An underground water sampling intelligent monitoring method based on big data collection

Abstract

The application discloses a kind of groundwater sampling intelligent monitoring methods based on big data collection, comprising the following steps: constructing the spatial grid unit set of groundwater monitoring area and monitoring well point set, and collecting groundwater monitoring dataset;Groundwater propagation network model is constructed;Abnormal detection is carried out, and the spatial grid unit corresponding to abnormal monitoring well point is determined;Get candidate propagation path set;Determine target reverse propagation path;Generate path sampling decision parameter set;Determine the sampling sequence and sampling time interval on target reverse propagation path by constructing time-constrained traveling salesman problem, generate path-oriented sampling scheme;Directional sampling is executed, and path sampling data is obtained and groundwater monitoring dataset is updated;Determine potential pollution source area and output corresponding monitoring result, improve the pertinence and sampling efficiency of groundwater abnormal propagation tracking.

Description

Technical Field

[0001] This invention relates to the field of groundwater monitoring and environmental information processing, and in particular to an intelligent monitoring method for groundwater sampling based on big data acquisition. Background Technology

[0002] Groundwater, as an important type of water resource, is widely used in various scenarios such as residential water supply, agricultural irrigation, and industrial production. Its water quality and propagation processes directly affect regional ecological environment security and water resource utilization security. Due to the characteristics of groundwater systems—high concealment, complex flow paths, delayed propagation processes, and the combined influence of geological structure, hydraulic connections, and local disturbances—it is often difficult to promptly and accurately identify the direction of propagation, the extent of diffusion, and potential pollution source areas when abnormal components or pollutant migration occur in groundwater. Therefore, establishing effective monitoring, tracking, and targeted sampling mechanisms for abnormal groundwater conditions has become a crucial technical problem that needs to be solved in the field of intelligent groundwater monitoring.

[0003] In existing technologies, groundwater monitoring is typically based on the deployment of monitoring wells, and groundwater quality parameter data are obtained through fixed-period sampling, manual inspection, and laboratory analysis. While this approach can reflect water quality changes at different monitoring wells over time to some extent, it is mostly characterized by single-well-point, discrete, and static data collection. Sampling tasks are usually carried out according to administrative divisions, well-point distribution conditions, or experience-based inspection routes, lacking the ability to dynamically analyze the internal propagation relationships of groundwater. In other words, existing technologies rely heavily on manual experience or established monitoring procedures at the sampling execution level, making it difficult to adaptively adjust the sampling targets, sampling sequence, and sampling time based on the characteristics of abnormal groundwater propagation, thus limiting monitoring efficiency and anomaly tracing capabilities. Summary of the Invention

[0004] One objective of this invention is to propose an intelligent monitoring method for groundwater sampling based on big data acquisition. This invention fully utilizes groundwater monitoring data analysis, spatial grid modeling, groundwater propagation network modeling, and path optimization calculation methods to model and reverse-track the abnormal propagation process of groundwater. Based on this, a time-constrained sampling path planning mechanism is constructed to form a path-oriented directional sampling scheme, achieving efficient identification of potential groundwater pollution source areas. This invention can adaptively determine the sampling object, sampling order, and sampling time based on the groundwater anomaly monitoring results, possessing advantages such as strong sampling targeting, reasonable path planning, high monitoring efficiency, and high accuracy in pollution source location.

[0005] A groundwater sampling intelligent monitoring method based on big data acquisition according to an embodiment of the present invention includes the following steps:

[0006] Construct a spatial grid cell set and a monitoring well set for the groundwater monitoring area, determine the spatial grid cell to which each monitoring well belongs, and collect groundwater monitoring datasets;

[0007] Based on groundwater monitoring datasets, the groundwater flow relationships between spatial grid cells are modeled to construct a groundwater propagation network model.

[0008] Based on the groundwater monitoring dataset, anomaly detection is performed on the groundwater quality parameters of the monitoring well points to identify abnormal monitoring well points and obtain the spatial grid cells corresponding to the abnormal monitoring well points.

[0009] Using the spatial grid cells corresponding to the anomaly monitoring well points as the starting nodes, and based on the groundwater propagation network model, a reverse propagation path search is performed to obtain a set of candidate propagation paths;

[0010] The candidate propagation path set is screened, and the target reverse propagation path is determined based on path connectivity and flow consistency;

[0011] Along the reverse propagation path of the target, the path average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell are extracted to generate a path sampling decision parameter set;

[0012] Based on the path sampling decision parameter set, a time-constrained traveling salesman problem is constructed to determine the sampling order and sampling time interval on the target back propagation path, and a path-guided sampling scheme is generated.

[0013] Perform directional sampling according to the path-guided sampling scheme, obtain path sampling data and update the groundwater monitoring dataset;

[0014] Based on the updated groundwater monitoring dataset, the groundwater propagation network model is updated to identify potential pollution source areas and output corresponding monitoring results.

[0015] Optionally, the boundary data of the groundwater monitoring area is obtained based on the spatial range of the groundwater monitoring area, and the groundwater monitoring area is divided into grids according to a unified spatial division scale to obtain spatial grid units with spatial location coordinates. The spatial coordinate information of each monitoring well point in the monitoring area is obtained, and the spatial grid unit to which each monitoring well point belongs is determined based on the positional relationship between the spatial coordinates of each monitoring well point and the spatial range of the spatial grid unit. Through the sensing devices and data acquisition terminals deployed at each monitoring well point, groundwater quality parameters, groundwater level, groundwater flow direction, groundwater flow velocity, formation permeability parameters, and historical monitoring records are collected. The collected data are processed with unified time stamping, and the data from different sources are linked and integrated in chronological order to form a groundwater monitoring dataset.

[0016] Optionally, the construction of the groundwater propagation network model specifically includes:

[0017] Based on the groundwater level in the groundwater monitoring data, calculate the corresponding groundwater hydraulic gradient between two adjacent spatial grid cells;

[0018] Calculate the groundwater flow intensity between adjacent spatial grid cells based on the formation permeability parameters in the groundwater monitoring data.

[0019] Based on the groundwater hydraulic gradient between adjacent spatial grid cells and the groundwater flow direction data in the groundwater monitoring dataset, the groundwater flow direction is determined. When the groundwater level of a spatial grid cell is greater than that of an adjacent spatial grid cell, a groundwater flow relationship is established from the spatial grid cell to the adjacent spatial grid cell. When the groundwater level of a spatial grid cell is less than that of an adjacent spatial grid cell, a reverse groundwater flow relationship is established.

[0020] By using groundwater flow intensity as the weight value of groundwater flow relationship and groundwater flow direction as the directional attribute of groundwater flow relationship, an association matrix between spatial grid cells is constructed.

[0021] Using spatial grid cells as nodes, groundwater flow relationships as connecting edges between nodes, and an association matrix as the network structure representation, a groundwater propagation network model is generated, with spatial grid cells as nodes and groundwater flow direction and intensity as edge attributes.

[0022] Optionally, obtaining the spatial grid cells corresponding to the anomaly monitoring well points specifically includes:

[0023] The groundwater quality parameters of each monitoring well point at each monitoring time are extracted from the groundwater monitoring dataset, and the various water quality parameters of each monitoring well point are classified and stored to form a water quality parameter data sequence divided by monitoring well point.

[0024] For each monitoring well point, the average level and fluctuation range of each water quality parameter at historical monitoring times are statistically analyzed.

[0025] The water quality parameters at the current monitoring time are compared with the average levels at the corresponding historical monitoring times to determine the degree of deviation of each water quality parameter from the historical state.

[0026] The deviation of various water quality parameters at the same monitoring well point is uniformly quantified to obtain the overall anomaly level of the monitoring well point at the current monitoring time. The monitoring well points with an overall anomaly level greater than the preset anomaly threshold are identified as abnormal monitoring well points. Based on the spatial grid unit to which each monitoring well point belongs, the spatial grid unit corresponding to the abnormal monitoring well point is obtained.

