Groundwater level prediction method and system based on big data

Abstract

The application discloses a groundwater level prediction method and system based on big data, relates to the technical field of groundwater pollution prevention and control and hydrological environment monitoring, and comprises the following steps: step S1, determining a pollution source position, setting a time stamp of a pollution migration state; step S2, obtaining a space-time correlation feature; and step S3, constructing a correlation relationship and obtaining a groundwater level of a target monitoring station. Through fusion analysis of data, the time stamp of the pollution migration state is established, accurate synchronous benchmarks are provided for subsequent groundwater level prediction, multi-dimensional data can be integrated into a structured space-time correlation feature by extracting a correlation matrix, the pollution diffusion data of the ground that is easy to monitor can indirectly serve the groundwater level prediction by direct mapping, the problems of insufficient data or difficult acquisition are solved, and the prediction cost is reduced.

Description

Technical Field

[0001] This invention relates to the technical field of groundwater pollution prevention and control and hydrological environment monitoring, and in particular to a groundwater level prediction method and system based on big data. Background Technology

[0002] In recent years, groundwater level prediction has been widely applied in landslide early warning, water resource management in agricultural irrigation areas, and prevention and control of urban groundwater subsidence. Deep learning models are deployed in the cloud and combined with lightweight statistical preprocessing and anomaly detection mechanisms to achieve high-precision real-time prediction of incomplete hydrological data. Multi-source big data such as satellite remote sensing, meteorological reanalysis data, and IoT sensors are integrated to construct a more comprehensive input feature set. An improved TANK hydrological model for landslides, combined with big data monitoring, enables rapid prediction of groundwater level fluctuations caused by rainfall infiltration. Hybrid interpolation schemes such as Z-score anomaly detection, sliding window segmentation, and STL-SARIMA interpolation are used to improve the continuity of input data. Principal component analysis, mutual information theory, and correlation analysis are used to screen key input variables.

[0003] Currently, Chinese invention patent CN120781025A discloses an efficient method for predicting the spatiotemporal distribution of groundwater levels at a grid scale in large-scale agricultural irrigation areas. This method constructs a SWAT model of the watershed where the irrigation area is located and a MODFLOW model covering the irrigation area, coupling the two to obtain a SWAT-MODFLOW model. Parameter corrections are then performed. A groundwater level dataset for the irrigation area is constructed through scenario combination and model simulation, used to train and select the most accurate LSTM model. The optimal control range of each LSTM model is determined through spatial compression and reconstruction, resulting in multiple LSTM models. Finally, these multiple LSTM models are used in parallel to predict the spatiotemporal distribution of groundwater levels at the grid scale of the irrigation area. This invention determines the optimal control range of each LSTM model through spatial compression and reconstruction, and ultimately achieves the prediction of the spatiotemporal distribution of groundwater levels at a grid scale in large-scale agricultural irrigation areas through parallel computation of multiple LSTM models. It improves prediction efficiency while maintaining the accuracy of groundwater level prediction. However, related technologies have limitations due to the single dimension of groundwater level prediction data and insufficient mining of spatiotemporal features, making it difficult to predict groundwater levels using readily available data. Summary of the Invention

[0004] The technical problem solved by this invention is: relying on pollution migration-related data, analyzing techniques to uncover the dynamic change patterns of water levels, obtaining the correlation between the dynamic change patterns of water levels and pollution migration, and realizing the prediction of groundwater levels based on the monitored pollution migration-related data.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: Firstly, a groundwater level prediction method based on big data, comprising the following steps: Step S1, determining the location of the pollution source, and setting a timestamp of the pollution migration status by combining meteorological data and historical pollution diffusion data; Step S2: Obtain the historical groundwater level corresponding to each timestamp, and obtain the spatiotemporal correlation features based on the historical groundwater level; Step S3: Based on the spatiotemporal correlation characteristics, construct the correlation between pollution migration status and groundwater level, and obtain the groundwater level of the target monitoring station based on the correlation and the current pollution migration status of each monitoring station.

[0006] As a preferred embodiment of the groundwater level prediction method based on big data described in this invention, step S1 includes the following sub-steps: step S11, setting the location of the pollution source according to the sewage discharge information of the area to be analyzed. Step S12: Obtain meteorological data and historical pollution diffusion data for the area to be analyzed; Step S13: Based on the distance between historical pollution data and pollution sources, set the affiliated source; Step S14: Set the pollution migration status based on the historical pollution data of the affiliated source; Step S15: Set pollution-related timestamps based on the pollution migration status.

[0007] As a preferred embodiment of the groundwater level prediction method based on big data described in this invention, step S13 includes the following sub-steps: step S131, transforming the location of the monitoring station and the location of the pollution source into the world coordinate system; Step S132: Calculate the straight-line distance between any monitoring station and each pollution source according to the Euclidean distance formula, and set the affiliated source based on the straight-line distance and the terrain features within the straight-line distance range; Step S133: Traverse each monitoring station to obtain the corresponding source of each station.

[0008] As a preferred embodiment of the groundwater level prediction method based on big data described in this invention, the pollution migration status includes not migrated, migrated but not transferred, and transferred. Pollution-related timestamps include the initial pollution timestamp and the initial transfer timestamp; Step S14 includes the following sub-steps: Step S141, based on the historical pollution diffusion data of the monitoring stations, mark the peak points, extract the timestamps of the peak points, obtain the timestamps of the peak points of each monitoring station under the source, and set the pollution migration order according to the timestamps of each peak point. Step S142: Obtain historical emission data of the affiliated sources, and set the pollution affiliation ratio based on historical emission data, meteorological data, and peak point timestamps; Step S143 includes the following sub-steps: resetting the pollution migration order according to the pollution affiliation ratio, and setting the pollution migration status based on the reset pollution migration order; Step S15 includes the following sub-steps: Step S151, for the unmigrated state, jump to the next set of historical contamination data and re-analyze the contamination migration state until the contamination migration state is converted to the migrated state or all historical contamination data has been jumped to, and mark the initial contamination timestamp. Step S152: For the migrated but not transferred state and the transferred state, mark the initial contamination timestamp.

[0009] As a preferred embodiment of the groundwater level prediction method based on big data described in this invention, step S2 includes the following sub-steps: step S21, according to the time sequence of the initial pollution timestamp of each monitoring station, the historical groundwater level corresponding to each monitoring station at the initial pollution timestamp is obtained, and N sets of water level sequences are obtained. Step S22: Based on the water level sequence and the coordinates of each monitoring station, obtain the spatiotemporal correlation characteristics.

