is confirmed according to the verification result, comprising:

[0094] S31, taking the position of the high matching candidate source as a starting point, collecting concentration time sequence data of each section along the downstream of the river;

[0095] S32, inputting the pollutant concentration time sequence data, pollutant type, pollutant concentration ratio and stable isotope ratio of the potential pollution source into a pre-constructed water dynamics-water quality coupling model to generate concentration time sequence data of the pollutant at different spatial points, and comparing the concentration time sequence data with the actually monitored concentration time sequence data of each section;

[0096] S33、If the comparison passes, determine that the high matching candidate source is the pollution source.

[0097] It should be noted that the pollutant migration trajectory is simulated by the river water power-water quality coupling model, the concentration time series data (concentration-time curve) of the high matching candidate source and the static characteristics are input into the model, the model is used to generate the concentration-time curve (concentration time series data) of the pollutant at different spatial points (such as 1km, 3km, 5km downstream of the candidate source), and the concentration time series data of each section is compared with the actual monitoring, and the spatial order of the concentration peak value (whether it decreases along the water flow direction) and the decay rate are focused on. If the trends are consistent, it indicates that the spatial migration path of the pollutant is consistent with the logic of the candidate source emission. Through the "theoretical simulation + actual data verification", the misjudgment caused by the "similar characteristics but not same source" caused by the dependence on the characteristics comparison is avoided, and the spatial migration correlation between the finally locked pollution source and the pollution event is ensured, and the accuracy of the tracing result is improved. For example: the water quality fingerprint, static characteristics and dynamic time series vector of enterprise A and the polluted water body are matched, but the river water power-water quality coupling model shows that its emissions need 5 hours to reach the pollution point, and the actual pollution breaks out within 1 hour, and the actual pollution source is the unregistered enterprise B upstream.

[0098] The river fine water power-water quality coupling model can simulate the water power parameters and the pollutant migration trajectory simultaneously. When the model is constructed, first, based on the high-precision topographic data (including cross section, underwater DEM and roughness zoning) of the monitoring river section, hydrological data (upstream flow, downstream water level), pollution source data (location of the discharge outlet, discharge amount and time series characteristics) and meteorological data, a two-dimensional or three-dimensional water power model (such as the shallow water equation) is used to simulate the river flow field (flow velocity, water depth, turbulence intensity), and a water quality module is simultaneously coupled to simulate the convection, diffusion and degradation process of the pollutant, and the key parameters such as roughness and degradation coefficient are calibrated by the measured water level, flow velocity and pollutant concentration data, so that the deviation between the model simulation results and the measured data is controlled within a preset threshold (such as water level error ≤10%), and finally a coupling model capable of accurately reproducing the pollutant migration trajectory is formed.

[0099] According to the embodiment of the application, the method further comprises:

[0100] The historical pollutant characteristic data of the potential pollution source is extracted from the pollution source characteristic fingerprint library, including the pollutant type, the pollutant concentration ratio, the stable isotope ratio and the time series characteristic vector;

[0101] The historical pollutant characteristic data is compared with the real-time pollutant characteristic data of the pollution source collected at present, to obtain the pollutant type matching degree, the pollutant concentration ratio deviation rate, the stable isotope ratio deviation rate and the time series characteristic vector cosine similarity;

[0102] The pollutant concentration proportion deviation rate and the stable isotope ratio deviation rate are weighted and averaged to calculate whether the calculation result is less than or equal to a preset deviation rate threshold value;

[0103] The pollutant species matching degree and the time sequence feature vector cosine similarity are weighted and averaged to calculate whether the calculation result is greater than or equal to a preset matching degree threshold value;

[0104] If one of the above judgments is no, the potential pollution source is subjected to multi-dimensional investigation and prompting.

[0105] It should be noted that the historical pollutant characteristic data in the potential pollution source characteristic fingerprint library is called, the "historical emission characteristics" and the "current emission characteristics" are compared and analyzed, and if the difference is too large, it is necessary to further investigate whether the enterprise has process changes, illegal discharge and the like.

[0106] The stable isotope ratio deviation rate can be calculated by (real-time value - historical value) / historical value x 100%; the pollutant species matching degree is represented by the ratio of the number of characteristic pollutant species common to real-time data and historical data to the number of all characteristic pollutant species in historical data; the pollutant concentration proportion deviation rate: for multiple characteristic pollutants, (real-time proportion - historical proportion) / historical proportion x 100% is calculated respectively, and the calculation results of all characteristic pollutants are added to calculate the average value.