[0027] Optionally, obtaining the candidate propagation path set specifically includes:

[0028] The spatial grid cell corresponding to the abnormal monitoring well point is determined as the starting node. In the groundwater propagation network model, the adjacent spatial grid cells connected to the starting node are extracted, and the corresponding groundwater flow relationship is obtained.

[0029] Based on the direction of groundwater flow, the groundwater flow relationships in the groundwater propagation network model are filtered to form a set of connection relationships for reverse propagation path search;

[0030] Starting from the initial node, the path is expanded on the set of connection relationships. The recursive search is performed upstream level by level according to the connection relationship between spatial grid units to generate a path sequence formed by the sequential connection of spatial grid units.

[0031] During the path expansion process, the groundwater flow intensity between adjacent spatial grid cells traversed by each path is accumulated to obtain the cumulative flow intensity corresponding to the path.

[0032] During path expansion, when the current path can no longer continue to expand upstream along the set of connections, the path expansion stops, and the path is identified as a candidate propagation path. All paths that have completed path expansion are then aggregated to form a set of candidate propagation paths.

[0033] Optionally, determining the target reverse propagation path specifically includes:

[0034] For each candidate propagation path in the candidate propagation path set, the path connectivity is calculated to obtain the path connectivity index;

[0035] Extract the groundwater flow direction between adjacent spatial grid cells in each candidate propagation path, and determine whether the groundwater flow direction between adjacent spatial grid cells is consistent with the reverse propagation path direction;

[0036] When the groundwater flow direction is consistent between adjacent spatial grid cells, the flow consistency index is calculated.

[0037] The groundwater flow intensity between all adjacent spatial grid cells in the candidate propagation path is accumulated and averaged with the number of connections between adjacent spatial grid cells in the candidate propagation path to obtain the path intensity index.

[0038] The path connectivity index, flow consistency index, and path strength index of each candidate propagation path are weighted and summed to obtain the comprehensive path evaluation value.

[0039] The comprehensive evaluation values ​​of each path in the candidate propagation path set are compared, and the candidate propagation path with the largest comprehensive evaluation value is selected as the target reverse propagation path.

[0040] Optionally, the generation of the path sampling decision parameter set specifically includes:

[0041] For each adjacent spatial grid cell in the reverse propagation path of the target, the corresponding groundwater flow intensity is extracted to form a path flow intensity sequence;

[0042] Based on the path flow intensity sequence, the flow intensity of all groundwater in the path is accumulated, and the average flow intensity parameter of the path is determined by combining the number of connections between adjacent spatial grid cells in the target reverse propagation path.

[0043] For each spatial grid cell in the path flow intensity sequence, the corresponding historical anomaly distribution features are extracted from the groundwater monitoring dataset to form a path anomaly distribution sequence.

[0044] Based on the path anomaly distribution sequence, all historical anomaly distribution features in the path are accumulated and combined with the number of connections between adjacent spatial grid cells in the target back propagation path to determine the path anomaly distribution parameters.

[0045] The groundwater flow intensity corresponding to each spatial grid cell is compared with the path average flow intensity parameter, and the corresponding historical anomaly distribution characteristics are compared with the path anomaly distribution parameters. The path sampling priority parameters of each spatial grid cell are obtained by comprehensive calculation.

[0046] The path sampling priority parameters corresponding to each spatial grid cell are associated with the spatial grid cell sequence information in the target backpropagation path to form a path sampling decision parameter set.

[0047] Optionally, the generation of the path-guided sampling scheme specifically includes:

[0048] Read the path sampling decision parameter set, and obtain the sequence information, path average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell in the target reverse propagation path, and treat each spatial grid cell as a node to be visited;

[0049] Based on the sequential relationship of the nodes to be visited, the path distances between the nodes to be visited along the reverse propagation path of the target are determined, forming a set of path distances;

[0050] Based on the path average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell in the path sampling decision parameter set, the priority of each node to be visited is sorted to obtain the access priority sequence of the nodes to be visited.

[0051] Based on the access priority sequence, the latest allowed access time is assigned to each node to be accessed, thus forming a time constraint condition.

[0052] Based on the nodes to be visited, the set of path distances, and the time constraints, a time-constrained traveling salesman problem is constructed, with the optimization objective being to visit all nodes to be visited while minimizing the path distance.

[0053] Solve the time-constrained traveling salesman problem to obtain the sampling order of each spatial grid cell;

[0054] Based on the sampling order, the sampling time interval between adjacent spatial grid cells is determined, and a path-guided sampling scheme is generated.

[0055] Optionally, updating the groundwater monitoring dataset specifically includes:

[0056] Read the sampling order, the sampling time corresponding to each spatial grid unit, and the sampling time interval between adjacent spatial grid units in the path-guided sampling scheme to determine the on-site sampling execution order of each spatial grid unit;

[0057] According to the on-site sampling execution order, the sampling positions corresponding to each spatial grid unit are reached in sequence, groundwater is sampled at the corresponding sampling time, and the groundwater quality parameters corresponding to each spatial grid unit are detected to obtain the path sampling data corresponding to each spatial grid unit.

[0058] The path sampling data corresponding to each spatial grid unit is associated with the corresponding spatial grid unit identifier, sampling time and sampling order to form the unit sampling record corresponding to each spatial grid unit;

[0059] The sampling records of each unit are sequentially collected according to the sampling order to form a path sampling dataset covering the target's backward propagation path;

[0060] The path sampling dataset and the groundwater monitoring dataset are merged to obtain the updated groundwater monitoring dataset.

[0061] Optionally, the updated groundwater monitoring dataset is input into the groundwater propagation network model to recalibrate the groundwater flow relationship between spatial grid cells and re-determine the anomalies of groundwater quality parameters corresponding to each spatial grid cell. Based on the updated distribution of anomaly monitoring data, the spatial grid cells corresponding to each anomaly monitoring well point, the target reverse propagation path, and the groundwater flow relationship between each spatial grid cell, the anomaly source location is located step by step along the reverse direction of groundwater propagation. The range of spatial grid cells upstream of the anomaly propagation result is determined. Spatial grid cells with high anomaly degree, strong propagation correlation, and located at the forefront of the reverse propagation path are identified as potential pollution source areas. The spatial grid cell identifiers corresponding to the potential pollution source areas, the corresponding anomaly results of groundwater quality parameters, the target reverse propagation path, and the corresponding monitoring results in the updated groundwater monitoring dataset are output.

[0062] The beneficial effects of this invention are:

[0063] This invention, based on a groundwater monitoring dataset, first constructs a spatial grid cell set and a groundwater propagation network model. Then, it performs a reverse propagation path search on the spatial grid cells corresponding to anomaly monitoring well points and selects the target reverse propagation path from the candidate propagation path set. This unifies the association between groundwater anomalies and their relationships with groundwater flow, spatial location, and propagation paths. This method allows for the direct inference of upstream spatial grid cells related to anomaly propagation from anomaly results, avoiding the problems of large sampling range, unclear investigation direction, and scattered sampling resources in traditional techniques, thus improving the directionality and effectiveness of groundwater anomaly tracking.

[0064] This invention extracts the path-average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell along the target's reverse propagation path, generating a path sampling decision parameter set. Based on this, a time-constrained traveling salesman problem is constructed to determine the sampling order and sampling time interval along the target's reverse propagation path, generating a path-oriented sampling scheme. In other words, this invention does not merely determine "which locations to sample," but further determines "the sampling order" and "the arrival time at the corresponding spatial grid cell for sampling." Therefore, this invention can collaboratively incorporate groundwater propagation characteristics, anomaly historical information, and on-site sampling execution requirements into the path planning process, making the sampling process no longer a simple spatial point traversal, but forming a directional sampling mechanism that considers the correlation of propagation paths, the timeliness of on-site execution, and the goal of minimizing the path length. Compared with existing technologies, this method can reduce invalid round trips and repeated sampling, shorten the sampling path length, improve the efficiency of on-site sampling operations, and enhance the continuous capture capability of anomaly propagation chains.