[0010] As a preferred embodiment of the groundwater level prediction method based on big data described in this invention, any set of water level sequences is selected. Set the values ​​for "not migrated", "migrated but not transferred", and "transferred" to 0, 1, and 2 respectively; Obtain the pollution transfer status of the monitoring stations corresponding to the water level sequence, and obtain the monitoring station number corresponding to the water level sequence; Construct a number vector based on the monitoring station number, a status vector based on the pollution transfer status, and a water level vector based on the water level sequence. The number vector, state vector and water level vector are concatenated from top to bottom to obtain the correlation matrix. The initial pollution timestamps of each group are traversed to obtain the correlation matrix corresponding to each group of initial pollution timestamps. Set the initial contamination timestamp as the label of the association matrix.

[0011] In a preferred embodiment of the groundwater level prediction method based on big data described in this invention, the correlation matrix is ​​set as a spatiotemporal correlation feature.

[0012] As a preferred embodiment of the groundwater level prediction method based on big data described in this invention, the correlation relationship is represented as the correspondence between spatiotemporal correlation features and the corresponding initial pollution timestamp. Obtain the current pollution migration status of each monitoring station, and based on the number corresponding to the monitoring station and the number corresponding to the pollution migration status, obtain the first two rows of elements in the spatial correlation matrix; Input the first two rows of elements in the spatial correlation matrix into the correlation relationship to obtain candidate spatiotemporal correlation features, obtain the water level vector in the candidate spatiotemporal correlation features, and set the water level vector as the first candidate groundwater level value. Obtain the historical pollution migration status from the first historical moment after the current moment to the current moment, and obtain the second groundwater level candidate value according to the method for obtaining groundwater level candidate values; Obtain the initial pollution timestamp corresponding to each candidate value, obtain the first groundwater level candidate value and the second groundwater level candidate value adjacent to the initial pollution timestamp, and determine the groundwater level of the target monitoring station based on the first groundwater level candidate value and the second groundwater level candidate value adjacent to the initial pollution timestamp.

[0013] Secondly, a groundwater level prediction system based on big data includes a setting module, an analysis module, and a mapping module; The setting module determines the location of the pollution source and sets a timestamp for the pollution migration status by combining meteorological data and historical pollution diffusion data. The analysis module obtains the historical groundwater level corresponding to each timestamp, and obtains spatiotemporal correlation features based on the historical groundwater level; The mapping module constructs a correlation between pollution migration status and groundwater level based on spatiotemporal correlation characteristics, and obtains the groundwater level of the target monitoring station based on the correlation and the current pollution migration status of each monitoring station.

[0014] The beneficial effects of this invention are as follows: By fusing and analyzing pollution source location, meteorological data, and historical pollution diffusion data, a timestamp of pollution migration status is established, which enables spatiotemporal alignment of the pollutant diffusion process with the dynamic changes in groundwater level, providing an accurate synchronization benchmark for subsequent water level prediction. By extracting the historical groundwater level corresponding to each timestamp and constructing a correlation matrix between water level sequence and pollution migration status, the intrinsic relationship between pollutant migration path and groundwater level response can be fully explored, integrating multidimensional data into structured spatiotemporal correlation features. By directly mapping pollution migration status with groundwater level, easily monitored surface pollution diffusion data can indirectly serve groundwater level prediction, solving the problem of insufficient or difficult-to-obtain groundwater level monitoring data in traditional methods and reducing prediction costs. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the basic process of a groundwater level prediction method based on big data, provided as an embodiment of the present invention. Detailed Implementation

[0016] 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 some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0017] It should be understood that the step numbers used herein are for ease of description only and are not intended to limit the order in which the steps are performed. It should also be understood that the terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention.

[0018] As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0019] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0020] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0021] Example, refer to Figure 1 As an embodiment of the present invention, a groundwater level prediction method based on big data is provided, including the following steps: Step S1, determine the location of the pollution source, and set a timestamp of the pollution migration status by combining meteorological data and historical pollution diffusion data; Step S2: Obtain the historical groundwater level corresponding to each timestamp, and obtain the spatiotemporal correlation features based on the historical groundwater level; Step S3: Based on the spatiotemporal correlation characteristics, construct the correlation between pollution migration status and groundwater level, and obtain the groundwater level of the target monitoring station based on the correlation and the current pollution migration status of each monitoring station.

[0022] More preferably, this invention establishes a timestamp of pollution migration status by fusing and analyzing pollution source location, meteorological data, and historical pollution diffusion data. This allows for spatiotemporal alignment of the pollutant diffusion process with dynamic changes in groundwater levels, providing an accurate synchronization benchmark for subsequent water level prediction. By extracting historical groundwater levels corresponding to each timestamp and constructing a correlation matrix between water level sequences and pollution migration status, the inherent connection between pollutant migration paths and groundwater level responses can be fully explored. Multidimensional data is integrated into structured spatiotemporal correlation features. By directly mapping pollution migration status to groundwater levels, easily monitored surface pollution diffusion data can indirectly serve groundwater level prediction, solving the problem of insufficient or difficult-to-obtain groundwater level monitoring data in traditional methods and reducing prediction costs.

[0023] Meteorological data includes wind speed, wind direction, precipitation, and evaporation. More preferably, the wind speed, wind direction, precipitation, and evaporation are all selected from the wind speed, wind direction, precipitation, and evaporation corresponding to the historical monitoring time points, so as to facilitate the linkage to determine the pollution migration status.

[0024] More preferably, wind speed, wind direction, precipitation, and evaporation are all expressed as the meteorological data from the previous historical monitoring time point to the next historical monitoring time point, and the average value of each meteorological data in between. For wind direction, the average value is represented as the average direction, which is a vector. For wind speed, precipitation, and evaporation, the average value is represented as the average value, which is a scalar.

[0025] Historical pollution diffusion data includes historical monitoring locations, corresponding historical monitoring time points, and corresponding historical pollutant concentrations.

[0026] More preferably, the historical monitoring location is represented by the longitude and latitude corresponding to the historical monitoring station; Historical monitoring time points refer to the time points at which historical monitoring stations monitored pollutant concentrations at historical monitoring locations.