[0107] The application also discloses a river pollutant tracing system, comprising a memory and a processor, the memory stores a river pollutant tracing method program, and the river pollutant tracing method program is executed by the processor to realize the following steps:

[0108] A potential pollution source characteristic fingerprint library is constructed by fusing an LSTM time sequence feature extraction model;

[0109] An abnormal water quality fingerprint of a monitored river section is identified;

[0110] If the identification is abnormal, an upstream water sample is collected and a water quality fingerprint is obtained by detection, and the water quality fingerprint is compared with the abnormal water quality fingerprint to determine a target river section according to the comparison result;

[0111] All enterprises in the target river section are subjected to real-time sampling, and the detection result data is compared with the detection result data of the abnormal water sample to determine a potential pollution source according to the comparison result;

[0112] The LSTM time sequence feature extraction model is used to determine a high-matching candidate source from the potential pollution source;

[0113] The high-matching candidate source is subjected to spatial displacement verification, and the pollution source is confirmed according to the verification result.

[0114] It should be noted that the present application builds a potential pollution source feature fingerprint library of a fusion LSTM (Long Short Term Memory network) time sequence feature extraction model, realizes all-round storage of enterprise pollution characteristics, not only contains static characteristics such as pollutant types, concentration ratios, stable isotope ratios, but also extracts dynamic time sequence vectors such as periodicity and trend characteristics of emissions through the LSTM algorithm, provides multi-dimensional benchmarks for tracing; through the joint determination of fluorescence peak position offset, new peak intensity ratio and abnormal duration, accurate identification of water quality anomalies is realized, and single index misjudgment is avoided; the intelligent strategy of "upstream sampling promotion + dichotomy sampling supplement" is used to locate the target river section, combined with the adjacent sampling point water quality fingerprint similarity change rate threshold, the pollution range is quickly narrowed, and the investigation efficiency is greatly improved; after screening potential pollution sources based on the triple comparison of pollutant types, concentration ratios and stable isotope ratios, high-matching candidate sources are further determined through LSTM time sequence feature vector comparison, the screening accuracy is improved from static characteristics and dynamic rules; finally, the river fine water dynamics-water quality coupling model is used to simulate the migration trajectory of pollutants from the candidate source to the monitoring section, and the simulated concentration time sequence data is compared and verified with the actual monitoring data, to establish the direct causal relationship between the pollution source and the pollution event. Form a complete traceability chain of "feature storage-exception identification-range locking-source screening-verification confirmation", solve the problems of time sequence feature missing, low matching accuracy and long tracing period of traditional methods, realize the accurate and efficient tracing of river pollutants, and it is especially suitable for industrial agglomeration area rivers, and can efficiently trace complex scenes such as intermittent illegal discharge and multi-source composite pollution.

[0115] According to the embodiment of the present application, the potential pollution source feature fingerprint library of the fusion LSTM time sequence feature extraction model is constructed, comprising:

[0116] Collecting the pollution samples of each enterprise in the monitoring river section at different pollution periods, obtaining the pollutant characteristic data, including pollutant types, pollutant concentration ratios, pollutant concentration time sequence data and stable isotope ratios;

[0117] Using a preset LSTM algorithm to train the pollutant concentration time sequence data in the sample, extracting time sequence feature vectors, including periodicity characteristic data and trend characteristic data, obtaining an LSTM time sequence feature extraction model;

[0118] Storing the pollutant characteristic data, time sequence feature vectors and LSTM time sequence feature extraction model into a database to generate a potential pollution source feature fingerprint library.

[0119] It should be noted that by collecting pollution samples of enterprises at different time periods, static characteristics such as types of pollutants, concentration ratios and stable isotope ratios are fused with time sequence characteristics (periodicity, trend) extracted by the LSTM to generate a potential pollution source characteristic fingerprint library, thereby providing a multi-dimensional benchmark for subsequent comparison. Compared with a traditional database that only stores static characteristics, the fingerprint library can capture the time sequence law of enterprise pollution, lay a data foundation for accurate matching of pollution sources, reduce misjudgments caused by ignoring time sequence characteristics, and solve the problem of distinguishing similar pollutants emitted by similar enterprises. The time sequence of pollutant concentration time sequence data is composed of pollutant concentration values collected at fixed time intervals.

