Groundwater pollution source tracing method based on multi-modal data fusion

Abstract

The present application belongs to the technical field of data processing, and relates to a groundwater pollution source tracing method based on multi-modal data fusion. The present application obtains the abandoned well cross-layer value, well sealing failure value, water head difference sequence, conductivity profile sequence and deep and shallow layer concentration sequence of the target site in each groundwater monitoring period, judges whether there is a risk of jump change of conduction, and screens a candidate well set; then generates a water head difference reversal value, a wellbore conduction value and a pollution layer position jump value, switches the deep independent source identification corresponding to the deep pollution anomaly to a conduction influence identification; finally, based on the conduction influence identification, the candidate area ranking value, the pollution source layer position value and the pollution migration path value are corrected to obtain the pollution source position value. The present application can reduce the misjudgment risk of deep pollution source caused by the conduction of abandoned well, and improve the reliability of the groundwater pollution source tracing result.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for tracing and tracking groundwater pollution sources based on multimodal data fusion. Background Technology

[0002] In groundwater pollution investigation, contaminated site remediation supervision, and industrial park environmental risk management, groundwater pollution source tracing technology based on multimodal data fusion is commonly used to determine the location, stratigraphic level, and migration path of pollution sources. This type of technology typically inputs monitoring well water quality data, groundwater level data, stratigraphic data, time-series pollutant concentration data, historical production data, and site spatial information into the tracing model. Through spatial overlay, time-series comparison, pollutant fingerprint matching, and groundwater flow direction analysis, it reverse-engineers candidate pollution source areas and pollution migration paths.

[0003] However, existing technologies, when performing data fusion, typically treat pollution anomalies in monitoring wells directly as evidence of pollution in the corresponding aquifer or planar location. This involves ranking pollution sources based on pollutant concentration, anomaly occurrence time, pollutant fingerprint similarity, and the spatial relationship between the monitoring well and the candidate pollution source area. This approach assumes that monitoring well anomalies can stably reflect the pollution status of the corresponding strata and is suitable for scenarios with relatively simple aquifer structures and continuous pollutant migration along conventional groundwater flows. However, in actual contaminated sites, abandoned wells, abandoned exploration boreholes, or old monitoring wells are common in old industrial areas, landfill perimeters, mining areas, and historically contaminated sites. These abandoned wells may penetrate both shallow and deep aquifers, and due to well shaft damage, missing casing, filter media penetration, or aging sealing materials, potential vertical channels may form. When the head difference between different aquifers changes due to high water season, low water season, heavy rainfall, or the start-up and shutdown of pumping wells, abandoned wells may form intermittent vertical channels driven by periodic changes in head difference, allowing shallow pollutants to enter deep aquifers or causing deep polluted water to flow back to shallow aquifers.

[0004] Because existing source tracing technologies typically fail to perform unified correlation analysis of stratified head changes, abandoned well spatial structure, wellbore conductivity profiles, and the lag response of shallow and deep aquifer pollution concentrations, misjudgments of pollution layers are prone to occur in the aforementioned scenarios. For example, after shallow pollutants enter deeper layers via abandoned wells, the system may misclassify deep-layer anomalies as independent deep-layer pollution sources; when shallow and deep monitoring wells exhibit alternating anomalies at different times, the system may misjudge multiple pollution sources acting together. This problem directly impacts the reliability of pollution source tracing results and remediation decisions. On one hand, the ranking of candidate pollution sources may be incorrect, ignoring genuine shallow-layer pollution sources; on the other hand, drainage wells, barrier walls, in-situ chemical injection points, or stratified monitoring wells may be deployed at incorrect depths, increasing remediation costs and reducing remediation effectiveness. In severe cases, the system may also underestimate the risk of deep aquifer pollution diffusion, affecting deep water supply wells and drinking water safety. Summary of the Invention

[0005] The purpose of this invention is to solve the problem that in the existing process of tracing groundwater pollution sources, the periodic opening of abandoned wells leads to the misjudgment of deep pollution anomalies as deep independent pollution sources, resulting in errors in the ranking of pollution source candidate areas, the determination of pollution source strata, and the determination of pollution migration paths. Therefore, this invention proposes a groundwater pollution source tracing method based on multimodal data fusion.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a groundwater pollution source tracing method based on multimodal data fusion, comprising the following steps: obtaining the abandoned well penetration value, well sealing failure value, hydraulic head difference sequence, conductivity profile sequence and deep and shallow concentration sequence of the target site in each groundwater monitoring cycle.

[0007] Based on the abandoned well penetration value, well sealing failure value, and head difference sequence, it is determined whether there is a risk of conduction jump at the target site. If so, a set of candidate wells is selected based on the abandoned well penetration value and well sealing failure value.

[0008] Based on the head difference sequence corresponding to each candidate conductive well in the candidate well set, a head difference inversion value is generated.

[0009] Based on the head difference reversal value and conductivity profile sequence, wellbore conduction value is generated.

[0010] Based on wellbore conductivity values ​​and shallow and deep layer concentration sequences, polluted layer jump values ​​are generated.

[0011] Deep pollution anomalies are identified based on shallow and deep concentration sequences, and the deep independent source identifiers corresponding to deep pollution anomalies are switched to conduction impact identifiers based on the pollution layer jump values.

[0012] Based on the conduction impact indicator, the candidate area ranking value, pollution source layer value, and pollution migration path value are corrected to obtain the pollution source location value.

[0013] In summary, by employing the aforementioned groundwater pollution source tracing method based on multimodal data fusion, the beneficial effects of this invention are as follows: This invention determines whether a target site has a risk of conduction jumps by using abandoned well penetration values, well sealing failure values, and hydraulic head difference sequences. When a risk exists, a candidate well set is selected based on the abandoned well penetration values ​​and well sealing failure values. This approach does not directly treat all abandoned wells as vertical conduction channels for source tracing; instead, it first excludes abandoned wells with only penetration structures but good sealing conditions, as well as abandoned wells with failed sealing but without forming cross-layer channels. This solves the problem of an overly broad scope for assessing the conduction risk of abandoned wells, which easily leads to misjudgments, and improves the targeting of candidate conduction well selection.

[0014] This invention generates a head difference inversion value through the head difference sequence corresponding to candidate conductive wells, and generates a wellbore conduction value based on the head difference inversion value and conductivity profile sequence. Furthermore, it combines deep and shallow layer concentration sequences to generate a contaminated layer jump value. This method correlates and judges head difference inversion, wellbore water quality interface movement, and the lag response of deep and shallow layer contamination concentrations, solving the problem of existing source tracing methods that directly identify independent deep-layer pollution sources based solely on pollution anomalies in deep monitoring wells. It can reduce the risk of misjudging contaminated layers caused by intermittent vertical conduction of abandoned wells.

