A method and system for accurately dividing the boundary of a groundwater pollution risk prevention and control area

Abstract

The present invention discloses a method and system for accurately demarcating the boundary of a groundwater pollution risk prevention and control area. By building a composite model based on meta-learning for groundwater pollution risk zoning, high-precision data on the distribution of groundwater pollution in non-sampled areas is filled to preliminarily demarcate the groundwater pollution risk prevention and control area; by calculating the spatial variation coefficient and based on the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the grid cell size, the grid refinement of the spatio-temporal variability of groundwater pollution risk is realized; by coupling the elements of groundwater richness, vulnerability, and source load, the boundary of the groundwater pollution risk prevention and control area is accurately defined; and borehole verification based on the spatial resolution of the prevention and control area is carried out to ensure the high precision and accuracy of the demarcation of the boundary of the groundwater pollution risk prevention and control area. The present invention solves the problems of difficult accurate characterization of the spatio-temporal variability of groundwater pollution and difficult accurate demarcation of the boundary of the groundwater pollution risk prevention and control area under small sample data.

Description

Technical Field

[0001] The present invention relates to the technical field of groundwater pollution risk prevention and control zone demarcation, and in particular to a method and system for accurately demarcating the boundaries of groundwater pollution risk prevention and control zones. Background Art

[0002] Groundwater is an important source of drinking water in my country. Accurate demarcation of groundwater pollution risk control zones is crucial for the prevention and control of groundwater pollution and the protection of groundwater resources. At present, my country's groundwater pollution risk control zones are mainly divided by spatially superimposing regional groundwater quality status, water richness, vulnerability, source load and other factors. However, there are still bottlenecks such as inaccurate boundary definition and low range resolution of groundwater pollution risk control zones. The main reasons include the small amount of monitoring data, low effectiveness, and low precision and accuracy of the "point-to-surface" prediction model.

[0003] In order to solve the problems of poor suitability and low precision of groundwater pollution risk control zone division due to the lack of a prediction model based on a small amount of data to characterize the spatiotemporal heterogeneity of groundwater pollution and a method to accurately define the boundaries of groundwater pollution risk control zones, a method for accurate demarcation of groundwater pollution risk control zones was proposed based on a small sample prototype network based on meta-learning. The method can realize the identification of groundwater pollution risk control zones with appropriate resolution and accurate boundaries, which is one of the important research trends in the field of groundwater pollution risk control zone division technology. Summary of the invention

[0004] The present invention overcomes the deficiencies of the prior art and provides a method and system for accurately dividing the boundaries of groundwater pollution risk prevention and control areas.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] The first aspect of the present invention provides a method for accurately dividing the boundaries of groundwater pollution risk prevention and control areas, comprising the following steps:

[0007] Build a meta-learning-based composite model for groundwater pollution risk zoning, fill in high-precision data on groundwater pollution distribution in non-sampling areas, and preliminarily delineate groundwater pollution risk prevention and control areas;

[0008] By calculating the spatial variation coefficient and based on the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the size of the grid unit, the spatiotemporal heterogeneity of groundwater pollution risk can be refined.

[0009] Couple the groundwater richness, vulnerability and source load factors to accurately define the boundaries of the groundwater pollution risk prevention and control zone.

[0010] Furthermore, in the present invention, a composite model for groundwater pollution risk zoning based on meta-learning is built to fill in high-precision data of groundwater pollution distribution in non-sampling areas, and preliminarily delineate groundwater pollution risk prevention and control areas, specifically including:

[0011] By selecting and cleaning no less than 10 classic groundwater migration mechanism models, a groundwater migration composite model covering the "source-flow-sink" relationship of 8 typical industries and various typical groundwater hydrogeological characteristics was integrated;

[0012] Based on the groundwater migration composite model, a prototype network in the meta-learning field is constructed, and the attention mechanism is coupled to build a composite model for groundwater pollution risk zoning based on meta-learning.

[0013] Integrate multi-year, multi-regional groundwater historical monitoring data, new well sampling data and online monitoring data, conduct exploratory data analysis, establish a spatiotemporal distribution feature set of groundwater pollution risk at different regional scales, hydrogeological conditions and pollution types, and split it into training and validation sets;

[0014] On the basis of maintaining the accurate simulation of well-fitted data by the composite model of groundwater pollution risk zoning, we targeted underfitted data samples and continuously improved the generalization ability and accuracy of the composite model of groundwater pollution risk zoning by optimizing the simulation effect of underfitted data sets. We also conducted uncertainty analysis on the simulation results of the model based on classification error, data completion error and model error.

[0015] Based on the background information of the study area and the sampling point data, a meta-learning-based composite model for groundwater pollution risk zoning is applied to supplement the groundwater pollution data of non-sampling points, characterize the spatiotemporal heterogeneity of regional groundwater pollution risks, and achieve the preliminary delineation of groundwater pollution risk prevention and control areas.