[0120] The emission periodicity characteristic data includes peak frequency, peak interval, phase and peak base ratio. The peak frequency refers to the number of emission concentration peak values appearing per unit time, directly reflects the activity degree of periodic emission, and can distinguish continuous pollution and intermittent pollution. The peak interval refers to the time difference between adjacent two emission peaks (for example, a peak appears every 8 hours), which is a core quantitative index of periodicity and can accurately identify fixed interval laws such as daily cycle and weekly cycle. The phase refers to the relative time position of the peak value in the cycle (for example, the peak value appears at 12 o'clock every day), which can distinguish the “time imprint” of different pollution sources (for example, the peak phase of similar factories may differ by 2 hours due to different production shifts). The peak base ratio refers to the ratio of the peak concentration to the baseline concentration (the average concentration in the non-peak period) in the cycle (for example, the peak value is 5 times the baseline value), which can reflect the intensity fluctuation range of periodic emission and enhance the feature recognition degree.

[0121] The trend characteristic data includes peak slope, change rate, trend duration and mutation node. The peak slope refers to the rate of emission concentration rising from the baseline value to the peak value (for example, rising from 1 mg / L to 5 mg / L in 1 hour, the slope is 4 mg / (L·h)), which can reflect the “start-up characteristics” of the pollution process (for example, the slope difference between instantaneous emission and slow accumulation emission). The change rate refers to the rate of concentration change with time in the overall trend (for example, increasing by 0.2 mg / L per day), which quantifies the “steepness” of the trend and distinguishes between slow deterioration and sharp over-standard. The trend duration refers to the duration of the same change trend (increasing / decreasing) (for example, a decreasing trend for 10 consecutive days), which can exclude short-term interference (such as instantaneous leakage) and focus on long-term stable emission characteristics. The mutation node refers to the time point at which the concentration trend changes significantly (for example, the concentration suddenly jumps from 2 mg / L to 8 mg / L on May 5), which can be associated with abnormal events of the pollution source (such as equipment failure, illegal discharge), and enhance the timeliness of the traceability.

[0122] According to the embodiment of the present application, the abnormal water quality fingerprint identification of the monitoring river section comprises:

[0123] Real-time acquisition of the water quality fingerprint of the river section, and processing of the water quality fingerprint of the normal water quality to obtain a fluorescence peak position offset, a new peak intensity ratio and an abnormal duration;

[0124] If the fluorescence peak position offset is greater than a preset offset threshold, the new peak intensity ratio is greater than a preset intensity ratio threshold, and the abnormal duration is greater than a preset time threshold, it is determined that the water quality is abnormal.

[0125] It should be noted that the multi-parameter determination of the abnormal water quality of the fluorescence peak position offset, the new peak intensity ratio and the abnormal duration can more sensitively and accurately capture the water quality anomaly, reduce the misjudgment risk caused by background fluctuation or accidental interference, and realize rapid locking of the pollution event.

[0126] According to the embodiment of the present application, if the anomaly is identified, the upstream water sample is collected and the water quality fingerprint is detected and obtained, and the water quality fingerprint is compared with the abnormal water quality fingerprint, and the target river section is determined according to the comparison result, comprising:

[0127] If the anomaly is identified, the water quality fingerprint of the upstream section is collected and compared with the water quality fingerprint of the water quality abnormal section;

[0128] If the comparison is successful, the sampling is continued to the upstream;

[0129] If the comparison is not successful, the water sample is re-collected at the two-position between the upstream section and the water quality abnormal section, and compared with the water quality fingerprint of the water quality abnormal section;

[0130] According to the water quality fingerprint comparison result, the same cycle of pollution tracing and investigation is carried out until the similarity change rate of the water quality fingerprints of two adjacent sampling points exceeds a preset change rate threshold;

[0131] The river section between the two adjacent sampling points is taken as the target river section.

[0132] It should be noted that the target river section is located by the strategy of "upstream sampling + two-point sampling", and the similarity change rate threshold of the adjacent sampling points, which can quickly narrow the pollution source range, and compared with the traditional sampling of the whole river section without target, the method greatly reduces the sampling amount and analysis cost, and improves the accuracy and efficiency of the target river section positioning.