[0027] More preferably, the historical pollutant concentrations include BOD5 concentration, COD concentration, SS concentration, ammonia nitrogen concentration, animal and vegetable oil concentration, petroleum hydrocarbon concentration, benzene concentration, toluene concentration, xylene concentration, total phosphorus concentration, total nitrogen concentration, and heavy metal concentration; The methods used by historical monitoring stations to monitor historical pollutant concentrations include: For BOD5 concentration (five-day biochemical oxygen demand concentration): Take the water sample to be tested, divide the water sample into two parts, immediately measure the initial dissolved oxygen (DO) concentration of one part, and incubate the other part in the dark at 20℃ for 5 days. Measure the remaining dissolved oxygen after incubation. BOD5 is expressed as the difference between the initial dissolved oxygen and the dissolved oxygen after 5 days. If the water sample contains toxins or has insufficient DO, dilute it with dilution water before measuring. BOD5 concentration reflects the total oxygen consumption of organic matter in the water that can be decomposed by microorganisms.

[0028] For COD concentration (chemical oxygen demand concentration): Add excess potassium dichromate oxidant, concentrated sulfuric acid and silver sulfate catalyst, heat and reflux for 2 hours to oxidize the reducing substances (mainly organic matter) in the water sample, titrate the remaining potassium dichromate with ferrous ammonium sulfate, calculate the oxygen equivalent based on the consumption, and set the oxygen equivalent as the COD concentration.

[0029] For SS concentration (suspended solids concentration): a known volume of water sample is filtered using a glass fiber filter membrane (pore size 0.45μm) of constant weight. The retentate along with the filter membrane is dried at 103~105℃ to constant weight. SS concentration is expressed as the ratio of the difference between the mass of the filter membrane after drying and the mass of the filter membrane before filtration to the volume of the filtered water sample.

[0030] For ammonia nitrogen concentration: After the water sample is pretreated by flocculation and sedimentation, Nessler's reagent (potassium mercuric iodide) is added, which reacts with ammonia to form a yellow-brown complex. The absorbance is measured at a wavelength of 420 nm. If the water sample contains interfering substances (such as residual chlorine and metal ions), distillation pretreatment is required.

[0031] For total phosphorus concentration: potassium persulfate (or sodium persulfate) is added and digested under high pressure to oxidize all organic phosphorus and polyphosphates to orthophosphate (PO4³⁻). Ammonium molybdate and ascorbic acid are added to generate a blue phosphomolybdic heteropolyacid complex. The absorbance is measured at a wavelength of 700 nm and compared with a standard series for quantification.

[0032] For total nitrogen concentration: add alkaline potassium persulfate and autoclave at 120~124℃ for 30 minutes to oxidize all organic nitrogen, ammonia nitrogen, and nitrite nitrogen to nitrate. Measure the nitrate absorbance using ultraviolet spectrophotometry (dual wavelengths of 220nm and 275nm), and calculate the total nitrogen concentration after subtracting interference.

[0033] For the concentration of animal and vegetable oils: acidify the water sample with hydrochloric acid to a pH less than or equal to 2, extract the oily substances in the sample with carbon tetrachloride (or n-hexane) as the extractant, pass the extract through a magnesium silicate adsorption column, where non-polar petroleum hydrocarbons are adsorbed and polar animal and vegetable oils pass through, collect the eluent, evaporate the solvent to dryness and weigh; or measure the characteristic absorption (2930 cm⁻¹, 2960 cm⁻¹, 3030 cm⁻¹) using infrared spectrophotometry, and express the concentration of animal and vegetable oils as the difference between the total oil concentration and the petroleum hydrocarbon concentration.

[0034] For petroleum hydrocarbon concentration: After extraction, the extract is passed through a magnesium silicate adsorption column, and the petroleum hydrocarbons on the adsorption column are eluted with a small amount of eluent. The eluent is collected and determined by infrared spectrophotometry (based on the stretching vibration absorption of CH bonds). The petroleum hydrocarbon concentration reflects the total amount of petroleum hydrocarbon pollutants.

[0035] For benzene, toluene, and xylene concentrations: volatile organic compounds in the water sample are purged with an inert gas (nitrogen / helium), adsorbed in a collection tube, desorbed by heating, and separated by a gas chromatograph. The three substances are then detected by a flame ionization detector (FID). The three substances are identified qualitatively based on their retention time and quantified by their peak area or peak height.

[0036] For heavy metal concentrations: add nitric acid (perchloric acid or hydrofluoric acid if necessary) for digestion, so that all suspended and organically complexed heavy metals are converted into ionic states. Then perform flame atomization (copper, zinc, lead, cadmium, etc.) or graphite furnace atomization (trace elements such as arsenic, mercury, selenium, etc.) and measure the absorbance at characteristic wavelengths of the elements. Common heavy metals include: mercury (Hg), cadmium (Cd), lead (Pb), chromium (Cr), arsenic (As), copper (Cu), zinc (Zn), nickel (Ni), and manganese (Mn).

[0037] The detection methods for historical pollutants are all existing technologies and will not be described in detail below.

[0038] Step S1 includes the following sub-steps: Step S11, set the location of the pollution source based on the sewage discharge information of the area to be analyzed; Step S12: Obtain meteorological data and historical pollution diffusion data for the area to be analyzed; Step S13: Based on the distance between historical pollution data and pollution sources, set the affiliated source; Step S14: Set the pollution migration status based on the historical pollution data of the affiliated source; Step S15: Set pollution-related timestamps based on the pollution migration status.

[0039] Step S13 includes the following sub-steps: Step S131, convert the location of the monitoring station and the location of the pollution source to the world coordinate system; More preferably, the location of the monitoring station and the location of the pollution source are both represented as the longitude, latitude and altitude of the corresponding geometric center point.

[0040] More preferably, the WGS-84 ellipsoid model is retrieved, the semi-major axis is set to 6,371,137 meters, the flattening is set to 1 / 298.257223563, and the eccentricity is set to... Where f is the flattening, the expression for the WGS-84 ellipsoid model is: ; in, , Let e ​​be the semi-major axis and e be the eccentricity. Latitude Longitude Altitude Let be the radius of curvature of the ellipsoid.

[0041] Step S132: Calculate the straight-line distance between any monitoring station and each pollution source according to the Euclidean distance formula, and set the affiliated source based on the straight-line distance and the terrain features within the straight-line distance range; More preferably, the expression for calculating the straight-line distance is: ; Where d is the straight-line distance. , and These represent the X-axis, Y-axis, and Z-axis coordinates of the monitoring station. , and These are the X-axis, Y-axis, and Z-axis coordinates of the pollution source, respectively.

[0042] More preferably, the method for setting the straight-line distance range includes: taking the geometric center point of the pollution source as the center and drawing a circle with an initial radius, and setting the circular area as the initial straight-line distance range. The initial radius is determined according to the type of pollution source. For point source pollution, it is set to 500 meters; for line source pollution, it is set to the radius of the envelope circle after expanding 200 meters outward from both ends of the pollution source in the length direction; and for area source pollution, it is set to the radius of the geometric circumcircle of the area source by 300 meters outward.