[0065] After completing the path-guided sampling scheme, this invention performs directional sampling according to the sampling sequence to obtain path sampling data. This path sampling data is then associated with spatial grid cell identifiers, sampling time, and sampling sequence to form cell sampling records. These records are further used to form a path sampling dataset, which is then merged with the original groundwater monitoring dataset to obtain an updated groundwater monitoring dataset. Therefore, this invention not only enables supplementary sampling of key areas under abnormal conditions but also achieves structured integration of path sampling data with existing monitoring data, allowing new sampling results to be directly fed back into the groundwater propagation network model update process. This method effectively overcomes the shortcomings of existing technologies, such as "lagging monitoring data updates" and "difficulty in coordinating supplementary sampling results with the original monitoring system," creating a closed loop between monitoring, tracking, sampling, and model updating, thereby improving the dynamic update capability and continuous analysis capability of the groundwater monitoring system.

[0066] This invention updates the groundwater propagation network model based on an updated groundwater monitoring dataset, identifies potential pollution source areas, and outputs corresponding monitoring results. This ensures that the identification of potential pollution source areas is not based on a single anomaly judgment or isolated sampling results, but rather on the combined effects of anomaly detection results, reverse propagation paths, path-guided sampling results, and data update results. Compared to existing technologies that rely solely on empirical judgments based on local monitoring wellpoint anomalies, this invention improves the accuracy, reliability, and specificity of potential pollution source area identification, reduces false positives and false negatives, and has higher practical application value for subsequent pollution investigation, pollution control, and optimization of groundwater monitoring schemes. Attached Figure Description

[0067] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0068] Figure 1 This is an overall flowchart of an intelligent monitoring method for groundwater sampling based on big data acquisition proposed in this invention.

[0069] Figure 2 This is a schematic diagram illustrating the construction of a groundwater propagation network model for a groundwater sampling intelligent monitoring method based on big data acquisition proposed in this invention.

[0070] Figure 3 This is a schematic diagram illustrating the construction of a path-guided sampling scheme for a groundwater sampling intelligent monitoring method based on big data acquisition proposed in this invention. Detailed Implementation

[0071] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0072] refer to Figures 1-3 A groundwater sampling intelligent monitoring method based on big data acquisition includes the following steps:

[0073] Construct a spatial grid cell set and a monitoring well set for the groundwater monitoring area, determine the spatial grid cell to which each monitoring well belongs, and collect groundwater monitoring datasets;

[0074] Based on groundwater monitoring datasets, the groundwater flow relationships between spatial grid cells are modeled to construct a groundwater propagation network model.

[0075] Based on the groundwater monitoring dataset, anomaly detection is performed on the groundwater quality parameters of the monitoring well points to identify abnormal monitoring well points and obtain the spatial grid cells corresponding to the abnormal monitoring well points.

[0076] Using the spatial grid cells corresponding to the anomaly monitoring well points as the starting nodes, and based on the groundwater propagation network model, a reverse propagation path search is performed to obtain a set of candidate propagation paths;

[0077] The candidate propagation path set is screened, and the target reverse propagation path is determined based on path connectivity and flow consistency;

[0078] Along the reverse propagation path of the target, the path average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell are extracted to generate a path sampling decision parameter set;

[0079] Based on the path sampling decision parameter set, a time-constrained traveling salesman problem is constructed to determine the sampling order and sampling time interval on the target back propagation path, and a path-guided sampling scheme is generated.

[0080] Perform directional sampling according to the path-guided sampling scheme, obtain path sampling data and update the groundwater monitoring dataset;

[0081] Based on the updated groundwater monitoring dataset, the groundwater propagation network model is updated to identify potential pollution source areas and output corresponding monitoring results.

[0082] In this embodiment, the boundary data of the groundwater monitoring area is obtained according to the spatial range of the groundwater monitoring area, and the groundwater monitoring area is divided into grids according to a unified spatial division scale to obtain spatial grid units with spatial location coordinates. The spatial coordinate information of each monitoring well point in the monitoring area is obtained, and the spatial grid unit to which each monitoring well point belongs is determined according to the positional relationship between the spatial coordinates of each monitoring well point and the spatial range of the spatial grid unit. After the assignment of monitoring well points and spatial grid units is determined, groundwater quality parameters, groundwater level, groundwater flow direction, groundwater flow velocity, formation permeability parameters and historical monitoring records are collected through sensing devices and data acquisition terminals deployed at each monitoring well point. The collected data are processed with unified time stamping, and the data from different sources are linked and integrated in chronological order to form a groundwater monitoring dataset.

[0083] In this embodiment, the construction of the groundwater propagation network model specifically includes:

[0084] Based on the groundwater level in the groundwater monitoring data, the corresponding groundwater hydraulic gradient between two adjacent spatial grid cells is calculated. The calculation process of the groundwater hydraulic gradient is as follows: subtract the groundwater level of the downstream spatial grid cell from the groundwater level of the upstream spatial grid cell, and then divide by the distance between the center points of the two spatial grid cells.

[0085] Based on the formation permeability parameters in the groundwater monitoring data, the groundwater flow intensity between adjacent spatial grid cells is calculated. The calculation process for groundwater flow intensity is as follows: the equivalent permeability coefficient between two spatial grid cells is multiplied by the groundwater hydraulic gradient. The equivalent permeability coefficient is the harmonic average of the formation permeability parameters corresponding to the two spatial grid cells.

[0086] Based on the groundwater hydraulic gradient between adjacent spatial grid cells and the groundwater flow direction data in the groundwater monitoring dataset, the groundwater flow direction is determined. When the groundwater level of a spatial grid cell is greater than that of an adjacent spatial grid cell, a groundwater flow relationship is established from the spatial grid cell to the adjacent spatial grid cell. When the groundwater level of a spatial grid cell is less than that of an adjacent spatial grid cell, a reverse groundwater flow relationship is established.

[0087] The process of determining the direction of groundwater flow is as follows: the direction of the groundwater hydraulic gradient is used as the initial flow direction. Groundwater flows from spatial grid cells with high water levels to spatial grid cells with low water levels. Groundwater flow direction data for the corresponding time is obtained and used as a reference for the actual flow direction. The direction of the groundwater hydraulic gradient is compared with the groundwater flow direction data. When the two directions are consistent, the direction is determined as the groundwater flow direction. When the two directions are inconsistent, the direction of the groundwater hydraulic gradient is corrected according to the groundwater flow direction data to obtain the final groundwater flow direction.

[0088] Groundwater flow intensity is used as the weight value of groundwater flow relationship, and groundwater flow direction is used as the directional attribute of groundwater flow relationship. A correlation matrix between spatial grid cells is constructed. The construction process of the correlation matrix is ​​as follows: when there is a directed connection relationship between two spatial grid cells from upstream to downstream, the element value is the corresponding groundwater flow intensity; when there is no directed connection relationship between two spatial grid cells, the element value is zero.

[0089] Using spatial grid cells as nodes, groundwater flow relationships as connecting edges between nodes, and an association matrix as the network structure representation, a groundwater propagation network model is generated, with spatial grid cells as nodes and groundwater flow direction and intensity as edge attributes.

[0090] In this embodiment, obtaining the spatial grid cells corresponding to the anomaly monitoring well points specifically includes:

[0091] The groundwater quality parameters of each monitoring well point at each monitoring time are extracted from the groundwater monitoring dataset, and the various water quality parameters of each monitoring well point are classified and stored to form a water quality parameter data sequence divided by monitoring well point.

[0092] The formation process of the water quality parameter data sequence is as follows: read the water quality parameters corresponding to each monitoring well point at different monitoring times from the groundwater monitoring dataset, sort the water quality parameters of each monitoring well point in chronological order, arrange the water quality parameter values ​​corresponding to different monitoring times in sequence, and process each water quality parameter of each monitoring well point into groups to obtain a water quality parameter data sequence with monitoring well point as the unit and time series as the structure.