[0016] Furthermore, in the present invention, by calculating the spatial variation coefficient, according to the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the grid unit size, the spatiotemporal heterogeneity grid of groundwater pollution risk is refined, specifically including:

[0017] Set N initial central starting points on the boundary of the preliminarily delineated groundwater pollution risk prevention and control area, calculate the spatial variation coefficients from the N initial central starting points to other points with a distance of h in all directions around them, and then increase the value of h by Δh in sequence, and continue to calculate the spatial variation coefficient;

[0018] The minimum value is taken from the calculated multiple spatial variation coefficients, and the feature space from the marked point to the corresponding initial central starting point is partitioned as a separate grid unit;

[0019] The root mean square error (RMSE), mean absolute error (MAE) and correlation coefficient (R 2 ) is used as the accuracy index to evaluate each feature space grid unit;

[0020] According to the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the size of the grid unit, and based on the requirements of spatial resolution, the spatiotemporal heterogeneity of groundwater pollution risk grid is refined.

[0021] Furthermore, in the present invention, the groundwater richness, vulnerability, and source load factors are coupled to accurately define the boundaries of the groundwater pollution risk prevention and control area, specifically including:

[0022] With the help of ArcGIS software, based on spatial analysis methods, multi-factor overlay analysis was performed to couple the spatiotemporal heterogeneity grid of groundwater pollution risk and the multi-factor information of groundwater richness, vulnerability, and source load;

[0023] Analyze the boundaries of groundwater pollution risk prevention and control zones and achieve accurate demarcation of the boundaries of groundwater pollution risk prevention and control zones.

[0024] Furthermore, the present invention carries out drilling verification based on the spatial resolution of the prevention and control zone to ensure high precision and accuracy in the demarcation of the boundaries of the groundwater pollution risk prevention and control zone.

[0025] The second aspect of the present invention provides a system for accurately dividing the boundaries of groundwater pollution risk prevention and control areas, specifically comprising:

[0026] The first meta-learning data filling module is used to preliminarily delineate groundwater pollution risk prevention and control areas;

[0027] The second coefficient of variation extraction module is used to refine the spatiotemporal heterogeneity grid of groundwater pollution risk;

[0028] The third multi-factor boundary analysis module is used to accurately define the boundaries of groundwater pollution risk prevention and control areas;

[0029] The fourth boundary drilling verification module is used to ensure the high precision and accuracy of the boundary demarcation of the groundwater pollution risk prevention and control zone.

[0030] The present invention has the following beneficial effects:

[0031] The present invention discloses a method and system for accurately dividing the boundaries of groundwater pollution risk control areas, builds a composite model for groundwater pollution risk zoning based on meta-learning, fills in high-precision data of groundwater pollution distribution in non-sampling areas, and preliminarily delineates groundwater pollution risk control areas; by calculating the spatial variation coefficient, based on the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the size of the grid unit, the spatiotemporal heterogeneity of groundwater pollution risk is refined; coupling the groundwater water richness, vulnerability, and source load elements, accurately defining the boundaries of groundwater pollution risk control areas; carrying out drilling verification based on the spatial resolution of the control area, ensuring the high precision and accuracy of the boundary division of the groundwater pollution risk control area. It solves the problem that it is difficult to accurately depict the spatiotemporal heterogeneity of groundwater pollution and the boundary of the groundwater pollution risk control area under small sample data, and realizes the precise division of the boundaries of the groundwater pollution risk control area with strong suitability and high precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, drawings of other embodiments can be obtained based on these drawings without paying creative work.

[0033] Figure 1 A schematic diagram of a process for accurately dividing the boundaries of groundwater pollution risk prevention and control areas is shown;

[0034] Figure 2 A schematic diagram of the structure of a system for accurately dividing the boundaries of groundwater pollution risk prevention and control areas is shown;

[0035] Figure 3 A schematic diagram of the process of a composite model for groundwater pollution risk zoning based on meta-learning is shown;

[0036] Figure 4 The evaluation results of the accuracy of the grid unit division of the spatiotemporal heterogeneity of groundwater pollution risk are shown;

[0037] Figure 5 A schematic diagram showing the division of grid cells for the temporal and spatial heterogeneity of groundwater pollution risk is shown;

[0038] Figure 6 The results of precise demarcation of the boundaries of groundwater pollution risk prevention and control areas are shown. DETAILED DESCRIPTION

[0039] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0041] like Figure 1 As shown, the first aspect of the present invention provides a method for accurately dividing the boundaries of groundwater pollution risk prevention and control areas, comprising the following steps:

[0042] S101: Build a meta-learning-based composite model for groundwater pollution risk zoning, fill in high-precision data on groundwater pollution distribution in non-sampling areas, and preliminarily delineate groundwater pollution risk prevention and control areas;

[0043] S102: By calculating the spatial variation coefficient, based on the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the grid unit size, the spatiotemporal heterogeneity of groundwater pollution risk is refined;

[0044] S103: Couple groundwater richness, vulnerability, and source load factors to accurately define the boundaries of groundwater pollution risk prevention and control areas;

[0045] S104: Carry out drilling verification based on the spatial resolution of the prevention and control zone to ensure high precision and accuracy in the demarcation of the boundaries of the groundwater pollution risk prevention and control zone.