[0015] This invention, when the pollution stratum jump value meets the conditions, switches the deep independent source identifier corresponding to the deep pollution anomaly to a conduction influence identifier, and corrects the candidate area ranking value, pollution source stratum value, and pollution migration path value based on the conduction influence identifier. This method can correct deep anomaly evidence from deep independent pollution sources to shallow pollution sources with conduction influence via abandoned wells, solving the problems of incorrect pollution source candidate area ranking, incorrect pollution source stratum determination, and incorrect pollution migration path determination, and improving the reliability of pollution source location values.

[0016] This invention re-determines the identifier switching value during subsequent groundwater monitoring cycles. When the identifier switching value continuously falls below a preset switching threshold for a preset number of recovery cycles, the conductivity impact identifier is restored to a deep independent source identifier. This method avoids the system maintaining the conductivity impact identifier for an extended period due to a single head difference reversal or a single pollution peak lag. It allows the pollution source location value to be dynamically updated according to changes in the conductivity status of abandoned wells, improving the adaptability of source tracing results in continuous monitoring scenarios. Attached Figure Description

[0017] Figure 1 A schematic diagram of the overall processing flow of the present invention is shown;

[0018] Figure 2 A schematic diagram of the process for switching to the conduction effect indicator according to the present invention is shown;

[0019] Figure 3The present invention illustrates a flowchart of the process for determining whether a target site has a risk of conduction switching. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The specific processing procedure of the present invention will be described below with reference to embodiments.

[0022] Example 1:

[0023] See Figure 1 The groundwater pollution source tracing method based on multimodal data fusion includes the following steps: obtaining the abandoned well penetration value, well sealing failure value, hydraulic head difference sequence, conductivity profile sequence and deep and shallow concentration sequence of the target site in each groundwater monitoring cycle.

[0024] Considering that the periodic opening of abandoned wells can cause changes in the contamination layer, it is necessary to rely on the structural state of the abandoned well, the head drive state between different aquifers, the mixing state of the water inside the well, and the response relationship of the pollution concentration between shallow and deep layers. Collecting only a single water quality value or a single water level value cannot distinguish the influence of deep independent pollution sources and the opening of abandoned wells. Therefore, it is necessary to collect and align the multiple monitoring values ​​related to the opening of abandoned wells in the target site in a unified manner.

[0025] In one specific embodiment, the groundwater monitoring cycle refers to the time interval for simultaneously collecting hydraulic head, water quality, and pollution concentration values ​​from shallow and deep aquifers within the target site.

[0026] Specifically, the abandoned well penetration value is used to characterize the degree to which an abandoned well simultaneously penetrates both shallow and deep aquifers. It can be determined based on the abandoned well depth, wellbore location, the top and bottom plate depths of the shallow and deep aquifers, and the overlap length between the abandoned well and each aquifer. In one specific embodiment, the abandoned well penetration value can be represented by a length value. First, based on the abandoned well depth, wellbore location, and the top and bottom plate depths of the shallow and deep aquifers, a first overlap length between the abandoned well and the shallow aquifer, and a second overlap length between the abandoned well and the deep aquifer, are determined respectively. Then, when both the first and second overlap lengths are greater than zero, the smaller of the first and second overlap lengths is determined as the abandoned well penetration value; when either the first or second overlap length is zero, the abandoned well penetration value is determined to be zero. The well sealing failure value is used to characterize the degree of sealing failure in abandoned wells and can be determined based on the absence of sealing materials, casing damage, filter media penetration, and missing well sealing records. In one specific embodiment, the well sealing failure value can be represented by a value between zero and one. First, failure sub-values ​​are set for the absence of sealing materials, casing damage, filter media penetration, and missing well sealing records. When the corresponding condition does not exist, the corresponding failure sub-value is set to zero. When the corresponding condition exists slightly and does not form a continuous channel, the corresponding failure sub-value is set to an intermediate value greater than zero and less than one. When the corresponding condition exists significantly and may form a continuous channel, the corresponding failure sub-value is set to one. Then, the maximum value among the multiple failure sub-values ​​is determined as the well sealing failure value. The head difference sequence includes multiple stratified head difference values ​​arranged according to the groundwater monitoring cycle, where the stratified head difference is the difference between the head value of shallow aquifers and the head value of deep aquifers. The conductivity profile sequence includes conductivity values ​​collected at different wellbore depths from the same candidate conductive well. The shallow and deep aquifer concentration sequences include the shallow aquifer pollution concentration values ​​and deep aquifer pollution concentration values ​​corresponding to the same pollutant index within the same groundwater monitoring period. The shallow and deep aquifer pollution concentration values ​​are stored using a unified timestamp for subsequent comparison of shallow and deep pollution peak times. When multiple pollutant indexes exist at the target site, corresponding shallow and deep aquifer concentration sequences are generated for each pollutant index, and the corresponding contaminated layer jump values ​​are calculated separately. Subsequently, the larger of the contaminated layer jump values ​​corresponding to each pollutant index can be used as the contaminated layer jump value within the target groundwater monitoring period. Thus, the abandoned well penetration value can characterize the effective penetration degree of an abandoned well simultaneously penetrating both shallow and deep aquifers using a length value, and the well sealing failure value can characterize the most severe failure factor in the sealed state of an abandoned well using a value between zero and one. Subsequently, the abandoned well penetration value and well sealing failure value are used to determine the structural risk value of the abandoned well and further used to screen candidate well sets.

[0027] See Figure 3Based on the abandoned well penetration value, well sealing failure value, and head difference sequence, it is determined whether there is a risk of conduction jump at the target site. If so, a set of candidate wells is selected based on the abandoned well penetration value and well sealing failure value.

[0028] It should be noted that the methods for determining whether there is a risk of conduction shift at the target site include: considering that there may be multiple abandoned wells at the target site, but not all abandoned wells will form a conduction channel between shallow and deep aquifers, the risk of contaminated layer shift cannot be directly determined solely based on the existence of abandoned wells. To avoid misjudgment, it is necessary to simultaneously examine whether the abandoned wells have the basis for cross-layer and well sealing failure, and whether there is a periodically changing hydraulic head driving force between different aquifers.

[0029] In one specific embodiment, firstly, the structural risk value of the abandoned well is determined based on the abandoned well penetration value and the well sealing failure value corresponding to each abandoned well. The larger the abandoned well penetration value and the larger the well sealing failure value, the greater the structural risk value of the abandoned well.