[0133] According to the embodiment of the present application, the detection result data is compared with the detection result data of the abnormal water sample, and the potential pollution source is determined according to the comparison result, comprising:

[0134] The detection result data includes the type of pollutants, the concentration ratio of pollutants and the stable isotope ratio;

[0135] The pollutant types, pollutant concentration ratios and stable isotope ratios of the enterprise water sample and the abnormal water sample are compared one by one;

[0136] The enterprise with the successful comparison of the pollutant types, the pollutant concentration ratios and the stable isotope ratios is selected as a potential pollution source.

[0137] It should be noted that, compared with single index comparison, the three comparisons of the pollutant types, the concentration ratios and the stable isotope ratios can significantly improve the specificity of the pollution source screening (for example, the stable isotope ratio has a "fingerprint" characteristic and is difficult to be tampered with by people) and reduce the misjudgment caused by similar pollutants emitted by different enterprises.

[0138] According to an embodiment of the present application, the determination of the high-matching candidate source from the potential pollution source by using the LSTM time sequence feature extraction model comprises:

[0139] The pollutant concentration time sequence data of the potential pollution source water sample and the abnormal water sample are acquired and input into the LSTM time sequence feature extraction model to obtain emission periodicity feature data and trend feature data;

[0140] The emission periodicity feature data and the trend feature data of the potential pollution source water sample and the abnormal water sample are compared, and the potential pollution source with the successful comparison is determined as the high-matching candidate source.

[0141] It should be noted that, the LSTM model is used to extract the emission periodicity, trend and other time sequence features of the potential pollution source and the abnormal water sample and to compare them, so that the high-matching candidate source is further screened from the "time law" aspect. Compared with the screening depending only on the static features, the supplement of the time sequence features can capture the behavior law of the enterprise pollution (such as the periodic emission caused by the production shift), exclude the interference sources that do not match the time sequence law of the abnormal event, and improve the reliability of the candidate source.

[0142] According to an embodiment of the present application, the spatial migration verification of the high-matching candidate source is performed, and the pollution source is confirmed according to the verification result, which comprises:

[0143] The concentration time sequence data of each section are collected downstream along the river course from the position of the high-matching candidate source as a starting point;

[0144] The pollutant concentration time sequence data, the pollutant types, the pollutant concentration ratios and the stable isotope ratios of the potential pollution source are input into a pre-constructed water dynamics-water quality coupling model to generate the concentration time sequence data of the pollutants at different spatial points, and the concentration time sequence data are compared with the actually monitored concentration time sequence data of each section;

[0145] If the comparison is passed, the high-matching candidate source is determined as the pollution source.

[0146] It should be noted that the concentration time series data (concentration curve changing with time) of the high matching candidate source is input into the river water dynamics-water quality coupling model to simulate the migration trajectory of the pollutant, the concentration curve changing with time (concentration time series data) of the pollutant at different spatial points (such as 1km, 3km and 5km downstream of the candidate source) is generated by using the model, and the concentration time series data of each section actually monitored is compared, and the spatial order of the concentration peak value (whether it decreases along the water flow direction) and the decay rate consistent with the model prediction are focused on. If the trends of the two are consistent, it indicates that the spatial migration path of the pollutant is consistent with the emission logic of the candidate source. Through the “theoretical simulation + actual data verification”, the misjudgment caused by the “similar characteristics but not same source” caused by the dependence on the characteristic comparison is avoided, and the spatial migration correlation between the finally locked pollution source and the pollution event is ensured, and the accuracy of the tracing result is improved. For example: the water quality fingerprint, static characteristics and dynamic time series vector of enterprise A and the polluted water body are matched, but the river water dynamics-water quality coupling model shows that the emissions of enterprise A need 5 hours to reach the pollution point, and the actual pollution occurs within 1 hour, and the actual pollution source is enterprise B which is not registered upstream.

[0147] The river fine water dynamics-water quality coupling model can simulate the water dynamics parameters and the migration trajectory of the pollutant simultaneously. When the model is constructed, first, based on the high-precision terrain data (including cross section, underwater DEM and roughness zoning) of the monitored river section, hydrological data (upstream flow, downstream water level), pollution source data (location of the discharge outlet, discharge amount and time series characteristics) and meteorological data, a two-dimensional or three-dimensional water dynamics model (such as the shallow water equation) is used to simulate the river flow field (flow velocity, water depth, turbulence intensity), and a water quality module is simultaneously coupled to simulate the convection, diffusion and degradation process of the pollutant, and the key parameters such as roughness and degradation coefficient are calibrated by using the measured water level, flow velocity and pollutant concentration data, so that the deviation between the simulation results of the model and the measured data is controlled within a preset threshold (such as water level error ≤10%), and finally the coupling model which can accurately reproduce the migration trajectory of the pollutant is formed.