[0093] For each monitoring well point, the average level and fluctuation range of each water quality parameter at historical monitoring times are statistically analyzed.

[0094] The statistical process for determining the average level and fluctuation range is as follows: statistical processing is performed on various water quality parameters. The average level of the water quality parameters during the historical monitoring process is obtained by accumulating the values ​​of the water quality parameters at each historical monitoring time and calculating the average value. Based on the differences between the values ​​of the water quality parameters at each historical monitoring time and the average level, the fluctuation range of the water quality parameters during the historical monitoring process is calculated. The fluctuation range is determined by the degree of deviation of each historical monitoring value from the average level and is used to characterize the magnitude of change of the water quality parameters over time.

[0095] The water quality parameters at the current monitoring time are compared with the average levels at the corresponding historical monitoring times to determine the degree of deviation of each water quality parameter from the historical state.

[0096] The process of determining the degree of deviation is as follows: For each monitoring well point, the various water quality parameters at the current monitoring time are compared with the average level of the water quality parameters at historical monitoring times. The difference between the value of the water quality parameter at the current time and the corresponding average level is calculated. The difference is compared with the fluctuation range of the water quality parameter. By judging the position of the difference in the fluctuation range, the degree of deviation of the water quality parameter from the historical state is determined.

[0097] The deviation of various water quality parameters at the same monitoring well point is uniformly quantified to obtain the overall anomaly level of the monitoring well point at the current monitoring time. The monitoring well points with an overall anomaly level greater than the preset anomaly threshold are identified as abnormal monitoring well points. Based on the spatial grid unit to which each monitoring well point belongs, the spatial grid unit corresponding to the abnormal monitoring well point is obtained.

[0098] In this embodiment, obtaining the candidate propagation path set specifically includes:

[0099] The spatial grid cell corresponding to the abnormal monitoring well point is determined as the starting node. In the groundwater propagation network model, the adjacent spatial grid cells connected to the starting node are extracted, and the corresponding groundwater flow relationship is obtained.

[0100] Based on the direction of groundwater flow, the groundwater flow relationships in the groundwater propagation network model are filtered to form a set of connection relationships for reverse propagation path search;

[0101] The screening process is as follows: the groundwater flow relationships are traversed, the flow relationship of groundwater from the downstream spatial grid cell to the upstream spatial grid cell is retained, and the flow direction is used as the constraint condition for path search. All groundwater flow relationships that satisfy the above flow direction constraint are summarized to form a set of connection relationships for reverse propagation path search.

[0102] Starting from the initial node, the path is expanded on the set of connection relationships. The recursive search is performed upstream level by level according to the connection relationship between spatial grid units to generate a path sequence formed by the sequential connection of spatial grid units.

[0103] During the path expansion process, the groundwater flow intensity between adjacent spatial grid cells along each path is accumulated to obtain the cumulative flow intensity corresponding to the path. The cumulative flow intensity is obtained by adding the groundwater flow intensity between each adjacent spatial grid cell in the path in sequence.

[0104] During path expansion, when the current path can no longer continue to expand upstream along the set of connections, the path expansion stops, and the path is identified as a candidate propagation path. All paths that have completed path expansion are then aggregated to form a set of candidate propagation paths.

[0105] In this embodiment, determining the target's backward propagation path specifically includes:

[0106] For each candidate propagation path in the candidate propagation path set, path connectivity is calculated. The path connectivity index is obtained by counting the number of effective connections between adjacent spatial grid cells in the path and calculating the ratio of the number of effective connections to the total number of spatial grid cells in the path. The path connectivity index characterizes the degree of continuous connection between spatial grid cells in the path.

[0107] Extract the groundwater flow direction between adjacent spatial grid cells in each candidate propagation path, and determine whether the groundwater flow direction between adjacent spatial grid cells is consistent with the reverse propagation path direction;

[0108] When the groundwater flow direction is consistent between adjacent spatial grid cells, the flow consistency index is calculated. The flow consistency index characterizes the degree of matching between the groundwater flow direction and the reverse propagation direction in the path.

[0109] The specific calculation process of the flow consistency index is as follows: the connections between all adjacent spatial grid cells in the candidate propagation path are counted, the number of connections between adjacent spatial grid cells whose groundwater flow direction is consistent with the reverse propagation direction is counted, and the total number of connections between adjacent spatial grid cells in the candidate propagation path is counted at the same time. The ratio of the number of connections with the same direction to the total number of connections is calculated to obtain the flow consistency index.

[0110] The groundwater flow intensity between all adjacent spatial grid cells in the candidate propagation path is accumulated and averaged with the number of connections between adjacent spatial grid cells in the candidate propagation path to obtain the path intensity index.

[0111] The path connectivity index, flow consistency index, and path strength index of each candidate propagation path are weighted and summed to obtain the comprehensive path evaluation value.

[0112] The comprehensive evaluation values ​​of each path in the candidate propagation path set are compared, and the candidate propagation path with the largest comprehensive evaluation value is selected as the target reverse propagation path.

[0113] In this embodiment, the generation of the path sampling decision parameter set specifically includes:

[0114] For each adjacent spatial grid cell in the reverse propagation path of the target, the corresponding groundwater flow intensity is extracted to form a path flow intensity sequence;

[0115] The specific process of forming the path flow intensity sequence is as follows: For the spatial grid cells arranged in the connection order in the target reverse propagation path, the groundwater flow intensity between each pair of adjacent spatial grid cells is obtained in sequence. According to the connection order of the spatial grid cells in the target reverse propagation path, the groundwater flow intensity obtained between each adjacent spatial grid cell is arranged in sequence. The groundwater flow intensity between each segment of adjacent spatial grid cells is recorded and combined in sequence to form the path flow intensity sequence.

[0116] Based on the path flow intensity sequence, the flow intensity of all groundwater in the path is accumulated, and the average flow intensity parameter of the path is determined by combining the number of connections between adjacent spatial grid cells in the target reverse propagation path.

[0117] For each spatial grid cell in the path flow intensity sequence, the corresponding historical anomaly distribution features are extracted from the groundwater monitoring dataset to form a path anomaly distribution sequence.

[0118] The path anomaly distribution sequence is formed as follows: For each spatial grid cell, the distribution data of anomalous monitoring well points corresponding to the spatial grid cell at historical monitoring times are extracted from the groundwater monitoring dataset. The existence of anomalous monitoring well points in each historical monitoring time of the spatial grid cell is statistically analyzed moment by moment. Based on the ratio between the number of monitoring times with anomalous monitoring well points and the total number of historical monitoring times of the spatial grid cell, the historical anomaly distribution characteristics of the spatial grid cell are determined. According to the arrangement order of spatial grid cells in the target reverse propagation path, the historical anomaly distribution characteristics corresponding to each spatial grid cell are arranged sequentially, and the arranged results are combined sequentially to form a path anomaly distribution sequence that corresponds one-to-one with the path flow intensity sequence.

[0119] Based on the path anomaly distribution sequence, all historical anomaly distribution features in the path are accumulated and combined with the number of connections between adjacent spatial grid cells in the target back propagation path to determine the path anomaly distribution parameters.

[0120] The groundwater flow intensity corresponding to each spatial grid cell is compared with the path average flow intensity parameter, and the corresponding historical anomaly distribution characteristics are compared with the path anomaly distribution parameters. The path sampling priority parameters of each spatial grid cell are obtained by comprehensive calculation.

[0121] The process of obtaining the path sampling priority parameter is as follows: The groundwater flow intensity corresponding to each spatial grid cell is compared with the path average flow intensity parameter one by one. By calculating the deviation of the groundwater flow intensity of each spatial grid cell from the path average flow intensity parameter, the relative intensity level of the spatial grid cell in terms of groundwater flow is determined. The historical anomaly distribution characteristics corresponding to each spatial grid cell are compared with the path anomaly distribution parameters one by one. By calculating the deviation of the historical anomaly distribution characteristics of each spatial grid cell from the path anomaly distribution parameters, the relative anomaly level of the spatial grid cell in terms of historical anomaly distribution is determined. The relative intensity level of groundwater flow and the relative anomaly level of historical anomaly distribution are jointly processed and superimposed in the same direction, so that spatial grid cells with high groundwater flow intensity and high historical anomaly distribution characteristics obtain high calculation results, while spatial grid cells with low groundwater flow intensity and low historical anomaly distribution characteristics obtain low calculation results. The joint processing result is used as the path sampling priority parameter for the corresponding spatial grid cell.