[0030] Specifically, to ensure that the structural risk value of abandoned wells simultaneously reflects both the degree of cross-contamination and the degree of well sealing failure, in one specific embodiment, the cross-contamination value and the well sealing failure value of the abandoned well are first normalized to obtain a normalized value for cross-contamination structure and a normalized value for well sealing failure. The normalized value for cross-contamination structure can be obtained by normalizing the cross-contamination value of the abandoned well based on a preset upper limit for cross-contamination, and the normalized value for well sealing failure can be obtained by normalizing the well sealing failure value based on a preset upper limit for failure. Subsequently, the smaller of the normalized value for cross-contamination structure and the normalized value for well sealing failure corresponding to the same abandoned well is determined as the structural risk value of the abandoned well. Therefore, only when an abandoned well simultaneously possesses a high degree of cross-contamination and a high degree of well sealing failure will its structural risk value be high.

[0031] Then, based on the positive and negative changes in the monitoring periods of adjacent groundwater in the head difference sequence, the number of changes in the direction of the head difference is determined. Specifically, when the sign of the head difference values ​​of two adjacent layers changes, it is considered that the direction of the head difference has changed once.

[0032] Subsequently, the periodic variation value of the head difference is determined based on the number of changes in the direction of the head difference and the maximum absolute value in the head difference sequence. The more times the head difference changes direction and the larger the maximum absolute value, the larger the periodic variation value of the head difference.

[0033] Specifically, to avoid misjudging periodic changes solely due to accidental changes in the direction of head difference or large single-time head difference amplitudes, in one specific embodiment, the number of head difference direction changes is normalized to obtain a normalized direction change value; the maximum absolute value in the head difference sequence is normalized to obtain a normalized head difference amplitude value; subsequently, the smaller of the normalized direction change value and the normalized head difference amplitude value is determined as the periodic head difference value. Thus, the periodic head difference value is constrained by both the number of direction changes and the head difference amplitude.

[0034] Specifically, the normalized value for the direction change can be obtained by dividing the number of changes in the direction of the head difference by a preset upper limit for the number of changes in the direction; the normalized value for the head difference amplitude can be obtained by dividing the maximum absolute value in the head difference sequence by a preset upper limit for the head difference amplitude. When the normalized value for the direction change or the normalized value for the head difference amplitude is greater than one, the corresponding normalized value is set to one. The preset upper limit for the number of changes in the direction of the head difference can be determined based on the larger value of the number of changes in the direction of the head difference during the historical groundwater monitoring period of the target site; the preset upper limit for the head difference amplitude can be determined based on the larger absolute value of the stratified head difference during the historical groundwater monitoring period of the target site.

[0035] Furthermore, the structural risk value and the periodic variation value of the head difference of the abandoned well are normalized to obtain the normalized value of the structural risk and the normalized value of the periodic variation. The smaller value between the normalized value of the structural risk and the normalized value of the periodic variation corresponding to the same abandoned well is determined as the conduction risk value.

[0036] Finally, when the conduction risk value corresponding to at least one abandoned well is greater than or equal to the preset conduction risk threshold, it is determined that there is a conduction jump risk at the target site; when the conduction risk values ​​corresponding to all abandoned wells are less than the preset conduction risk threshold, it is determined that there is no conduction jump risk at the target site.

[0037] Specifically, the preset conduction risk threshold can be determined based on manually confirmed conduction samples of abandoned wells, records of abandoned well sealing verification, and records of changes in stratified head differences within the historical monitoring period of the target site. When there are few manually confirmed samples for a particular plot, the preset conduction risk threshold can be set to a value between zero and one, and adjusted according to the results of subsequent on-site verification.

[0038] It should be noted that the method for selecting the candidate well set based on the cross-layer value and sealing failure value of abandoned wells includes the following: Since candidate connected wells are the objects of subsequent calculations of head difference reversal value, wellbore continuity value, and contaminated layer jump value, including all abandoned wells in the calculation would increase the risk of misjudgment and the computational burden. Therefore, it is necessary to first exclude abandoned wells that only have cross-layer structures but are in good sealing condition, as well as abandoned wells that have failed to form cross-layer channels.

[0039] In one specific embodiment, firstly, a preset perforation upper limit value, a preset failure upper limit value, and a preset screening threshold are obtained. The preset perforation upper limit value and the preset failure upper limit value are both positive values.

[0040] Then, the cross-layer value corresponding to each abandoned well is divided by the preset cross-layer upper limit value to obtain the cross-layer normalization value. When the cross-layer normalization value is greater than one, the cross-layer normalization value is set to one.

[0041] Subsequently, the failure value of the sealing of each abandoned well is divided by the preset failure upper limit to obtain the failure normalization value. When the failure normalization value is greater than one, the failure normalization value is set to one.

[0042] Furthermore, the smaller of the cross-layer normalization value and the failure normalization value corresponding to the same abandoned well is determined as the single-well conduction screening value.

[0043] Finally, when the single-well conduction screening value is greater than or equal to the preset screening threshold, the corresponding abandoned well is identified as a candidate conduction well, and a candidate well set is formed based on all candidate conduction wells. The candidate well set is used to subsequently extract the corresponding head difference sequence and generate a head difference inversion value. To ensure that each candidate well in the candidate well set corresponds to the head difference sequence, in one specific embodiment, firstly, the spatial location of the wellhead and the depth of the wellbore extension of the candidate well are obtained; then, based on the spatial location of the wellhead of the candidate well, the shallow aquifer monitoring well and the deep aquifer monitoring well closest to the candidate well are selected within the target site, or the monitoring locations of the shallow aquifer and the deep aquifer respectively are selected among the monitoring wells in the same layer; subsequently, the shallow aquifer head value and the deep aquifer head value within the same groundwater monitoring cycle are obtained, and the difference between the shallow aquifer head value and the deep aquifer head value is determined as the layer head difference value corresponding to the candidate well; further, multiple layer head differences are arranged in chronological order according to the groundwater monitoring cycle to obtain the head difference sequence corresponding to the candidate well. This establishes a clear data correspondence between candidate conductive wells, head difference sequences, and head difference reversal values.

[0044] Specifically, the preset cross-layer upper limit, preset failure upper limit, and preset screening threshold can be determined based on historical well data, well sealing acceptance records, on-site verification records, and manually confirmed samples of conductive wells at the target site. When manually confirmed samples are lacking in the area, the maximum value among the cross-layer values ​​of all abandoned wells at the target site can be used as the preset cross-layer upper limit, the maximum value among the well sealing failure values ​​can be used as the preset failure upper limit, and the preset screening threshold can be set to a value between zero and one. Subsequently, the preset screening threshold can be adjusted based on the on-site verification results.

[0045] Based on the head difference sequence corresponding to each candidate conductive well in the candidate well set, a head difference inversion value is generated.

[0046] It should be noted that the method for generating the head difference reversal value based on the head difference sequence corresponding to each candidate well in the candidate well set includes the following: Considering that vertical conduction of abandoned wells is usually driven by changes in the direction of the head difference between shallow and deep aquifers, judging solely based on the magnitude of the head difference within a single groundwater monitoring cycle could easily lead to misinterpreting ordinary water level fluctuations as the driving force for conduction. Therefore, it is necessary to identify whether a valid reversal of the head difference direction has occurred within continuous groundwater monitoring cycles.