[0043] More preferably, the method for setting the membership source includes an exclusion method and a setting method; The exclusion method includes elimination. If there are mountains, hills or artificial dikes within the straight-line distance between the monitoring station and the pollution source, and the altitude of the mountains, hills or artificial dikes is higher than the average height of the topographic profile formed by the line connecting the pollution source and the monitoring station, then it is determined that there is a topographic barrier between the monitoring station and the pollution source, and the monitoring station is excluded. If there is an underground aquifer fault zone, aquitard, or karst development zone between the monitoring station and the pollution source, and the fault zone or aquitard blocks the migration path of pollutants in the groundwater, then it is determined that the monitoring station and the pollution source are not hydrologically connected, and the monitoring station is excluded. If a monitoring station and a pollution source belong to different surface water basins or groundwater basins, and the watershed is located within a straight-line distance between the two, then the monitoring station is determined to belong to a different watershed than the pollution source, and the monitoring station is excluded.

[0044] The setup method includes: after excluding topographical barriers, hydrological disconnections, and differences in watershed affiliation, sorting the remaining pollution sources in ascending order according to their straight-line distance, selecting the pollution source with the smallest straight-line distance, and setting this pollution source as w as the affiliated source of the monitoring station.

[0045] More preferably, digital elevation data within the straight-line distance range between the monitoring station and the pollution source is obtained. Topographic elevation points are extracted at sampling intervals along the line connecting the monitoring station and the pollution source to generate a topographic profile curve. A straight line segment is formed by connecting the elevation points of the monitoring station and the elevation points of the pollution source. The area enclosed by the straight line segment and the topographic profile curve is calculated. The area is divided by the horizontal projection length of the straight line segment to obtain the average height of the topographic profile. If there are mountains, hills, or artificial embankments within the straight-line distance range, the elevation of the geometric center point of the mountain, hill, or artificial embankment is extracted. If the elevation of the center point is higher than the average height of the topographic profile, it is determined that there is a topographic barrier between the monitoring station and the pollution source, and the monitoring station is excluded.

[0046] More preferably, the sampling interval is dynamically set according to the straight-line distance. When the straight-line distance is less than or equal to 1000 meters, the sampling interval is set to 10 meters. When the straight-line distance is greater than 1000 meters but less than or equal to 5000 meters, the sampling interval is set to 50 meters. When the straight-line distance is greater than 5000 meters, the sampling interval is set to 100 meters.

[0047] More preferably, the hydrogeological survey report of the area to be analyzed is retrieved, and the distribution maps of aquifers, fault zones, and karst development areas within the range connecting the monitoring station and the pollution source are extracted. If a fault zone exists within the range, its nature is determined to be either a water-conducting fault zone or a water-impermeable fault zone. If it is a water-impermeable fault zone, it is determined that the fault zone blocks the migration path of pollutants in groundwater. If a karst development area exists within the range, its development level is determined to be strong, moderate, or weak. If it is strong and an underground river system exists in the karst development area... If the flow direction of the underground river system is not directed from the pollution source to the monitoring station, then it is determined that the karst development area blocks the migration path of pollutants in the groundwater. If there is an impermeable layer within the line, and the lithology of the impermeable layer is determined to be clay, shale, or dense limestone, and the continuous thickness of the impermeable layer is greater than the thickness threshold, then it is determined that the impermeable layer blocks the migration path of pollutants in the groundwater. If any of the above-mentioned blocking situations are determined to exist, then it is determined that the monitoring station and the pollution source are not hydrologically connected, and the monitoring station is excluded.

[0048] More preferably, the thickness threshold is set according to the type of pollutant: 5 meters for heavy metal pollutants, 3 meters for organic pollutants, and 2 meters for nutrient pollutants.

[0049] More preferably, surface water system vector data and groundwater flow field data of the area to be analyzed are obtained. For surface water basins, the nearest surface river is extracted along the line connecting the monitoring station and the pollution source. It is determined whether the monitoring station and the pollution source are located on the same side of the river. If they are located on different sides and the river watershed is within the straight-line distance between them, it is determined that the monitoring station and the pollution source belong to different surface water basins. For groundwater basins, the groundwater flow direction of the monitoring station and the pollution source is extracted according to the groundwater flow field data. If the groundwater flow directions of the two point to different discharge areas and the groundwater watershed between the two discharge areas is within the straight-line distance between them, it is determined that the monitoring station and the pollution source belong to different groundwater basins. When it is determined that they belong to different surface water basins or groundwater basins, it is determined that the monitoring station and the pollution source belong to different basins, and the monitoring station is excluded.

[0050] More preferably, the method for determining the groundwater watershed includes setting up a virtual profile within the range of the line connecting the monitoring station and the pollution source, extracting the water head value of each point along the profile direction, connecting adjacent points with equal water head values ​​to form an iso-water head line, and connecting the apexes of the iso-water head line that bulge towards the direction of higher water head as the groundwater watershed.

[0051] Step S133: Traverse each monitoring station to obtain the corresponding source of each station.

[0052] More preferably, the monitoring stations are tagged based on their source affiliation, and the tag for each monitoring station is Q. k , where k is a natural number, representing the k-th pollution source.

[0053] Pollution migration status includes not migrated, migrated but not transferred, and transferred; Pollution-related timestamps include the initial pollution timestamp and the initial transfer timestamp; Step S14 includes the following sub-steps: Step S141, based on the historical pollution diffusion data of the monitoring stations, mark the peak points, extract the timestamps of the peak points, obtain the timestamps of the peak points of each monitoring station under the source, and set the pollution migration order according to the timestamps of each peak point. More preferably, the first duration is set as a sliding time window. Historical pollution diffusion data from any monitoring station is slid across the data in reverse chronological order to obtain data segments. These data segments are then numbered according to the sliding order, and the data segment number is denoted as P. ij Where i is the i-th monitoring station, j is the data segment obtained by the j-th sliding, and i and j are both natural numbers; Obtain any data segment, extract the time series of historical pollutant concentrations of any type in the data segment, select the maximum value in the time series and record it as the peak point, and mark the peak point as the first marker; Iterate through each data segment to obtain the peak point of each data segment.

[0054] More preferably, the timestamps of each peak point of the monitoring station are extracted, and the timestamps of each peak point of each monitoring station are extracted by traversing each monitoring station.