[0122] The path sampling priority parameters corresponding to each spatial grid cell are associated with the spatial grid cell sequence information in the target backpropagation path to form a path sampling decision parameter set;

[0123] The specific process of forming the path sampling decision parameter set is as follows: according to the connection order of spatial grid cells in the target backpropagation path, each spatial grid cell is sequentially numbered to clarify the positional relationship of each spatial grid cell in the path; the path sampling priority parameter corresponding to each spatial grid cell is matched one-to-one with its sequential position in the target backpropagation path, so that each spatial grid cell has both path position attribute and priority attribute; according to the order of the target backpropagation path, the path sampling priority parameters corresponding to each spatial grid cell are arranged in an orderly manner, and the arranged results are combined to form the path sampling decision parameter set.

[0124] In this embodiment, the generation of the path-guided sampling scheme specifically includes:

[0125] Read the path sampling decision parameter set, and obtain the sequence information, path average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell in the target reverse propagation path, and treat each spatial grid cell as a node to be visited;

[0126] Based on the sequential relationship of the nodes to be visited, the path distances between the nodes to be visited along the reverse propagation path of the target are determined, forming a set of path distances;

[0127] The process of forming the path distance set is as follows: Based on the connection relationship between the nodes to be visited in the target's reverse propagation path, determine whether there is a reachable connection between any two nodes to be visited along the target's reverse propagation path; when there is a connection between two nodes to be visited along the target's reverse propagation path, take the node to be visited upstream as the starting node and the node to be visited downstream as the ending node, and count the number of edges traversed from the starting node to the ending node segment by segment according to the actual connection order of the target's reverse propagation path to obtain the path distance between the two nodes to be visited; perform the above calculation on all nodes to be visited that satisfy the order relationship to obtain a one-to-one path distance value; summarize the multiple path distance values ​​according to the corresponding node pair relationship to form the path distance set;

[0128] Based on the path average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell in the path sampling decision parameter set, the priority of each node to be visited is sorted to obtain the access priority sequence of the nodes to be visited.

[0129] The priority ranking process is as follows: For each node to be visited, the spatial grid cell to which the node belongs is determined, and the path average flow intensity parameter corresponding to the spatial grid cell is read from the path sampling decision parameter set; historical anomaly distribution features associated with the location of the node to be visited are extracted, and the path average flow intensity parameter and historical anomaly distribution features are used as the ranking criteria for the nodes to be visited. The priority evaluation result for each node to be visited is calculated. The calculation method is to perform joint quantization processing on the path average flow intensity parameter and the historical anomaly distribution features. The path average flow intensity parameter is numerically extracted to obtain the flow intensity value representing the level of path flow activity in the area where the node to be visited is located. The historical anomaly distribution features are numerically extracted to obtain the anomaly distribution value representing the degree of historical anomaly aggregation in the area where the node to be visited is located. The flow intensity value and the anomaly distribution value are weighted and summed to obtain the priority evaluation result corresponding to the node to be visited. The nodes to be visited are arranged in order from high to low according to the priority evaluation results of each node to be visited to obtain the order of visits for each node. All the ranked nodes to be visited are combined in sequence to form the access priority sequence of the nodes to be visited.

[0130] Based on the access priority sequence, the latest allowed access time is assigned to each node to be accessed, thus forming a time constraint condition.

[0131] The specific process for forming time constraints is as follows: Based on the access priority sequence of the nodes to be visited, determine the sorting position of each node and obtain the path distance between each node along the target's reverse propagation path. Using the node with the highest priority in the sequence as the starting node, and combining the path distance value corresponding to the starting node, calculate the latest allowed access time. The latest allowed access time is the time value corresponding to the cumulative path distance required to reach the node from the target node along the reverse propagation path. Following the priority sequence, process each subsequent node in turn. For each node, determine the path distance between it and the nodes already sorted before it, and sum the path distance with the latest allowed access time of the preceding nodes to obtain the cumulative access time of the current node. Use this cumulative access time as the latest allowed access time of the node. Summarize the latest allowed access times for all nodes to form the time constraints for each node.

[0132] Based on the nodes to be visited, the set of path distances, and the time constraints, a time-constrained traveling salesman problem is constructed, with the optimization objective being to visit all nodes to be visited while minimizing the path distance.

[0133] The construction process of the time-constrained Traveling Salesman Problem is as follows: Each node to be visited is treated as a visitable object in the time-constrained Traveling Salesman Problem; the path distance between nodes along the backward propagation path to the target is taken as the path cost between any two visitable objects; and the latest allowed visit time for each node to be visited is taken as the time constraint condition that each visitable object must satisfy. Starting from the target node, a visit path covering all nodes to be visited is established, stipulating that each node to be visited is only allowed to be visited once on this visit path, and after the visit is completed, the process continues to the next node to be visited until all nodes have been visited. For any candidate visit path... The path is calculated by accumulating the path distances between adjacent nodes in the order of visits to obtain the actual arrival time of each node to be visited. The actual arrival time is then compared with the latest allowed access time of the corresponding node. If the actual arrival time of a node to be visited is greater than the latest allowed access time, the candidate path is determined to be a feasible path that does not meet the time constraint. If the actual arrival time of a node to be visited is less than or equal to the corresponding latest allowed access time, the candidate path is determined to be a feasible path that meets the time constraint. Among all feasible paths, the optimization objective is to minimize the total path distance to all nodes to be visited, thus forming a time-constrained Traveling Salesman Problem.

[0134] Solve the time-constrained traveling salesman problem to obtain the sampling order of each spatial grid cell;

[0135] The sampling order of each spatial grid cell is obtained as follows: After constructing the time-constrained traveling salesman problem, each node to be visited is mapped to its corresponding spatial grid cell, and the path distance between nodes and the latest allowed access time for each node are used as inputs for the solution. Candidate access paths covering all nodes to be visited are generated according to different access orders. For each candidate access path, the cumulative path distance from the starting node to each node to be visited is calculated in sequence, and the actual arrival time of each node to be visited is obtained based on the cumulative path distance. The actual arrival time of each node to be visited is compared with the corresponding latest allowed access time one by one, and infeasible candidate access paths that are greater than the latest allowed access time are eliminated, while feasible candidate access paths that satisfy all time constraints are retained. The total path distance of all feasible candidate access paths is compared, and the feasible candidate access path with the smallest total path distance is selected as the optimal access path. The sampling order of each spatial grid cell is determined according to the order in which the nodes to be visited in the optimal access path appear.

[0136] Based on the sampling order, the sampling time interval between adjacent spatial grid cells is determined, and a path-guided sampling scheme is generated;

[0137] The specific process of generating the path-guided sampling scheme is as follows: After obtaining the sampling order of each spatial grid unit, adjacent spatial grid units are sequentially paired into consecutive sampling unit pairs according to the sampling order; for each pair of adjacent spatial grid units, the corresponding path distance between them is read, and combined with the actual movement speed when performing the sampling task, the path travel time required to move from the previous spatial grid unit to the next spatial grid unit is calculated. Then, the node dwell time, equipment start-up time, data recording time, and end processing time required to complete the sampling operation in the previous spatial grid unit are added to obtain the sampling time interval from the completion of sampling in the previous spatial grid unit to the start of sampling in the next spatial grid unit; according to the sampling order, the arrangement order of each spatial grid unit, the sampling time interval between each pair of adjacent spatial grid units, the expected arrival time of each spatial grid unit, and the expected start time of sampling are sequentially associated and uniformly arranged to form a complete temporal result, which is used as the path-guided sampling scheme.