[0047] In one specific embodiment, firstly, the head difference values ​​of two adjacent layers are extracted from the head difference sequence according to the groundwater monitoring cycle.

[0048] Then, the absolute values ​​of the head difference between two adjacent layers are compared with the preset effective head difference threshold. When the absolute values ​​of the head difference between two adjacent layers are both greater than or equal to the preset effective head difference threshold, and the product of the head difference between the two adjacent layers is less than zero, the next groundwater monitoring cycle is determined as the head difference reversal cycle.

[0049] Subsequently, based on all head difference reversal cycles, the number of head difference reversals is determined.

[0050] Furthermore, based on the sum of the absolute values ​​of the stratified head differences before and after each head difference reversal cycle, the single reversal amplitude value is determined, and the maximum value among all single reversal amplitude values ​​is determined as the maximum reversal amplitude value.

[0051] Finally, the number of head difference reversals and the maximum reversal amplitude are normalized to obtain normalized values ​​for the number of reversals and the maximum reversal amplitude. The smaller of these two values ​​is determined as the head difference reversal value. This head difference reversal value is used to subsequently generate wellbore conductivity values ​​in conjunction with the conductivity profile sequence. Specifically, the number of head difference reversals can be divided by a preset upper limit for the number of reversals to obtain the normalized value for the number of reversals; the maximum reversal amplitude can be divided by a preset upper limit for the maximum reversal amplitude to obtain the normalized value for the maximum reversal amplitude. Both the preset upper limit for the number of reversals and the preset upper limit for the maximum reversal amplitude are positive values ​​and can be determined based on head difference reversal records during the historical groundwater monitoring period of the target site or manually confirmed valid reversal samples.

[0052] Specifically, the preset effective head difference threshold can be determined based on the water level gauge measurement error, the natural water level fluctuation amplitude of shallow and deep aquifers during the groundwater monitoring cycle, and manually confirmed effective head difference reversal samples. This eliminates misjudgments of head difference reversal caused by instrument errors or short-term, minor water level fluctuations.

[0053] Based on the head difference reversal value and conductivity profile sequence, wellbore conduction value is generated.

[0054] It should be noted that the method for generating wellbore conduction values ​​based on head difference reversal values ​​and conductivity profile sequences includes: since head difference reversal only indicates that there is a conduction driving force between different aquifers, it cannot directly prove that vertical water exchange has occurred inside the candidate conducting well. Therefore, it is also necessary to combine the conductivity profile sequence to determine whether the water quality interface inside the wellbore moves with head difference reversal.

[0055] In one specific embodiment, firstly, based on the head difference reversal cycle, the pre-reversal profile value and post-reversal profile value are extracted from the conductivity profile sequence. Specifically, when a candidate well cannot obtain a pre-reversal profile value or post-reversal profile value due to blockage, collapse, or residual sealing, the corresponding wellbore conductivity value is not generated, and the candidate well is marked as a well to be reviewed. The well to be reviewed is not included in the calculation of the identifier switching value within the local groundwater monitoring cycle. The pre-reversal profile value is the wellbore conductivity profile value most recently collected before the head difference reversal cycle, and the post-reversal profile value is the wellbore conductivity profile value most recently collected after the head difference reversal cycle.

[0056] Then, based on the positive and negative changes in the stratified head difference before and after the head difference reversal cycle, the head difference reversal direction value is determined. The head difference reversal direction value is used to indicate whether the head driving direction is from the shallow aquifer to the deep aquifer, or from the deep aquifer to the shallow aquifer.

[0057] Subsequently, based on the conductivity difference between adjacent well depths in the pre-reversal profile, the intermediate depth between adjacent well depths with the largest conductivity difference is determined as the pre-reversal conductivity interface depth. Similarly, based on the conductivity difference between adjacent well depths in the post-reversal profile, the intermediate depth between adjacent well depths with the largest conductivity difference is determined as the post-reversal conductivity interface depth. Further, when the maximum conductivity difference between adjacent well depths in either the pre-reversal or post-reversal profile is less than a preset interface identification threshold, it is determined that there is no identifiable conductivity interface in the corresponding conductivity profile sequence. Therefore, no conductivity interface movement value is generated for the candidate well within the current groundwater monitoring cycle, and the well conduction value corresponding to the candidate well is set to zero. Specifically, the preset interface identification threshold can be determined based on conductivity measurement error, conductivity fluctuation range under static well conditions, and manually confirmed well water quality interface samples.

[0058] Furthermore, the conductivity interface movement value is determined based on the absolute value of the difference between the conductivity interface depth value before and after reversal. In one specific embodiment, the wellbore depth position starts from the wellhead and gradually increases downwards along the wellbore. When the head difference reversal direction value indicates that the head driving direction is from the shallow aquifer to the deep aquifer, if the conductivity interface depth value after reversal is greater than the conductivity interface depth value before reversal, then the conductivity interface movement direction is determined to be consistent with the head difference reversal direction value. When the head difference reversal direction value indicates that the head driving direction is from the deep aquifer to the shallow aquifer, if the conductivity interface depth value after reversal is less than the conductivity interface depth value before reversal, then the conductivity interface movement direction is determined to be consistent with the head difference reversal direction value; otherwise, the conductivity interface movement direction is determined to be inconsistent with the head difference reversal direction value.

[0059] Finally, when the direction of movement corresponding to the conductivity interface movement value is consistent with the direction of reversal of the head difference, the direction consistency value is set to one; otherwise, the direction consistency value is set to zero. Specifically, the conductivity interface movement value is normalized to obtain a normalized conductivity interface movement value; wherein, the normalized conductivity interface movement value can be determined by dividing the conductivity interface movement value by a preset upper limit value for interface movement. The preset upper limit value for interface movement is a value greater than zero, which can be determined based on the vertical distance between the shallow and deep aquifers traversed by the candidate well, or based on the larger value of the conductivity interface movement value within the historical groundwater monitoring period. When the normalized conductivity interface movement value is greater than one, the normalized conductivity interface movement value is set to one. When the direction consistency value is zero, the wellbore connection value is set to zero; when the direction consistency value is one, the smaller value between the head difference reversal value and the normalized conductivity interface movement value is determined as the wellbore connection value. The wellbore connection value is used to subsequently combine the shallow and deep concentration sequences to generate contaminated layer jump values.

[0060] Based on wellbore conductivity values ​​and shallow and deep layer concentration sequences, polluted layer jump values ​​are generated.