[0148] According to the embodiment of the present application, further comprising:

[0149] The historical pollutant characteristic data of the pollution source is extracted from the potential pollution source characteristic fingerprint library, including the pollutant type, the pollutant concentration ratio, the stable isotope ratio and the time series characteristic vector;

[0150] The historical pollutant characteristic data and the real-time pollutant characteristic data of the pollution source currently collected are compared, and the pollutant type matching degree, the pollutant concentration ratio deviation rate, the stable isotope ratio deviation rate and the time series characteristic vector cosine similarity are obtained;

[0151] The pollutant concentration ratio deviation rate and the stable isotope ratio deviation rate are weighted and averaged to calculate whether the calculation result is less than or equal to a preset deviation rate threshold;

[0152] The pollution species matching degree and the time sequence feature vector cosine similarity are weighted and averaged to determine whether the calculation result is greater than or equal to a preset matching degree threshold;

[0153] If one of the above judgments is false, the potential pollution source is searched in multiple dimensions.

[0154] It should be noted that the historical pollutant characteristic data in the potential pollution source characteristic fingerprint library is called, the "historical emission characteristics" and the "current emission characteristics" are compared and analyzed, and if the difference is too large, it is necessary to further investigate whether the enterprise has process changes, illegal discharge and the like.

[0155] The stable isotope ratio deviation rate can be calculated by (real-time value - historical value) / historical value * 100%; the pollution species matching degree is represented by the ratio of the number of characteristic pollution species common to the real-time data and the historical data to the number of all characteristic pollution species in the historical data; and the pollution concentration proportion deviation rate is calculated for multiple characteristic pollutants, respectively, (real-time proportion - historical proportion) / historical proportion * 100%, and the calculation results of all characteristic pollutants are added to calculate the average value.

[0156] The river pollutant tracing method and system provided by the application can effectively exclude interference sources and accurately lock the real pollution source even in the presence of multiple suspected enterprises and intermittent pollution, and solve the problems of low tracing efficiency and insufficient matching accuracy of traditional methods in complex pollution situations.

[0157] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can be implemented in other manners. The above described device embodiments are merely schematic, for example, the division of the units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling, or direct coupling or communication connection between various components can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0158] The units described as separate parts above can or can not be physically separate, the parts displayed as units can or can not be physical units; they can be located in one place, or distributed on multiple network units; part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0159] In addition, each functional unit in each embodiment of the application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be realized in the form of hardware or in the form of hardware plus software functional unit.

Claims

1. A method for tracing the source of river pollutants, characterized in that, The method comprises the following steps: Collecting pollutant samples of each enterprise in the monitoring river section at different pollution periods to obtain pollutant characteristic data, including pollutant types, pollutant concentration proportions, pollutant concentration time series data and stable isotope ratios; Training the pollutant concentration time series data in the samples by using a preset LSTM algorithm to extract time series feature vectors, including periodicity characteristic data and trend characteristic data, and obtaining an LSTM time series feature extraction model; Storing the pollutant characteristic data, time series feature vectors and LSTM time series feature extraction model in a database to generate a potential pollution source characteristic fingerprint library; Real-time collecting water quality fingerprints of the monitoring river section, and obtaining a fluorescence peak position offset, an added peak intensity ratio and an abnormal duration by combining the water quality fingerprints of normal water quality; If the fluorescence peak position offset is greater than a preset offset threshold, the added peak intensity ratio is greater than a preset intensity ratio threshold, and the abnormal duration is greater than a preset time threshold, it is determined that the water quality is abnormal; If the abnormality is identified, the water quality fingerprint of the upstream section is collected and compared with the water quality fingerprint of the section where the water quality is determined to be abnormal; If the comparison is successful, the sampling continues to advance upstream; If the comparison is unsuccessful, water samples are collected again at a two-division position between the upstream section and the section where the water quality is determined to be abnormal, and compared with the water quality fingerprint of the section where the water quality is determined to be abnormal; The same cycle of pollution tracing is performed according to the water quality fingerprint comparison result until the similarity change rate of the water quality fingerprints of two adjacent sampling points exceeds a preset change rate threshold; The river section between the two adjacent sampling points is taken as a target river section; Real-time sampling is performed on all enterprises in the target river section, the detection result data is compared with the detection result data of the abnormal water sample, and a potential pollution source is determined according to the comparison result; The LSTM time series feature extraction model is used to determine a high-matching candidate source from the potential pollution source; Spatial displacement verification is performed on the high-matching candidate source, and the pollution source is confirmed according to the verification result.