[0138] In this embodiment, updating the groundwater monitoring dataset specifically includes:

[0139] Read the sampling order, the sampling time corresponding to each spatial grid unit, and the sampling time interval between adjacent spatial grid units in the path-guided sampling scheme to determine the on-site sampling execution order of each spatial grid unit;

[0140] According to the on-site sampling execution order, the sampling positions corresponding to each spatial grid unit are reached in sequence, groundwater is sampled at the corresponding sampling time, and the groundwater quality parameters corresponding to each spatial grid unit are detected to obtain the path sampling data corresponding to each spatial grid unit.

[0141] The specific process for obtaining path sampling data is as follows: following the on-site sampling execution order, proceed sequentially to the sampling locations corresponding to each spatial grid unit to be sampled, and collect groundwater samples at the planned sampling time corresponding to the spatial grid unit; after sampling is completed, test the water quality parameters of the collected groundwater samples, including pollutant concentration, pH, conductivity, dissolved oxygen, turbidity, and other groundwater monitoring indicators related to pollution identification; record the test results corresponding to each spatial grid unit in a one-to-one correspondence with the unit number, sampling location, sampling time, and sampling execution order of the spatial grid unit to form the path sampling data corresponding to the spatial grid unit;

[0142] The path sampling data corresponding to each spatial grid unit is associated with the corresponding spatial grid unit identifier, sampling time and sampling order to form the unit sampling record corresponding to each spatial grid unit;

[0143] The process of forming unit sampling records is as follows: After obtaining the path sampling data corresponding to each spatial grid unit, the spatial grid unit identification information, actual sampling time, on-site sampling order, and groundwater quality parameter detection results corresponding to each path sampling data are extracted. The above contents are uniformly collected according to the correspondence of the same spatial grid unit and the same sampling behavior. Taking each spatial grid unit as a unit, the spatial grid unit identification, sampling time, sampling order, and various water quality parameter detection values ​​in the path sampling data are bound one by one to generate the unit sampling record corresponding to the spatial grid unit.

[0144] The sampling records of each unit are sequentially collected according to the sampling order to form a path sampling dataset covering the target's backward propagation path;

[0145] The path sampling dataset and the groundwater monitoring dataset are merged to obtain the updated groundwater monitoring dataset;

[0146] The process of obtaining the updated groundwater monitoring dataset is as follows: The spatial grid cell identifiers, sampling locations, sampling times, sampling sequences, and groundwater quality parameter detection results corresponding to each data point in the path sampling dataset are organized and standardized according to the same data organization method as the groundwater monitoring dataset. The standardized path sampling data is then incorporated into the groundwater monitoring dataset line by line. For data in the groundwater monitoring dataset that already has historical records of the same spatial grid cell or monitoring location, the original historical monitoring records are retained, and the new data obtained from the current path sampling is added as new time-series monitoring records to the corresponding data sequence. For data corresponding to spatial grid cells not included in the original groundwater monitoring dataset, the path sampling data corresponding to the spatial grid cell is directly written into the groundwater monitoring dataset to form new unit monitoring records. After all path sampling data has been written, the updated groundwater monitoring dataset, containing both the original groundwater monitoring data and the current path sampling data, is obtained.

[0147] In this embodiment, the updated groundwater monitoring dataset is input into the groundwater propagation network model. The groundwater flow relationship between spatial grid cells is recalibrated, and the abnormality of groundwater quality parameters corresponding to each spatial grid cell is re-determined. Based on the updated distribution of abnormal monitoring data, the spatial grid cells corresponding to each abnormal monitoring well point, the target reverse propagation path, and the groundwater flow relationship between each spatial grid cell, the location of the abnormality source is located step by step along the reverse direction of groundwater propagation. The range of spatial grid cells upstream of the abnormal propagation result is determined. Spatial grid cells with high abnormality, strong propagation correlation, and located at the forefront of the reverse propagation path are identified as potential pollution source areas. The spatial grid cell identifier corresponding to the potential pollution source area, the abnormal result of the corresponding groundwater quality parameter, the target reverse propagation path, and the corresponding monitoring result in the updated groundwater monitoring dataset are output.

[0148] Example 1: This example illustrates the monitoring of groundwater in a chemical industrial park. Located in a plain area, the park's northwest side contains fine chemical production facilities, raw material storage areas, and wastewater storage facilities, while its southeast side includes farmland irrigation areas and village water intake wells. Due to the shallow groundwater depth and high aquifer permeability in this area, contaminants can easily migrate along local dominant channels once polluted, leading to a wider pollution spread. Existing monitoring methods rely primarily on periodic sampling from fixed monitoring wells. While this can detect anomalies at some wells, it often requires additional sampling of surrounding areas based on manual experience, resulting in unclear sampling direction, unreasonable sampling sequence, inaccurate on-site sampling time allocation, and low efficiency in tracing pollution sources. Therefore, this example employs the aforementioned intelligent groundwater sampling monitoring method based on big data acquisition to conduct targeted tracking and sampling monitoring of groundwater anomalies in this area.

[0149] The groundwater monitoring area selected in this embodiment covers approximately six square kilometers. Following a unified spatial division rule, it was constructed into a set of 144 spatial grid units, each with a side length of 200 meters. There were originally 36 monitoring wells within the monitoring area, located at the park boundary, around the production unit area, around the wastewater storage facility, along the downstream migration direction of groundwater, and near external sensitive receptors. First, each monitoring well was matched with its corresponding spatial grid unit, and groundwater monitoring data for twelve consecutive months was collected. The monitoring data included groundwater level, flow direction, hydraulic gradient, and groundwater quality parameters, primarily chemical oxygen demand (COD), ammonia nitrogen, total dissolved solids (TDS), chloride ions, volatile organic compounds (VOCs), and heavy metals. By analyzing historical monitoring data, it was found that three monitoring wells located downstream on the east side of the park showed abnormally high VOC concentrations in the most recent monitoring period. Two of these wells also exhibited intermittent abnormal fluctuations in their corresponding spatial grid units during historical months.

[0150] First, based on the groundwater monitoring dataset, the groundwater flow relationships between spatial grid cells are modeled to construct a groundwater propagation network model. This model reflects the propagation direction and correlation strength of groundwater between different spatial grid cells. Then, anomaly detection is performed on the groundwater quality parameters corresponding to each monitoring well point to identify anomalous monitoring well points, and the spatial grid cells corresponding to these anomalous monitoring well points are used as starting nodes. Based on these starting nodes, a reverse propagation path search is performed according to the groundwater propagation network model, resulting in multiple candidate propagation paths. Subsequently, the candidate propagation paths are filtered based on path connectivity and flow consistency to determine a target reverse propagation path. This target reverse propagation path extends gradually upstream from the anomalous spatial grid cell located downstream of the park, ultimately pointing to several spatial grid cells on the northwest side of the park near the raw material storage tank area and the wastewater temporary storage facility.

[0151] Along the reverse propagation path from the target, the path-average flow intensity parameters and historical anomaly distribution characteristics are extracted from the involved spatial grid cells to form a path sampling decision parameter set. Spatial grid cells with higher path-average flow intensity and higher frequency of historical anomaly distributions are assigned higher access priority. Simultaneously, combining the path distance relationships between each node to be visited along the reverse propagation path from the target, a path distance set is formed. Then, based on the path-average flow intensity parameters, historical anomaly distribution characteristics, and the path distance set, each node to be visited is comprehensively ranked to obtain an access priority sequence. Next, according to the access priority sequence, a latest allowed access time is assigned to each node to be visited, forming a time constraint. Then, a time-constrained traveling salesman problem is constructed using each node to be visited, the path distance set, and the time constraint. After solving, a sampling order that satisfies the time constraint and has a smaller total path distance is obtained. Furthermore, the sampling time interval between adjacent spatial grid cells is determined, ultimately generating a path-guided sampling scheme.