[0061] It should be noted that the method for generating contaminated layer jump values ​​based on wellbore conductivity and shallow / deep concentration sequences includes the following: Since deep contamination anomalies may originate from independent deep contamination sources or from shallow contaminants entering deep aquifers via candidate conductive wells, the contamination source layer cannot be determined solely based on the increase in deep contamination concentration. To distinguish between these two scenarios, it is necessary to determine whether the shallow contamination peak occurs before the deep contamination peak and whether there is a concentration response correlation between the two.

[0062] In one specific embodiment, firstly, based on the concentration sequence of shallow and deep layers, the shallow pollution peak, the shallow pollution peak time, the deep pollution peak, and the deep pollution peak time are determined.

[0063] Then, when the peak time of deep pollution is later than the peak time of shallow pollution, the time difference between the peak times of deep and shallow pollution is determined as the deep lag interval. The deep lag interval characterizes the time delay in the transmission of the shallow pollution concentration response to the deep pollution concentration response. When the peak time of deep pollution is not later than the peak time of shallow pollution, it is determined that the deep and shallow pollution concentration responses do not satisfy the time relationship of lag transmission from shallow to deep, and the lag value is set to zero.

[0064] Subsequently, when the deep hysteresis interval value is within the preset hysteresis time range, the hysteresis coincidence value is set to one; otherwise, the hysteresis coincidence value is set to zero.

[0065] Furthermore, the difference between shallow and deep pollution peak values ​​is determined based on the absolute value of the difference between the shallow and deep pollution peak values. When the difference between shallow and deep peak values ​​is less than or equal to a preset peak difference threshold, the peak difference coincidence value is set to one; otherwise, the peak difference coincidence value is set to zero.

[0066] Finally, the smaller value among the hysteresis coincidence value, peak difference coincidence value, and wellbore continuity value is determined as the contaminated layer jump value. This contaminated layer jump value is used to subsequently determine whether to switch the deep independent source identifier corresponding to the deep contamination anomaly to a continuity impact identifier.

[0067] Specifically, the preset lag time range can be determined based on the vertical distance between shallow and deep aquifers within the target site, the depth of candidate conductive wells, and historical pollution concentration response records; the preset peak difference threshold can be determined based on the historical concentration fluctuation range of the same pollutant in shallow and deep aquifers, detection errors, and dilution effects. This avoids misjudgments of pollution layer jump values ​​caused by sampling errors or single concentration fluctuations.

[0068] Before switching the label, it is necessary to identify deep pollution anomalies. Considering that deep pollution anomalies are the basis for subsequent determination of whether the label of a deep independent source needs to be switched, it cannot be determined directly based on a single concentration fluctuation.

[0069] In one specific embodiment, firstly, the pollution concentration value of the deep aquifer is extracted based on the concentration sequence of shallow and deep layers. Then, the background concentration value of the deep aquifer and a preset deep aquifer anomaly threshold are obtained. Subsequently, the pollution concentration value of the deep aquifer is compared with the background concentration value of the deep aquifer. When the difference between the two is greater than or equal to the preset deep aquifer anomaly threshold, the increased pollution concentration of the deep aquifer within the corresponding groundwater monitoring period is determined as a deep pollution anomaly. Further, the deep pollution anomaly is used to subsequently determine whether its corresponding evidence attribute remains a deep independent source identifier or switches to a conduction impact identifier.

[0070] Specifically, the background concentration value of deep aquifers can be determined based on the pollution concentration values ​​of deep aquifers during historical groundwater monitoring periods without anomalies at the target site, or based on the pollution concentration values ​​of deep aquifers from upstream background monitoring wells. The preset deep aquifer anomaly threshold can be determined based on the detection error of deep aquifer pollution concentration, the historical concentration fluctuation range, and manually confirmed samples of deep pollution anomalies. This avoids misjudging increased deep aquifer pollution concentrations as deep pollution anomalies due to detection errors or short-term concentration fluctuations.

[0071] See Figure 2 Based on the pollution layer jump value, the deep independent source identifier corresponding to the deep pollution anomaly is switched to the conduction impact identifier.

[0072] It should be noted that the method for switching the deep independent source identifier corresponding to the deep pollution anomaly to the conduction impact identifier based on the pollution layer jump value includes: determining the peak time of shallow pollution and the peak time of deep pollution based on the deep and shallow concentration sequences.

[0073] When the peak time of deep pollution is later than the peak time of shallow pollution, the deep lag interval value is determined, and the indicator switching value is determined based on the deep lag interval value, the pollution layer jump value, the wellbore conduction value, and the head difference reversal value.

[0074] When the peak time of deep pollution is not later than the peak time of shallow pollution, it is determined that the deep pollution anomaly does not satisfy the hysteresis response relationship from the shallow aquifer to the deep aquifer via the candidate conductive well. The identifier switching value is set to zero, and the deep independent source identifier corresponding to the deep pollution anomaly is retained.

[0075] When the identifier switching value is greater than or equal to the preset switching threshold, the deep independent source identifier corresponding to the deep pollution anomaly is switched to the conduction impact identifier; otherwise, the deep independent source identifier corresponding to the deep pollution anomaly is retained.

[0076] Specifically, the preset switching threshold can be determined based on manually confirmed samples of abandoned well conduction impact from historical groundwater monitoring cycles at the target site. When manually confirmed samples are lacking for the site, the preset switching threshold can be determined based on the conservative judgment intervals corresponding to the preset lower lag limit, preset upper lag limit, preset jump limit, preset conduction limit, and preset reversal limit. In one specific embodiment, since the identification switching value is determined by the minimum value among the lag coincidence value, contaminated layer jump normalization value, well conduction normalization value, and head difference reversal normalization value, the preset switching threshold can be set to a value between zero and one. When manually confirmed samples are lacking for the site, the preset switching threshold can be initially set to a conservative threshold greater than zero and less than one, and after subsequent on-site verification and confirmation of the abandoned well conduction impact, the preset switching threshold can be adjusted based on the number of erroneous switching and missed switching.

[0077] Furthermore, the method for determining the switching value includes: obtaining the preset lower hysteresis limit, the preset upper hysteresis limit, the preset jump limit, the preset conduction limit, and the preset reversal limit.

[0078] Specifically, the preset lower lag limit and the preset upper lag limit are the lower and upper limits of the preset lag time range, respectively, and are used to limit the reasonable lag between the peak time of deep pollution and the peak time of shallow pollution.

[0079] When the deep hysteresis interval value is greater than or equal to the preset lower hysteresis limit and less than or equal to the preset upper hysteresis limit, the hysteresis coincidence value is set to one; otherwise, the hysteresis coincidence value is set to zero.

[0080] Divide the pollution level jump value by the preset jump upper limit value to obtain the pollution level jump normalization value.

[0081] Divide the wellbore conduction value by the preset conduction upper limit value to obtain the wellbore conduction normalization value.

[0082] Divide the head difference reversal value by the preset reversal upper limit value to obtain the head difference reversal normalization value.