[0055] More preferably, obtain any group of monitoring stations with the same label, obtain each peak point with the same j value in the group of stations, obtain the timestamp of the peak points in the group, sort the timestamps of the peak points in the group in instantaneous order, extract the number of each monitoring station after sorting, i.e., the i value, and set the number string composed of each i value as the pollution migration order.

[0056] To explain the order of pollution migration, for example, when the number string is 3, 1, 4, 2, it indicates that the pollution first peaks at monitoring station number 3, and then spreads to stations 1, 4, and 2 in sequence. This sequence is not a simple mapping of spatial proximity, but a temporal topology influenced by the groundwater flow field, the heterogeneity of the medium, and the pollution source.

[0057] More preferably, the first duration is 24 hours to ensure coverage of typical pollution diffusion cycles. The first duration can be dynamically adjusted according to regional hydrogeological conditions. For example, it can be shortened to 12 hours in highly permeable aquifers and extended to 48 hours in poorly permeable aquifers.

[0058] Step S142: Obtain historical emission data of the affiliated sources, and set the pollution affiliation ratio based on historical emission data, meteorological data, and peak point timestamps; More preferably, the historical emission data includes historical emission time points, historical emission amounts, and historical emission types.

[0059] More preferably, the historical emission types of the source are obtained, and the pollution affiliation ratio of historical pollutant concentrations not included in the historical emission types is set to 0.

[0060] More preferably, the historical pollutant concentrations included in any historical emission type are selected, and the method for setting the pollution affiliation ratio of the historical pollutant concentrations includes selecting the pollution migration sequence corresponding to any affiliation source and obtaining the monitoring station ranked first. Based on the coordinates of the source and the coordinates of the monitoring station, construct a vector pointing from the source to the monitoring station, denoted as the reference vector; Calculate the time difference between the peak point timestamp and each historical emission time point, and select the historical emission time point corresponding to the absolute value of the smallest time difference as the key time point; Count the number of historical monitoring time points between the key time point and the peak point, and use the difference between this number and 1 to set the number of wind directions; Based on the calculation method of vector sum, the comprehensive wind direction is obtained, and based on the comprehensive wind direction, the wind direction vector is obtained; Obtain the angle between the reference vector and the risk vector, obtain the cosine value of the angle, and set the cosine value as the pollution membership percentage; More preferably, when the cosine value is negative, the pollution membership percentage is set to 0.

[0061] More preferably, the method for setting the comprehensive wind direction includes obtaining the wind direction at each historical monitoring time point between the key time point and the peak point timestamp, summing each wind direction as a vector, and setting the direction of the vector sum as the comprehensive wind direction.

[0062] More preferably, the historical pollutant concentrations at each monitoring station are iterated to obtain the pollution affiliation ratio of each historical pollutant concentration to its respective source. By traversing all monitoring stations, the pollution affiliation ratio of each historical pollutant concentration at each monitoring station to its respective source is obtained.

[0063] Step S143 includes the following sub-steps: resetting the pollution migration order according to the pollution affiliation ratio, and setting the pollution migration status based on the reset pollution migration order; More preferably, the original pollution migration sequence is obtained, and the peak point of any historical pollution concentration at each monitoring station in the original pollution migration sequence is obtained; The expression for diffusion concentration is: ; Where U represents historical emissions. It is a constant. , Here, y represents the lateral diffusion coefficient and vertical diffusion coefficient, respectively; y is the product of the straight-line distance between the monitoring station and its affiliated source and the proportion of pollution belonging to that source; z is the product of the straight-line distance between the monitoring station and its affiliated source and the sine of the included angle; and H is the height of the emission outlet of the affiliated source, i.e., the vertical distance from the geometric center point of the emission outlet to the ground. For overall wind speed, The diffusion concentration at this monitoring station, To calculate the total precipitation, To account for the total evaporation, For distance gradient, is a constant, and d is the straight-line distance between the monitoring station and the source it belongs to.

[0064] More preferably, the setting logic for comprehensive wind speed, comprehensive precipitation, and comprehensive evaporation is the same, and they are collectively referred to as comprehensive meteorological data. The method for setting comprehensive meteorological data includes counting the number of historical monitoring time points between the timestamps of key time points and peak points, and setting the difference between this number and 1 as the number of meteorological data. Based on the calculation method of scalar sum, comprehensive meteorological data are obtained.

[0065] More preferably, the ratio of diffusion concentration to peak point is calculated as an influence coefficient; The impact coefficients are sorted in descending order, and the monitoring station numbers corresponding to the descending-order impact coefficients are obtained, thus obtaining the new pollution migration order for each monitoring station.

[0066] More preferably, the method for setting the pollution migration status includes selecting any peak point of any monitoring station, using the peak point as a reference, filtering in both the reverse time direction and the sequential time direction, selecting the monotonicity boundary points in both the reverse time direction and the sequential time direction, recording them as terminal values, and marking the terminal values ​​with a second mark. Construct the correspondence between the first tag and the second tag.

[0067] More preferably, the monotonicity dividing point is represented as the historical pollutant concentration at which the value changes from monotonically decreasing to monotonically increasing in the chronological or reverse chronological direction from that peak point. When there are 0 terminal values, jump to the peak point corresponding to the next data segment and repeat the terminal value acquisition step. When all data segments are jumped and there are 0 terminal values, set the pollution migration status to the non-migrated status. When there is one terminal value, jump to the peak point corresponding to the next data segment and repeat the terminal value acquisition step. When all data segments are jumped and there is one terminal value in each segment, set the pollution migration status to the state of being migrated but not transferred. When there are two terminal values, the pollution migration status is set to the state that it has been migrated and transferred.

[0068] More preferably, the logical order of setting the pollution migration status includes determining whether there are two terminal values; If there are two terminal values, the pollution migration status is "has been migrated and has been transferred"; If there are no two terminal values, determine if there is one terminal value. If there is one terminal value, the pollution migration status is "has been migrated and has not been transferred"; When there is no terminal value, the pollution migration status is set to the unmigrated state.

[0069] Step S15 includes the following sub-steps: Step S151, for the unmigrated state, jump to the next set of historical contamination data and re-analyze the contamination migration state until the contamination migration state is converted to the migrated state or all historical contamination data has been jumped to, and mark the initial contamination timestamp. More preferably, if all historical pollution data is accessed and no initial pollution timestamp is marked, then the initial pollution timestamp of the corresponding monitoring station is marked as 0.

[0070] Step S152: For the migrated but not transferred state and the transferred state, mark the initial contamination timestamp; More preferably, the initial contamination timestamp is represented as the timestamp corresponding to the monotonic boundary point that precedes the time sequence.