[0152] According to the path-guided sampling scheme, on-site sampling personnel, carrying portable sampling devices and rapid on-site testing equipment, sequentially arrived at the corresponding sampling locations of each spatial grid unit along the target reverse propagation path, completing groundwater sampling and testing groundwater quality parameters at the corresponding sampling time, thus obtaining path sampling data. The path sampling data not only includes the water quality parameter test results corresponding to each spatial grid unit, but also simultaneously records the spatial grid unit identifier, sampling time, and sampling order, forming a unit sampling record. All unit sampling records are summarized to form a path sampling dataset, which is then merged with the original groundwater monitoring dataset to obtain an updated groundwater monitoring dataset. Based on the updated groundwater monitoring dataset, the groundwater propagation network model is updated, and the direction of pollution propagation and potential pollution source areas are reassessed. Ultimately, three consecutive spatial grid units between the southern edge of the raw material storage tank area on the northwest side of the park and the sewage temporary storage facility were identified as having the highest pollution correlation, constituting a potential pollution source area. Subsequent on-site verification revealed that an underground pipeline interface in this area showed signs of sealing aging, and there were traces of leakage under the nearby surface hardened layer, consistent with the identification results of this invention.

[0153] To further verify the beneficial effects of this invention, the method of this invention was compared with the conventional manual experience-based supplementary sampling method previously used in the park. The conventional method typically expands sampling radially outwards from the anomalous well point, with on-site personnel determining the sampling points and order based on experience, lacking integrated correlation analysis with the groundwater propagation path. This embodiment selected similar anomalous events for comparative statistical analysis, and the results are shown in Table 1 below.

[0154] Table 1 Comparison of the effects of the method of the present invention and conventional supplementary sampling methods

[0155] Average number of sampling spatial grid cells per anomaly event 15 9 Reduced by 40.0% Average sampling path length for a single anomaly event 8.6 kilometers 5.1 km Reduced by 40.7% Average time from anomaly detection to completion of sampling in key areas 6.5 hours 4.0 hours Shortened by 38.5% Average sampling personnel input per event 6 people 4 people Reduced by 33.3% Average number of rounds required to initially identify potential pollution source areas 3 rounds Round 1 Reduced by 66.7% Key spatial grid cells complete sampling ratio on time 68.4% 94.7% An increase of 26.3 percentage points Accuracy of identifying potential pollution source areas 74.2% 91.8% An increase of 17.6 percentage points The consistency rate between abnormal propagation paths and on-site verification results 70.5% 89.6% An increase of 19.1 percentage points Average number of invalid sampling points per event 5 1 Reduced by 80.0% Contribution of the updated groundwater monitoring dataset to subsequent model corrections 0.61 0.84 Increased by 37.7%

[0156] As shown in Table 1, when similar levels of anomaly are observed, the method of this invention reduces the number of actual sampling spatial grid units on-site from an average of 15 to 9, a reduction of 40%; the total sampling path length is reduced from 8.6 kilometers to 5.1 kilometers; the total time from anomaly detection to completion of sampling in key areas is shortened from 6.5 hours to 4.0 hours; the number of rounds required to initially pinpoint potential pollution source areas is reduced from 3 to 1; the accuracy rate of potential pollution source area identification is increased from 74.2% to 91.8%; and the consistency between the anomaly propagation path and subsequent verification results is significantly improved. These data demonstrate that this invention effectively solves the problems of unclear sampling direction after anomaly occurrence, excessive sampling points, low on-site efficiency, and long pollution source investigation cycles in existing technologies.

[0157] From the perspective of path sampling data quality, this invention couples the sampling order with the groundwater propagation path, allowing upstream high-risk spatial grid cells to be prioritized for access within a more reasonable time window, thereby reducing the interference of local pollution plume changes caused by time delays on the judgment results. In this embodiment, the average time interval between adjacent sampling points is controlled between 18 and 35 minutes under the path-guided sampling scheme, and key spatial grid cells are all sampled within the limited latest allowable access time, ensuring better temporal consistency of path sampling data in the same round. The updated groundwater monitoring dataset can more accurately reflect the continuous changes in the anomaly propagation chain, thereby improving the reliability of groundwater propagation network model updates.

[0158] This embodiment verifies, through its practical application in a groundwater anomaly monitoring scenario in a chemical industrial park, that the present invention can not only rapidly deduce the target's reverse propagation path based on anomaly monitoring well points, but also further optimize the path through time constraints to form a path-guided sampling scheme that is executable on-site. This makes the sampling behavior more concentrated, the path more reasonable, and the sampling efficiency higher. Furthermore, it can support the accurate identification of potential pollution source areas with the updated groundwater monitoring dataset, demonstrating strong engineering application value.

[0159] 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 technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A groundwater sampling intelligent monitoring method based on big data acquisition, characterized in that, Includes the following steps: Construct a spatial grid cell set and a monitoring well set for the groundwater monitoring area, determine the spatial grid cell to which each monitoring well belongs, and collect groundwater monitoring datasets; Based on groundwater monitoring datasets, the groundwater flow relationships between spatial grid cells are modeled to construct a groundwater propagation network model. Based on the groundwater monitoring dataset, anomaly detection is performed on the groundwater quality parameters of the monitoring well points to identify abnormal monitoring well points and obtain the spatial grid cells corresponding to the abnormal monitoring well points. Using the spatial grid cells corresponding to the anomaly monitoring well points as the starting nodes, and based on the groundwater propagation network model, a reverse propagation path search is performed to obtain a set of candidate propagation paths; The candidate propagation path set is screened, and the target reverse propagation path is determined based on path connectivity and flow consistency; Along the reverse propagation path of the target, the path average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell are extracted to generate a path sampling decision parameter set; Based on the path sampling decision parameter set, a time-constrained traveling salesman problem is constructed to determine the sampling order and sampling time interval on the target back propagation path, and a path-guided sampling scheme is generated. Perform directional sampling according to the path-guided sampling scheme, acquire path sampling data, and update the groundwater monitoring dataset; Based on the updated groundwater monitoring dataset, the groundwater propagation network model is updated to identify potential pollution source areas and output corresponding monitoring results.

2. The intelligent monitoring method for groundwater sampling based on big data acquisition according to claim 1, characterized in that, Based on the spatial scope of the groundwater monitoring area, regional boundary data is obtained, and the groundwater monitoring area is divided into grids according to a unified spatial division scale to obtain spatial grid units with spatial location coordinates. The spatial coordinate information of each monitoring well point within the monitoring area is obtained, and the spatial grid unit to which each monitoring well point belongs is determined based on the positional relationship between the spatial coordinates of each monitoring well point and the spatial scope of the spatial grid unit. Through sensing devices and data acquisition terminals deployed at each monitoring well point, groundwater quality parameters, groundwater level, groundwater flow direction, groundwater flow velocity, formation permeability parameters, and historical monitoring records are collected. All types of collected data are uniformly processed with time stamps, and data from different sources are correlated and integrated in chronological order to form a groundwater monitoring dataset.

3. The intelligent monitoring method for groundwater sampling based on big data acquisition according to claim 1, characterized in that, The construction of the groundwater propagation network model specifically includes: Based on the groundwater level in the groundwater monitoring data, calculate the corresponding groundwater hydraulic gradient between two adjacent spatial grid cells; Calculate the groundwater flow intensity between adjacent spatial grid cells based on the formation permeability parameters in the groundwater monitoring data. Based on the groundwater hydraulic gradient between adjacent spatial grid cells and the groundwater flow direction data in the groundwater monitoring dataset, the groundwater flow direction is determined. When the groundwater level of a spatial grid cell is greater than that of an adjacent spatial grid cell, a groundwater flow relationship is established from the spatial grid cell to the adjacent spatial grid cell. When the groundwater level of a spatial grid cell is less than that of an adjacent spatial grid cell, a reverse groundwater flow relationship is established. By using groundwater flow intensity as the weight value of groundwater flow relationship and groundwater flow direction as the directional attribute of groundwater flow relationship, an association matrix between spatial grid cells is constructed. Using spatial grid cells as nodes, groundwater flow relationships as connecting edges between nodes, and an association matrix as the network structure representation, a groundwater propagation network model is generated, with spatial grid cells as nodes and groundwater flow direction and intensity as edge attributes.