[0083] The minimum value among the hysteresis coincidence value, the pollution layer jump normalization value, the wellbore conduction normalization value, and the head difference reversal normalization value is determined as the identifier switching value.

[0084] Specifically, the preset jump upper limit, preset conduction upper limit, and preset reversal upper limit are all values ​​greater than zero. The preset jump upper limit can be determined based on the largest jump value of the contaminated stratum within the historical groundwater monitoring period of the target site, or a manually confirmed sample of contaminated stratum jumps. The preset conduction upper limit can be determined based on the largest conduction value of the candidate conduction wells within the target site, or a manually confirmed sample of conduction wells. The preset reversal upper limit can be determined based on the largest reversal value of the head difference in the head difference sequence, or a manually confirmed valid sample of head difference reversal. When manually confirmed samples are lacking for the site, the maximum value from the corresponding historical monitoring values ​​can be used as the preset jump upper limit, preset conduction upper limit, and preset reversal upper limit, respectively. When the contaminated stratum jump normalization value is greater than one, the contaminated stratum jump normalization value is set to one; when the well conduction normalization value is greater than one, the well conduction normalization value is set to one; when the head difference reversal normalization value is greater than one, the head difference reversal normalization value is set to one.

[0085] By determining the minimum value among the hysteresis coincidence value, the polluted layer jump normalization value, the wellbore conduction normalization value, and the head difference reversal normalization value as the identifier switching value, the identifier switching can be constrained by deep hysteresis, polluted layer jump, wellbore conduction, and head difference reversal simultaneously, thus avoiding false switching caused by a single abnormal pollutant concentration or a single head fluctuation.

[0086] In one specific embodiment, the candidate pollution source area can be determined based on the target site's historical production data, pollutant storage areas, leakage records, site spatial information, and abnormal pollution distribution from monitoring wells. Specifically, spatial areas with records of pollutant storage, use, discharge, or historical leakage are used as initial candidate pollution sources. These initial candidate pollution sources are then screened based on abnormal pollution distribution from monitoring wells to obtain a ranking of candidate pollution sources.

[0087] Before revising the candidate region ranking values, it is necessary to determine the initial ranking values ​​for each pollution source candidate region. Similarly, before revising the pollution source stratigraphic values ​​and pollution migration path values, it is necessary to determine the corresponding initial pollution source stratigraphic values ​​and initial pollution migration path values. In one specific embodiment, firstly, the initial pollution source stratigraphic values ​​are determined based on the aquifers where pollution anomalies are located in the shallow and deep concentration sequences. Then, the initial pollution migration path values ​​are determined based on the spatial relationship between the pollution source candidate regions and monitoring wells, the flow direction in shallow aquifers, and the flow direction in deep aquifers. Subsequently, the initial pollution source stratigraphic values ​​are used as the pollution source stratigraphic values, and the initial pollution migration path values ​​are used as the pollution migration path values ​​for subsequent correction when the conduction impact indicator is established. Considering that the ranking of pollution source candidate regions is usually influenced by the degree of pollution concentration anomaly, the time of pollution peak occurrence, and the spatial distance between the pollution source candidate regions and monitoring wells, it is necessary to first generate an initial ranking basis that can be corrected.

[0088] In one specific embodiment, firstly, the degree of pollution concentration anomaly corresponding to each monitoring well is determined based on the concentration sequence of deep and shallow layers.

[0089] Specifically, the degree of pollution concentration anomaly can be determined based on the difference between the pollution concentration value and the corresponding background concentration value of the aquifer. When there are multiple pollutant indicators, the concentration anomaly value corresponding to each pollutant indicator can be determined separately, and the maximum value among the concentration anomalies can be taken as the degree of pollution concentration anomaly.

[0090] Then, based on the peak times of shallow and deep pollution, the time of pollution anomaly occurrence is determined.

[0091] Subsequently, spatial proximity values ​​are determined based on the spatial distance between the candidate pollution source area and the corresponding monitoring well.

[0092] Furthermore, based on the degree of pollution concentration anomaly, the time of occurrence of the pollution anomaly, and spatial proximity, the initial ranking value of each pollution source candidate area is determined. Among them, the higher the degree of pollution concentration anomaly, the earlier the pollution anomaly occurred, and the closer the spatial distance, the higher the initial ranking value of the corresponding pollution source candidate area.

[0093] Specifically, in one embodiment, the degree of pollution concentration anomaly, the time of occurrence of the pollution anomaly, and the spatial proximity value are normalized to obtain a normalized value for concentration anomaly, a normalized value for time, and a normalized value for spatial proximity. Then, the average of these three values ​​is determined as the initial ranking value for the corresponding pollution source candidate area. The earlier the pollution anomaly occurs, the larger the time normalized value; the smaller the spatial distance between the pollution source candidate area and the corresponding monitoring well, the larger the spatial proximity normalized value. The time normalized value can be determined based on the chronological order of pollution anomalies within the same groundwater monitoring cycle, with the pollution source candidate area having the earliest occurrence time corresponding to the largest time normalized value. The spatial proximity normalized value can be determined based on the spatial distance between the pollution source candidate area and the corresponding monitoring well, with the pollution source candidate area having the smallest spatial distance corresponding to the largest spatial proximity normalized value. Thus, the initial ranking value can simultaneously reflect the intensity of the pollution anomaly, the chronological order of anomalies, and the degree of spatial proximity.

[0094] Finally, the initial ranking value of each pollution source candidate area is used as the candidate area ranking value, which will be used for subsequent correction when the conduction impact indicator is established.

[0095] Based on the conduction impact indicator, the candidate area ranking value, pollution source layer value, and pollution migration path value are corrected to obtain the pollution source location value.

[0096] It should be noted that the method for obtaining the pollution source location value by correcting the candidate area ranking value, pollution source stratigraphic value, and pollution migration path value based on the conduction impact indicator includes the following: Considering that when the conduction impact indicator is valid, deep pollution anomalies are no longer suitable as direct evidence of deep independent pollution sources, if the deep candidate area ranking value is still increased according to the conventional source tracing method, it is easy to lead to errors in the pollution source stratigraphic position and the depth of remediation site deployment. Therefore, it is necessary to directly transfer the conduction impact indicator to the correction process of the pollution source candidate area ranking, pollution source stratigraphic position, and pollution migration path.

[0097] In one specific embodiment, firstly, based on the conduction impact identifier, the corresponding candidate conduction well is determined as the target conduction well; wherein, the corresponding candidate conduction well refers to the candidate conduction well that participates in generating the wellbore conduction value, the contaminated layer jump value and the identifier switching value, and triggers the conduction impact identifier.

[0098] Specifically, when there are multiple candidate wells with corresponding conduction impact indicators within the same groundwater monitoring cycle, each candidate well is identified as the target well, and the shallow correction candidate area determination, sorting correction, and conduction migration path value correction processes are executed respectively.