[0071] Step S2 includes the following sub-steps: Step S21, based on the time sequence of the initial pollution timestamp of each monitoring station, obtain the historical groundwater level corresponding to each monitoring station at the initial pollution timestamp, and obtain N sets of water level sequences. More preferably, the initial contamination timestamp that is ranked first is obtained and recorded as the reference timestamp; The historical groundwater levels of each monitoring station at the reference timestamp are obtained and recorded as the first water level sequence. Based on the order of the initial pollution timestamps, the water level sequences from the second group to the Nth group are obtained, where N is a natural number.

[0072] More preferably, the water level sequence is arranged according to the order of pollution transfer. Step S22: Based on the water level sequence and the coordinates of each monitoring station, the spatiotemporal correlation characteristics are obtained.

[0073] Select any set of water level sequences; Set the values ​​for "not migrated", "migrated but not transferred", and "transferred" to 0, 1, and 2 respectively; Obtain the pollution transfer status of the monitoring stations corresponding to the water level sequence, and obtain the monitoring station number corresponding to the water level sequence; Construct a number vector based on the monitoring station number, a status vector based on the pollution transfer status, and a water level vector based on the water level sequence. The number vector, state vector and water level vector are concatenated from top to bottom to obtain the correlation matrix. The initial pollution timestamps of each group are traversed to obtain the correlation matrix corresponding to each group of initial pollution timestamps. Set the initial contamination timestamp as the label of the association matrix.

[0074] More preferably, the number of columns in each spatial correlation feature vector is kept equal, and the columns are automatically adjusted according to the monitoring station number, so that the columns of each spatial correlation feature vector represent the data of the same monitoring station.

[0075] More preferably, the eigenvalues ​​of the spatial correlation matrix are obtained based on the eigenvalue decomposition principle of linear algebra.

[0076] More preferably, since the initial pollution timestamps are discrete data, the method for truncating the feature values ​​includes: performing linear interpolation on the feature value sequence to generate an equally spaced time series; when the interval between adjacent timestamps is greater than the first duration, cubic spline interpolation is used to supplement it, and the feature values ​​of the original timestamp positions are retained after interpolation, and only the missing segments are filled.

[0077] Set the correlation matrix as a spatiotemporal correlation feature.

[0078] Step S3: Based on the spatiotemporal correlation characteristics, construct the correlation between pollution migration status and groundwater level, and obtain the groundwater level of the target monitoring station based on the correlation and the current pollution migration status of each monitoring station.

[0079] The correlation is represented as the correspondence between spatiotemporal correlation features and the corresponding initial pollution timestamp; Obtain the current pollution migration status of each monitoring station, and based on the number corresponding to the monitoring station and the number corresponding to the pollution migration status, obtain the first two rows of elements in the spatial correlation matrix; Input the first two rows of elements in the spatial correlation matrix into the correlation relationship to obtain candidate spatiotemporal correlation features, obtain the water level vector in the candidate spatiotemporal correlation features, and set the water level vector as the first candidate groundwater level value. Obtain the historical pollution migration status from the first historical moment after the current moment to the current moment, and obtain the second groundwater level candidate value according to the method for obtaining groundwater level candidate values.

[0080] Obtain the initial pollution timestamp corresponding to each candidate value, obtain the first groundwater level candidate value and the second groundwater level candidate value adjacent to the initial pollution timestamp, and determine the groundwater level of the target monitoring station based on the first groundwater level candidate value and the second groundwater level candidate value adjacent to the initial pollution timestamp.

[0081] More preferably, this application can achieve the same data results through another parallel extraction method that is substantially the same as the method in this application: replacing the initial pollution timestamp with the pollution transfer timestamp, resetting the spatiotemporal correlation features, and obtaining the groundwater level based on the spatiotemporal correlation features.

[0082] More preferably, the first groundwater level candidate value and the second groundwater level candidate value adjacent to the initial pollution timestamp are set as candidate groups, and the number of candidate groups is counted. When the number of candidate groups is greater than or equal to 1, the first time period is traced back to obtain candidate values ​​that are adjacent in time sequence until the number of candidate groups is equal to 1. Then, the corresponding first groundwater level candidate value is obtained and set as the target water level sequence. When the number of candidate groups is equal to 1, the corresponding first groundwater level candidate value is obtained and set as the target water level sequence.

[0083] Obtain the target monitoring station number, obtain its column number in the spatial correlation matrix based on the target monitoring station number, obtain the corresponding water level in the target water level sequence based on the column number, and set the water level as the groundwater level of the target monitoring station.

[0084] More preferably, the Binhai Chemical Industrial Park was selected as the area to be analyzed. Five monitoring wells, M1 to M5, were deployed within the park. The monitoring period was from 00:00 on January 1, 2024 to 24:00 on January 31, 2024, with data collected every 6 hours, totaling 124 historical monitoring time points. Two pollution sources were identified within the park. The coordinates of the geometric center of each monitoring well are as follows: More preferably, the method for setting the location of the pollution source includes retrieving the sewage discharge registration information of the park, setting the geometric center of the sewage outlet A as W1, setting the geometric center of the sewage channel B as W2, where W1 is a point source and W2 is a surface source.

[0085] More preferably, an example of meteorological data (average value for the period from 00:00 to 06:00 on January 5, 2024): Average wind speed: 3.2 m / s, average wind direction: east to north (vector composite direction angle 45°), average precipitation: 0 mm, average evaporation: 1.5 mm.

[0086] More preferably, an example of historical pollution diffusion data (COD concentration in well M1, unit mg / L): Retrieve the WGS-84 ellipsoid model. Taking M1 as an example, φ is 39.9005∘, λ is 116.4010∘, and h is 11m. The radius of curvature of the ramusoidal circle is 387112.4m. Substituting this into the ellipsoidal model, we obtain the coordinates of M1 as (−2171204.3, 4385231.7, 4072890.6). Similarly, we obtain the coordinates of W1, W2, and M2~M5. Initial straight-line distance range: W1 point source initial radius 500 m; W2 surface source geometric outer circle radius 300 m, expanding outward by 300 m, total 600 m; The steps of the exclusion method include: there is a mountain within the range of the line connecting M2 and W1, the elevation of the geometric center of the mountain is 35 m, and the average height of the topographic profile connecting the pollution source and the monitoring station is 28 m. Since 35 is greater than 28, the terrain is determined to be an obstacle, and M2 and W1 are excluded. There is a karst development area within the line connecting M3 and W2. It is a strongly developed area and the underground river flows from M3 to W2 (not from the pollution source to the monitoring station). It is determined that the pollutant migration path is blocked, so M3 and W2 are excluded. M5 and W2 belong to different surface water basins, and the river watershed is located within the line connecting the two, so M5 and W2 are excluded; Sort the remaining candidate sources in ascending order of linear distance and take the minimum value.