4. The intelligent monitoring method for groundwater sampling based on big data acquisition according to claim 1, characterized in that, The specific steps for obtaining the spatial grid cells corresponding to the anomaly monitoring well points include: The groundwater quality parameters of each monitoring well point at each monitoring time are extracted from the groundwater monitoring dataset, and the various water quality parameters of each monitoring well point are classified and stored to form a water quality parameter data sequence divided by monitoring well point. For the water quality parameter data sequence of each monitoring well point, the average level and fluctuation range of each water quality parameter at the historical monitoring time are statistically analyzed; The water quality parameters at the current monitoring time are compared with the average levels at the corresponding historical monitoring times to determine the degree of deviation of each water quality parameter from the historical state. The deviation of various water quality parameters at the same monitoring well point is uniformly quantified to obtain the overall anomaly level of the monitoring well point at the current monitoring time. The monitoring well points with an overall anomaly level greater than the preset anomaly threshold are identified as abnormal monitoring well points. Based on the spatial grid unit to which each monitoring well point belongs, the spatial grid unit corresponding to the abnormal monitoring well point is obtained.

5. The intelligent monitoring method for groundwater sampling based on big data acquisition according to claim 1, characterized in that, The specific steps to obtain the candidate propagation path set include: The spatial grid cell corresponding to the abnormal monitoring well point is determined as the starting node. In the groundwater propagation network model, the adjacent spatial grid cells connected to the starting node are extracted, and the corresponding groundwater flow relationship is obtained. Based on the direction of groundwater flow, the groundwater flow relationships in the groundwater propagation network model are filtered to form a set of connection relationships for reverse propagation path search; Starting from the initial node, the path is expanded on the set of connection relationships. The recursive search is performed upstream level by level according to the connection relationship between spatial grid units to generate a path sequence formed by the sequential connection of spatial grid units. During the path expansion process, the groundwater flow intensity between adjacent spatial grid cells traversed by each path is accumulated to obtain the cumulative flow intensity corresponding to the path. During path expansion, when the current path can no longer continue to expand upstream along the set of connections, the path expansion stops, and the path is identified as a candidate propagation path. All paths that have completed path expansion are then aggregated to form a set of candidate propagation paths.

6. The intelligent monitoring method for groundwater sampling based on big data acquisition according to claim 1, characterized in that, The determination of the target reverse propagation path specifically includes: For each candidate propagation path in the candidate propagation path set, the path connectivity is calculated to obtain the path connectivity index; Extract the groundwater flow direction between adjacent spatial grid cells in each candidate propagation path, and determine whether the groundwater flow direction between adjacent spatial grid cells is consistent with the reverse propagation path direction; When the groundwater flow direction is consistent among adjacent spatial grid cells, the flow consistency index is calculated. The groundwater flow intensity between all adjacent spatial grid cells in the candidate propagation path is accumulated and averaged with the number of connections between adjacent spatial grid cells in the candidate propagation path to obtain the path intensity index. The path connectivity index, flow consistency index, and path strength index of each candidate propagation path are weighted and summed to obtain the comprehensive path evaluation value. The comprehensive evaluation values ​​of each path in the candidate propagation path set are compared, and the candidate propagation path with the largest comprehensive evaluation value is selected as the target reverse propagation path.

7. The intelligent monitoring method for groundwater sampling based on big data acquisition according to claim 1, characterized in that, The generation of the path sampling decision parameter set specifically includes: For each adjacent spatial grid cell in the reverse propagation path of the target, the corresponding groundwater flow intensity is extracted to form a path flow intensity sequence; Based on the path flow intensity sequence, the flow intensity of all groundwater in the path is accumulated, and the average flow intensity parameter of the path is determined by combining the number of connections between adjacent spatial grid cells in the target reverse propagation path. For each spatial grid cell in the path flow intensity sequence, the corresponding historical anomaly distribution features are extracted from the groundwater monitoring dataset to form a path anomaly distribution sequence. Based on the path anomaly distribution sequence, all historical anomaly distribution features in the path are accumulated and combined with the number of connections between adjacent spatial grid cells in the target back propagation path to determine the path anomaly distribution parameters. The groundwater flow intensity corresponding to each spatial grid cell is compared with the path average flow intensity parameter, and the corresponding historical anomaly distribution characteristics are compared with the path anomaly distribution parameters. The path sampling priority parameters of each spatial grid cell are obtained by comprehensive calculation. The path sampling priority parameters corresponding to each spatial grid cell are associated with the spatial grid cell sequence information in the target backpropagation path to form a path sampling decision parameter set.

8. The intelligent monitoring method for groundwater sampling based on big data acquisition according to claim 1, characterized in that, The generation of the path-guided sampling scheme specifically includes: Read the path sampling decision parameter set, and obtain the sequence information, path average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell in the target reverse propagation path, and treat each spatial grid cell as a node to be visited; Based on the sequential relationship of the nodes to be visited, the path distances between the nodes to be visited along the reverse propagation path of the target are determined, forming a set of path distances; Based on the path average flow intensity parameters and historical anomaly distribution characteristics of each spatial grid cell in the path sampling decision parameter set, the priority of each node to be visited is sorted to obtain the access priority sequence of the nodes to be visited. Based on the access priority sequence, the latest allowed access time is assigned to each node to be accessed, thus forming a time constraint condition. Based on the nodes to be visited, the set of path distances, and the time constraints, a time-constrained traveling salesman problem is constructed, with the optimization objective being to visit all nodes to be visited while minimizing the path distance. Solve the time-constrained traveling salesman problem to obtain the sampling order of each spatial grid cell; Based on the sampling order, the sampling time interval between adjacent spatial grid cells is determined, and a path-guided sampling scheme is generated.

9. The intelligent monitoring method for groundwater sampling based on big data acquisition according to claim 1, characterized in that, The update of the groundwater monitoring dataset specifically includes: Read the sampling order, the sampling time corresponding to each spatial grid unit, and the sampling time interval between adjacent spatial grid units in the path-guided sampling scheme to determine the on-site sampling execution order of each spatial grid unit; According to the on-site sampling execution order, the sampling positions corresponding to each spatial grid unit are reached in sequence, groundwater is sampled at the corresponding sampling time, and the groundwater quality parameters corresponding to each spatial grid unit are detected to obtain the path sampling data corresponding to each spatial grid unit. The path sampling data corresponding to each spatial grid unit is associated with the corresponding spatial grid unit identifier, sampling time and sampling order to form the unit sampling record corresponding to each spatial grid unit; The sampling records of each unit are sequentially collected according to the sampling order to form a path sampling dataset covering the target's backward propagation path; The path sampling dataset and the groundwater monitoring dataset are merged to obtain the updated groundwater monitoring dataset.

10. The intelligent monitoring method for groundwater sampling based on big data acquisition according to claim 1, characterized in that, The updated groundwater monitoring dataset is input into the groundwater propagation network model to recalibrate the groundwater flow relationships between spatial grid cells and re-determine the anomalies in groundwater quality parameters corresponding to each spatial grid cell. Based on the updated distribution of anomaly monitoring data, the spatial grid cells corresponding to each anomaly monitoring well point, the target reverse propagation path, and the groundwater flow relationships between spatial grid cells, the anomaly source location is located step by step along the reverse direction of groundwater propagation. The range of spatial grid cells upstream of the anomaly propagation results is determined. Spatial grid cells with high anomaly degree, strong propagation correlation, and located at the forefront of the reverse propagation path are identified as potential pollution source areas. The spatial grid cell identifiers corresponding to the potential pollution source areas, the corresponding anomaly results of groundwater quality parameters, the target reverse propagation path, and the corresponding monitoring results in the updated groundwater monitoring dataset are output.

Images