[0099] Then, based on the spatial distance between the target well and each pollution source candidate area and the direction of water flow in the shallow aquifer, shallow modified candidate areas are determined from the pollution source candidate areas upstream of the target well.

[0100] Specifically, the flow direction in shallow aquifers can be determined based on shallow aquifer contour lines or shallow aquifer head gradients. For each pollution source candidate area, firstly, the candidate migration direction from the pollution source candidate area to the target well is determined; then, the candidate migration direction is compared with the shallow aquifer flow direction; subsequently, when the angle between the candidate migration direction and the shallow aquifer flow direction is less than or equal to a preset upstream angle threshold, and the spatial distance between the pollution source candidate area and the target well is less than or equal to a preset shallow correction distance threshold, the pollution source candidate area is determined as a shallow correction candidate area; when the above conditions are not met, the pollution source candidate area is not determined as a shallow correction candidate area. This ensures that the shallow correction candidate area is simultaneously constrained by the shallow aquifer flow direction and spatial distance, preventing the incorrect ranking of pollution source candidate areas downstream of the target well or too far away. The preset upstream angle threshold can be determined based on the historical fluctuation range of the shallow aquifer flow direction within the target site. The preset shallow correction distance threshold can be determined based on the historical effective response distance between the pollution source candidate area and the monitoring well within the target site, the permeability of the shallow aquifer, and the pollutant migration capacity.

[0101] In one specific embodiment, the preset upstream angle threshold is an angle value greater than zero and less than or equal to 90 degrees. When the historical fluctuation range of the shallow aquifer flow direction is large, the preset upstream angle threshold is increased accordingly; when the shallow aquifer flow direction is stable, the preset upstream angle threshold is decreased accordingly. The preset shallow correction distance threshold is a distance value greater than zero, which can be determined based on the effective migration distance between the confirmed pollution source candidate area and the response monitoring well within the target site. When the plot lacks confirmed samples, the larger value of the historical effective response distance between the pollution source candidate area and the monitoring well within the target site can be selected based on the permeability of the shallow aquifer and the pollutant migration capacity.

[0102] Subsequently, a preset sorting correction value is obtained. The sorting value of the candidate area corresponding to the deep pollution anomaly is reduced by the preset sorting correction value, and the sorting value of the candidate area corresponding to the shallow correction candidate area is increased by the preset sorting correction value to obtain the corrected candidate area sorting value. When the corrected candidate area sorting value exceeds the preset sorting value range, the corrected candidate area sorting value is restricted to the preset sorting value range.

[0103] Specifically, the preset ranking correction value can be determined based on the identifier switching value. In one specific embodiment, firstly, a preset maximum ranking correction value is obtained; then, the product of the identifier switching value and the preset maximum ranking correction value is determined as the preset ranking correction value; subsequently, the preset ranking correction value is used to reduce the ranking value of the candidate area corresponding to the deep pollution anomaly and to increase the ranking value of the candidate area corresponding to the shallow correction candidate area. The preset maximum ranking correction value can be determined based on the value range of the candidate area ranking value and the manually reviewed abandoned well conduction impact sample. The preset ranking value range is consistent with the value range of the initial ranking value to prevent the corrected candidate area ranking value from exceeding the comparable range.

[0104] Furthermore, the pollution source stratum value is corrected from the deep aquifer to the shallow aquifer corresponding to the shallow correction candidate area, thus obtaining the corrected pollution source stratum value.

[0105] Finally, the pollution migration path value is corrected to the conduction migration path value from the shallow corrected candidate area to the deep aquifer through the target conduction well, and the pollution source location value is determined based on the corrected candidate area ranking value, the corrected pollution source layer value, and the conduction migration path value.

[0106] Specifically, the location corresponding to the shallow modified candidate region with the highest modified candidate region ranking value is determined as the pollution source location value. When there are multiple shallow modified candidate regions whose modified candidate region ranking values ​​are all greater than or equal to a preset output threshold, multiple pollution source location values ​​are output in descending order of modified candidate region ranking values, and associated with the corresponding modified pollution source layer value and conduction migration path value.

[0107] The preset output threshold can be determined based on the range of the initial sorting values, the number of candidate pollution source samples confirmed manually, and the number of candidate locations to be output for the target site. When there are few manually confirmed samples for a plot, the preset output threshold can be set to a value between zero and one, and adjusted according to the on-site verification results.

[0108] Furthermore, in order to avoid the system remaining on for a long time and affecting the indicator due to a single head difference reversal or a single pollution peak lag, an indicator recovery process is also set up in one specific embodiment.

[0109] First, when switching the deep independent source identifier corresponding to the deep pollution anomaly to the conduction impact identifier, save the sorting value, the layer value, and the migration path value before the switch.

[0110] Then, during subsequent groundwater monitoring cycles, the label switching values ​​are redefined.

[0111] Subsequently, when the identifier switching value is continuously less than the preset switching threshold for a preset recovery cycle number, it is determined that the conduction effect of the abandoned well is no longer continuous, and the conduction effect identifier corresponding to the deep pollution anomaly is restored to the deep independent source identifier.

[0112] Specifically, the preset number of recovery cycles is a positive integer greater than or equal to two, which can be determined based on the length of the groundwater monitoring cycle and the cycle of change in the hydraulic head difference at the target site. When the groundwater monitoring cycle is short or the hydraulic head difference changes rapidly, the preset number of recovery cycles can be appropriately increased to avoid frequent restoration of markers due to short-term fluctuations; when the groundwater monitoring cycle is long and the hydraulic head difference changes stably, the preset number of recovery cycles can be appropriately reduced.

[0113] Furthermore, the candidate area ranking value, pollution source stratigraphic value, and pollution migration path value are restored to the ranking value, stratigraphic value, and migration path value before the switchover, respectively.

[0114] Finally, in the next groundwater monitoring cycle after the deep independent source identification is restored, the candidate area ranking value will continue to be updated based on the deep and shallow concentration sequences, and the pollution source location value will continue to be determined based on the candidate area ranking value, pollution source stratigraphic value, and pollution migration path value.