[0087] More preferably, the result of the membership source setting is as follows: Taking the W1 group (M1, M3, M5) as an example, the first time period is 24 hours. The maximum value of the COD time series is extracted from the data segment Pi1 (i is the monitoring well number) on January 5th using a sliding time window: Sort by timestamp instantaneous order, the pollution migration sequence number string is 1,3,5, indicating that the pollution was transferred from M1 to M3 and then to M5.

[0088] Taking M1 as an example, the calculation process for the pollution affiliation ratio is described as follows: Reference vector: pointing from W1 to M1; Key time point: The historical emission time point with the smallest absolute value of the time difference from the peak point of 1 / 5 06:00 is 1 / 500:00; The number of historical monitoring time points between 00:00 and 06:00 is 2 (including endpoints), and the number of wind directions is 1; During this period, there is only one wind direction vector, with the overall wind direction being northeast at 45°. The cosine of the angle between the wind direction vector and the reference vector is 0.82, and the pollution belonging to this vector is 0.82.

[0089] More preferably, taking the COD sequence (time order) of M1 as an example: [12,35,28,20,15,18] (1 / 5 00:00 to 1 / 6 06:00); Using the peak point 35 (1 / 5 06:00) as the baseline, the time sequence is reversed (going backwards): 35→12 (decreasing), before 12 is 15 (changing from decreasing to increasing), and 15 is the monotonicity dividing point, denoted as the terminal value A. The time sequence direction (towards the future) is: 35→28→20→15 (decreasing), after 15 is 18 (changing from decreasing to increasing), and 18 is the monotonic dividing point, denoted as the terminal value B; There are two terminal values. Set the M1 pollution migration status to be migrated and transferred (status value 2). Similarly, if there is one M3 terminal value, it is set to have been migrated and not transferred (state value 1); if there are zero M5 terminal values ​​(still zero after jumping to all segments), it is set to not be migrated (state value 0).

[0090] More preferably, the initial contamination timestamps are: the timestamp corresponding to the terminal value A with the earlier time sequence of M1 is 1 / 4 18:00; the timestamp corresponding to the terminal value of M3 is 1 / 5 06:00; Initial transfer timestamp: The terminal value B, which is later in the M1 time sequence, corresponds to the timestamp 1 / 6 00:00; Direction of transfer: M1 points in a straight line towards the next monitoring station M3 in the pollution migration sequence, that is, from M1 to M3.

[0091] More preferably, the initial pollution timestamps of each monitoring well are obtained and sorted by time: Using T1 as the reference timestamp, obtain the historical groundwater level (in meters) of each monitoring well at time T1: This group is arranged according to the W1 pollution migration sequence (1,3,5) and the remaining number sequence, and is denoted as the first group of water level sequences. Similarly, the second to fifth groups of water level sequences are obtained.

[0092] For the first group of water level sequences: Numbering vector: [1,3,5,2,4], State vector (corresponding to pollution migration state at any given time): [2,1,0,0,0], Water level vector: [12.5,11.8,13.2,10.5,9.8]; Concatenate the three vectors from top to bottom to obtain a 3×5 incidence matrix: ; Set the initial pollution timestamp T1 as the label of the correlation matrix. Iterate through T2 to T5 to obtain 5 correlation matrices, and the columns of each matrix are automatically aligned according to the monitoring well number (the first column is always M1, the second column is M3, and so on).

[0093] More preferably, step S3 is illustrated with an example: Prediction target: The groundwater level of monitoring well M3 is predicted as of 2024-01-2012:00 (current time). Current input: The current pollution migration status of M3 is that it has been migrated but not transferred (status value 1), and the M3 number is 3.

[0094] Input ID 3 and state 1 into the association database, and retrieve the column in each association matrix where the first row is ID 3 and the second row is state 1. Assume a successful match is found in the second association matrix (T2=1 / 506:00); In the second group of correlation matrices, the water level corresponding to column M3 is 11.5m, which is recorded as the first candidate value for groundwater level; Tracing back to the first time period (24 hours), i.e., 12:00 on January 19, 2024. Obtain the pollution migration status of M3 at this time, assuming it is still in status 1, the corresponding M3 water level in the correlation matrix is ​​11.2m, which is recorded as the second candidate value of groundwater level; The first candidate value corresponds to timestamp T2, and the second candidate value corresponds to timestamp T_a. If T2 and T_a are not adjacent in the sorted timestamp sequence, the process continues to trace back until a unique pair of adjacent timestamps is found. Assuming that the final unique candidate group is the second group, its water level vector is the target water level sequence: [12.2, 11.5, 13.0, 10.3, 9.6]; Based on the number 3 of M3, its column number in the correlation matrix is ​​obtained as the second column. The value of the second column, 11.5m, is extracted from the target water level sequence and set as the predicted groundwater level of monitoring well M3 at the current moment.

[0095] More preferably, the operating logic of the system modules is described below: The module is set to execute S11~S15, including: Receives sewage discharge information, meteorological data, and historical pollution diffusion data; calls the WGS-84 coordinate transformation submodule to complete ECEF coordinate calculation; calls the source affiliation determination submodule to execute topographic, hydrological, and watershed triple exclusion logic; outputs labels for each monitoring well and timestamps of pollution migration status. The analysis module executes S21~S22, including: N sets of water level sequences are generated by sorting according to the initial pollution timestamp; number vector, state vector, and water level vector are constructed and concatenated into an incidence matrix; linear interpolation or cubic spline interpolation is performed on missing timestamps to ensure matrix column alignment.

[0096] The mapping module executes S3: receiving the current pollution migration status of each monitoring well; matching the dual conditions of number and status in the association matrix library; performing 24-hour backtracking and adjacent timestamp locking; and outputting the predicted groundwater level of the target monitoring well.

[0097] In this embodiment, the time points from January 20th to January 25th, 2024 for well M3 are predicted and compared with the measured water level: The above operational data shows that, based on the mapping relationship between pollution migration status and spatiotemporal correlation characteristics, groundwater levels can be effectively predicted through spatiotemporal analysis of pollution status at monitoring stations without relying on complex hydrological mechanism models.