2. The river pollution backtracking method according to claim 1, characterized in that, The real-time sampling of all enterprises in the target river section, the comparison of the detection result data with the detection result data of the abnormal water sample, and the determination of the potential pollution source according to the comparison result, comprise: The detection result data includes pollutant types, pollutant concentration proportions and stable isotope ratios; The pollutant types, pollutant concentration proportions and stable isotope ratios of the water samples of the enterprises and the abnormal water sample are compared one by one; The enterprises whose pollutant types, pollutant concentration proportions and stable isotope ratios are all successfully compared are selected as potential pollution sources.

3. The river pollution backtracking method according to claim 2, wherein, The use of the LSTM time series feature extraction model to determine a high-matching candidate source from the potential pollution source comprises: Obtaining pollutant concentration time series data of the potential pollution source water sample and the abnormal water sample, and inputting the data into the LSTM time series feature extraction model to process and obtain periodicity characteristic data and trend characteristic data; The periodicity characteristic data and trend characteristic data of the potential pollution source water sample and the abnormal water sample are compared, and the potential pollution source whose comparison is successful is taken as a high-matching candidate source.

4. The river pollution backtracking method of claim 3, wherein, The spatial displacement verification of the high-matching candidate source and the confirmation of the pollution source according to the verification result comprise: Starting from the position of the high-matching candidate source, collecting concentration time series data of each section along the downstream of the river course; The concentration time series data of pollutants of potential pollution sources, the pollutant types, the pollutant concentration ratios and the stable isotope ratios are input into a pre-constructed hydrodynamic-water quality coupling model to generate the concentration time series data of pollutants at different spatial points, and the concentration time series data is compared with the actually monitored concentration time series data of each section; If the comparison is passed, the high-matching candidate source is determined as the pollution source.

5. A river pollutant tracing system employing a river pollutant tracing method according to any one of claims 1 to 4, characterized in that, The river pollutant source tracing method comprises the following steps: Collecting the pollutant samples of each enterprise in the monitoring river section at different pollution periods to obtain the pollutant characteristic data, including the pollutant types, the pollutant concentration ratios, the pollutant concentration time series data and the stable isotope ratios; Training the pollutant concentration time series data in the samples by using a preset LSTM algorithm to extract the time series feature vectors, including the periodic characteristic data and the trend characteristic data, and obtaining an LSTM time series feature extraction model; Storing the pollutant characteristic data, the time series feature vectors and the LSTM time series feature extraction model in a database to generate a potential pollution source characteristic fingerprint library; Real-time collecting the water quality fingerprints of the monitoring river section, and obtaining the fluorescence peak position offset, the new peak intensity ratio and the abnormal duration by combining the water quality fingerprints of the normal water quality; If the fluorescence peak position offset is greater than a preset offset threshold, the new peak intensity ratio is greater than a preset intensity ratio threshold, and the abnormal duration is greater than a preset time threshold, the water quality is determined to be abnormal; If the abnormality is identified, the section where the water quality is determined to be abnormal is taken as a starting point, the water quality fingerprints of the upstream sections are collected and compared with the water quality fingerprints of the section where the water quality is determined to be abnormal; If the comparison is successful, the sampling is continued to the upstream; If the comparison is unsuccessful, the water sample is collected again at the two-position between the upstream section and the section where the water quality is determined to be abnormal, and compared with the water quality fingerprints of the section where the water quality is determined to be abnormal; The same cycle of pollution source tracing is performed according to the water quality fingerprint comparison result until the similarity change rate of the water quality fingerprints of two adjacent sampling points exceeds a preset change rate threshold; The river section between the two adjacent sampling points is taken as a target river section; The real-time sampling is performed on all the enterprises in the target river section, the detection result data is compared with the detection result data of the abnormal water sample, and the potential pollution source is determined according to the comparison result; The high-matching candidate source is determined from the potential pollution source by using the LSTM time series feature extraction model; The spatial migration verification is performed on the high-matching candidate source, and the pollution source is confirmed according to the verification result.

Images