[0115] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the groundwater pollution source tracing method and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for tracing and tracking groundwater pollution sources based on multimodal data fusion, characterized in that, Includes the following steps: Acquire the abandoned well penetration values, well sealing failure values, hydraulic head difference sequences, electrical conductivity profile sequences, and shallow and deep concentration sequences for the target site in each groundwater monitoring cycle; Based on the abandoned well penetration value, well sealing failure value and head difference sequence, determine whether there is a risk of conduction jump at the target site. If so, screen the candidate well set based on the abandoned well penetration value and well sealing failure value. Based on the head difference sequence corresponding to each candidate conductive well in the candidate well set, generate the head difference inversion value; Based on the head difference reversal value and conductivity profile sequence, the wellbore conduction value is generated; Based on wellbore conductivity values ​​and shallow and deep layer concentration sequences, polluted layer jump values ​​are generated. Deep pollution anomalies are identified based on shallow and deep concentration sequences, and deep independent source identifiers corresponding to deep pollution anomalies are switched to conduction impact identifiers based on pollution layer jump values. Based on the conduction impact indicator, the candidate area ranking value, pollution source layer value, and pollution migration path value are corrected to obtain the pollution source location value; Methods for generating head difference inversion values ​​include: Extract the head difference between two adjacent layers from the head difference sequence; When the absolute value of the difference between two adjacent layers of water head is greater than or equal to the preset effective water head difference threshold, and the product of the difference between two adjacent layers of water head difference is less than zero, the next groundwater monitoring cycle is determined as the water head difference reversal cycle. The maximum reversal amplitude value is determined based on the number of head difference reversals and the stratified head difference value before and after the head difference reversal cycle, and the head difference reversal value is determined accordingly. Methods for generating wellbore continuity values ​​include: Based on the head difference reversal cycle, the pre-reversal profile value and post-reversal profile value are extracted from the conductivity profile sequence. The direction of head difference reversal is determined based on the positive and negative changes in the stratified head difference before and after the head difference reversal cycle. Based on the conductivity difference between adjacent wellbore depths in the profile values ​​before and after reversal, the conductivity interface depth values ​​before and after reversal are determined. The conductivity interface shift value is determined based on the conductivity interface depth value before and after reversal. The wellbore conduction value is determined based on the conductivity interface shift value, the head difference reversal direction value, and the head difference reversal value.

2. The groundwater pollution source tracing method based on multimodal data fusion according to claim 1, characterized in that, Methods for switching the deep independent source identifier corresponding to a deep contamination anomaly to a conduction impact identifier include: Based on the concentration sequences of shallow and deep layers, the peak times of shallow and deep pollution were determined. When the peak time of deep pollution is later than the peak time of shallow pollution, the deep lag interval value is determined, and the indicator switching value is determined based on the deep lag interval value, the pollution layer jump value, the wellbore conduction value and the head difference reversal value. When the identifier switching value is greater than or equal to the preset switching threshold, the deep independent source identifier corresponding to the deep pollution anomaly is switched to the conduction impact identifier; otherwise, the deep independent source identifier corresponding to the deep pollution anomaly is retained.

3. The groundwater pollution source tracing method based on multimodal data fusion according to claim 2, characterized in that, The methods for determining the identifier switching value include: Obtain the preset hysteresis lower limit, preset hysteresis upper limit, preset jump upper limit, preset conduction upper limit, and preset reversal upper limit; When the deep hysteresis interval value is between the preset lower hysteresis limit and the preset upper hysteresis limit, the hysteresis coincidence value is set to one; otherwise, it is set to zero. Based on preset jump upper limit, preset conduction upper limit, and preset reversal upper limit, the jump value of the contaminated layer, the conduction value of the well shaft, and the reversal value of the head difference are normalized to obtain the normalized value of the jump value of the contaminated layer, the normalized value of the conduction value of the well shaft, and the normalized value of the reversal value of the head difference. The minimum value among the hysteresis coincidence value, the pollution layer jump normalization value, the wellbore conduction normalization value, and the head difference reversal normalization value is determined as the identifier switching value.

4. The groundwater pollution source tracing method based on multimodal data fusion according to claim 1, characterized in that, Methods for determining whether a target site has a continuity switching risk include: The structural risk value of each abandoned well is determined based on its cross-layer value and sealing failure value. The periodic variation value of the head difference is determined based on the number of changes in the direction of the head difference in the head difference sequence and the maximum absolute value in the head difference sequence. The conduction risk value is determined based on the structural risk value of abandoned wells and the periodic variation value of hydraulic head difference. When the conduction risk value corresponding to any abandoned well is greater than or equal to the preset conduction risk threshold, it is determined that there is a conduction jump risk at the target site.

5. The groundwater pollution source tracing method based on multimodal data fusion according to claim 1, characterized in that, Methods for selecting candidate well sets include: Obtain the preset layer penetration limit value, preset failure limit value, and preset filtering threshold; Based on preset upper limits for penetration and preset upper limits for failure, the penetration value and well sealing failure value of abandoned wells are normalized to obtain normalized penetration value and normalized failure value. The smaller value between the cross-layer normalization value and the failure normalization value of the same abandoned well is determined as the single-well conduction screening value. Abandoned wells with a single-well conduction screening value greater than or equal to a preset screening threshold are identified as candidate conduction wells, and a candidate well set is formed based on all candidate conduction wells.

6. The groundwater pollution source tracing method based on multimodal data fusion according to claim 1, characterized in that, Methods for generating contamination stratum jump values ​​include: The peak value of shallow pollution, the time of the peak value of shallow pollution, the peak value of deep pollution, and the time of the peak value of deep pollution were determined based on the concentration sequences of shallow and deep layers. When the peak time of deep pollution is later than the peak time of shallow pollution, the deep lag interval value is determined. The difference between shallow and deep pollution peak values ​​is determined based on the absolute value of the difference between the shallow and deep pollution peak values. The contaminated stratum jump value is determined based on the deep hysteresis interval, the difference between deep and shallow peak values, and the wellbore conductivity.

7. The groundwater pollution source tracing method based on multimodal data fusion according to claim 1, characterized in that, Methods for obtaining pollution source location values ​​include: The candidate wells that trigger the conduction impact indicator are identified as the target wells; The shallow modified candidate area is determined based on the spatial distance between the target well and the pollution source candidate area and the water flow direction of the shallow aquifer. The candidate region ranking value is reduced based on the preset ranking correction value, and the candidate region ranking value is increased based on the shallow corrected candidate region, thus obtaining the corrected candidate region ranking value. The pollution source stratigraphic value is corrected to the shallow aquifer corresponding to the shallow correction candidate area to obtain the corrected pollution source stratigraphic value; the pollution migration path value is corrected to the conduction migration path value from the shallow correction candidate area to the deep aquifer through the target well. The location value of the pollution source is determined based on the corrected candidate region ranking value, the corrected pollution source stratigraphic value, and the conductive migration path value.

8. The groundwater pollution source tracing method based on multimodal data fusion according to claim 2, characterized in that, Also includes: When switching to the conduction impact indicator, save the sorting value, layer value, and migration path value before the switch; The label switching value will be re-determined during subsequent groundwater monitoring cycles; When the identifier switching value is continuously less than the preset switching threshold for a preset number of recovery cycles, the conduction impact identifier will be restored to the deep independent source identifier, and the sorting value, layer position value, and migration path value before switching will be restored.

Images