[0098] More preferably, this invention establishes a timestamp of pollution migration status by fusing and analyzing pollution source location, meteorological data, and historical pollution diffusion data. This allows for spatiotemporal alignment of the pollutant diffusion process with dynamic changes in groundwater levels, providing an accurate synchronization benchmark for subsequent water level prediction. By extracting historical groundwater levels corresponding to each timestamp and constructing a correlation matrix between water level sequences and pollution migration status, the inherent connection between pollutant migration paths and groundwater level responses can be fully explored. Multidimensional data is integrated into structured spatiotemporal correlation features. By directly mapping pollution migration status to groundwater levels, easily monitored surface pollution diffusion data can indirectly serve groundwater level prediction, solving the problem of insufficient or difficult-to-obtain groundwater level monitoring data in traditional methods and reducing prediction costs.

[0099] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0100] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A groundwater level prediction method based on big data, characterized in that, The steps include: Step S1, determining the location of the pollution source and setting a timestamp for the pollution migration status by combining meteorological data and historical pollution diffusion data; Step S2: Obtain the historical groundwater level corresponding to each timestamp, and obtain the spatiotemporal correlation features based on the historical groundwater level; Step S3: Based on the spatiotemporal correlation characteristics, construct the correlation between pollution migration status and groundwater level, and obtain the groundwater level of the target monitoring station based on the correlation and the current pollution migration status of each monitoring station.

2. The groundwater level prediction method based on big data as described in claim 1, characterized in that, Meteorological data includes wind speed, wind direction, precipitation, and evaporation. Historical pollution diffusion data includes historical monitoring locations, corresponding historical monitoring time points, and corresponding historical pollutant concentrations.

3. The groundwater level prediction method based on big data as described in claim 1, characterized in that, Step S1 includes the following sub-steps: Step S11, set the location of the pollution source based on the sewage discharge information of the area to be analyzed; Step S12: Obtain meteorological data and historical pollution diffusion data for the area to be analyzed; Step S13: Based on the distance between historical pollution data and pollution sources, set the affiliated source; Step S14: Set the pollution migration status based on the historical pollution data of the affiliated source; Step S15: Set pollution-related timestamps based on the pollution migration status.

4. The groundwater level prediction method based on big data as described in claim 3, characterized in that, Step S13 includes the following sub-steps: Step S131, convert the location of the monitoring station and the location of the pollution source to the world coordinate system; Step S132: Calculate the straight-line distance between any monitoring station and each pollution source according to the Euclidean distance formula, and set the affiliated source based on the straight-line distance and the terrain features within the straight-line distance range; Step S133: Traverse each monitoring station to obtain the corresponding source of each station.

5. The groundwater level prediction method based on big data as described in claim 3, characterized in that, Pollution migration status includes not migrated, migrated but not transferred, and transferred; Pollution-related timestamps include the initial pollution timestamp and the initial transfer timestamp; Step S14 includes the following sub-steps: Step S141, based on the historical pollution diffusion data of the monitoring stations, mark the peak points, extract the timestamps of the peak points, obtain the timestamps of the peak points of each monitoring station under the source, and set the pollution migration order according to the timestamps of each peak point. Step S142: Obtain historical emission data of the affiliated sources, and set the pollution affiliation ratio based on historical emission data, meteorological data, and peak point timestamps; Step S143 includes the following sub-steps: resetting the pollution migration order according to the pollution affiliation ratio, and setting the pollution migration status based on the reset pollution migration order; Step S15 includes the following sub-steps: Step S151, for the unmigrated state, jump to the next set of historical contamination data and re-analyze the contamination migration state until the contamination migration state is converted to the migrated state or all historical contamination data has been jumped to, and mark the initial contamination timestamp. Step S152: For the migrated but not transferred state and the transferred state, mark the initial contamination timestamp.

6. The groundwater level prediction method based on big data as described in claim 1, characterized in that, Step S2 includes the following sub-steps: Step S21, based on the time sequence of the initial pollution timestamp of each monitoring station, obtain the historical groundwater level corresponding to each monitoring station at the initial pollution timestamp, and obtain N sets of water level sequences. Step S22: Based on the water level sequence and the coordinates of each monitoring station, obtain the spatiotemporal correlation characteristics.

7. The groundwater level prediction method based on big data as described in claim 6, characterized in that, Select any set of water level sequences; Set the values ​​for "not migrated", "migrated but not transferred", and "transferred" to 0, 1, and 2 respectively; Obtain the pollution transfer status of the monitoring stations corresponding to the water level sequence, and obtain the monitoring station number corresponding to the water level sequence; Construct a number vector based on the monitoring station number, a status vector based on the pollution transfer status, and a water level vector based on the water level sequence. The number vector, state vector and water level vector are concatenated from top to bottom to obtain the correlation matrix. The initial pollution timestamps of each group are traversed to obtain the correlation matrix corresponding to each group of initial pollution timestamps. Set the initial contamination timestamp as the label of the association matrix.

8. The groundwater level prediction method based on big data as described in claim 7, characterized in that, Set the correlation matrix as a spatiotemporal correlation feature.

9. The groundwater level prediction method based on big data as described in claim 1, characterized in that, The correlation is represented as the correspondence between spatiotemporal correlation features and the corresponding initial pollution timestamp; Obtain the current pollution migration status of each monitoring station, and based on the number corresponding to the monitoring station and the number corresponding to the pollution migration status, obtain the first two rows of elements in the spatial correlation matrix; Input the first two rows of elements in the spatial correlation matrix into the correlation relationship to obtain candidate spatiotemporal correlation features, obtain the water level vector in the candidate spatiotemporal correlation features, and set the water level vector as the first candidate groundwater level value. Obtain the historical pollution migration status from the first historical moment after the current moment to the current moment, and obtain the second groundwater level candidate value according to the method for obtaining groundwater level candidate values; Obtain the initial pollution timestamp corresponding to each candidate value, obtain the first groundwater level candidate value and the second groundwater level candidate value adjacent to the initial pollution timestamp, and determine the groundwater level of the target monitoring station based on the first groundwater level candidate value and the second groundwater level candidate value adjacent to the initial pollution timestamp.

10. A groundwater level prediction system based on big data, the system being used to execute the groundwater level prediction method based on big data as described in claim 1, characterized in that, It includes a settings module, an analysis module, and a mapping module; The setting module determines the location of the pollution source and sets a timestamp for the pollution migration status by combining meteorological data and historical pollution diffusion data. The analysis module obtains the historical groundwater level corresponding to each timestamp, and obtains spatiotemporal correlation features based on the historical groundwater level; The mapping module constructs a correlation between pollution migration status and groundwater level based on spatiotemporal correlation characteristics, and obtains the groundwater level of the target monitoring station based on the correlation and the current pollution migration status of each monitoring station.

Images