[0046] The present invention solves the problem that it is difficult to accurately depict the spatiotemporal heterogeneity of groundwater pollution and to accurately define the boundaries of groundwater pollution risk prevention and control zones under small sample data, and realizes the precise division of the boundaries of groundwater pollution risk prevention and control zones with strong suitability and high precision, providing practical engineering reference value and prevention and control priorities for the prevention and control and remediation of groundwater pollution and the protection and sustainable utilization of groundwater resources.

[0047] Among them, further, S101, in one embodiment of the present invention, a composite model for groundwater pollution risk zoning based on meta-learning is built, high-precision data of groundwater pollution distribution in non-sampling areas is filled, and groundwater pollution risk prevention and control areas are preliminarily delineated, specifically including:

[0048] By selecting and cleaning no less than 10 classic groundwater migration mechanism models, a groundwater migration composite model covering the "source-flow-sink" relationship of 8 typical industries and various typical groundwater hydrogeological characteristics was integrated;

[0049] It should be noted that the eight typical industries refer to industrial pollution sources, mining areas, hazardous waste disposal sites, landfills, gas stations, agricultural pollution sources, golf courses, and surface sewage. The relationship between groundwater "source-flow-sink" may vary greatly in different regions, different groundwater hydrogeological characteristics, and different industries;

[0050] It should be noted that the classic groundwater migration mechanism model embeds the knowledge of groundwater "source-flow-sink" under certain conditions. Therefore, it may be possible to more accurately simulate the spatiotemporal heterogeneity of groundwater pollution risk under a certain region, certain hydrogeological conditions, and certain pollution source. However, there is no groundwater migration mechanism model that covers the "source-flow-sink" relationship of 8 typical industries and various typical groundwater hydrogeological characteristics. Therefore, it is currently impossible to use a groundwater migration mechanism model that can be simultaneously applicable to any region, any hydrogeological conditions, and any pollution source. Accurate simulation of the spatiotemporal heterogeneity of groundwater pollution risk. By selecting and cleaning no less than 10 classic groundwater migration mechanism models and integrating the groundwater migration composite model, it is possible to achieve the aggregation and coupling of the "source-flow-sink" relationship of 8 typical industries and various typical groundwater hydrogeological characteristics, and use this as a priori knowledge item of the meta-learning prototype network;

[0051] Based on the groundwater migration composite model, a meta-learning prototype network is constructed and coupled with the attention mechanism, thereby building a composite model for groundwater pollution risk zoning based on meta-learning;

[0052] It should be noted that the prototype network in the field of meta-learning is an efficient small sample learning method. Its basic idea is to create a prototype representation for each classification, and for a query that needs to be classified, the distance between the prototype vector of the classification and the query point is calculated to determine the softmax probability result. Specifically, assuming that the dataset is D, the representation of the sample inside it is {(x 1 ,y 1 ), (x 2 ,y 2 ),…,(x n ,y n )}, where x represents a vector and y represents a classification label; for each classification, n sample points are randomly generated from the total sample set, which is the support set S; n sample points are randomly selected from the total sample set for each classification to generate the query set Q; for the sample points of the support set S, the encoding formula is used Generate a prototype representation for each category. The encoding formula of this study is It is a CNN. For each category, all sample codes are summed and averaged, and the result is used as the prototype representation of the category, which is: Similarly, the query set also generates sample encodings and calculates the distance from the query set to the prototype representation of the support set. The minimum distance is used to determine which category the query sample belongs to. The distance calculation can usually use the Euclidean distance or COS similarity Etc.; Finally, calculate the probability p that the query sample belongs to each category w (y=k|x), where the softmax calculation method is used to convert the distance into probability form: Finally, the loss function of the prototype network in the meta-learning domain is calculated Will Substitution You can get:

[0053] It should be noted that choosing different distance metrics may bring different performance to the model. In order to study the distance metric setting, we use the Euclidean distance function and the cosine distance function to train the prototype network we proposed. After testing, the convergence of the Euclidean distance function is significantly better than that of the cosine distance function. The Euclidean distance model shows higher test accuracy than the cosine distance model. Therefore, the Euclidean distance is more suitable for the prototype network of this study;

[0054] It should be noted that the attention mechanism improves the feature representation of the network by focusing on important features and ignoring unnecessary information. It has been widely used in computer vision tasks. Spatial attention can use the spatial relationship of features to select. First, prior knowledge and attention mechanism are introduced into the meta-learning CNN network. Prior knowledge is responsible for helping the meta-learning network express input data in the representation space, while the attention mechanism enables meta-learning to focus on key data features in the representation space. The feature extraction network that introduces the attention module can extract high-level and discriminative features, further reducing the misjudgment rate. Spatial attention mechanism representation: \({M}_{SAM}\in{R}^{1\timesH\timesW}\);

[0055] Integrate multi-year, multi-regional groundwater historical monitoring data, new well sampling data and online monitoring data, conduct exploratory data analysis, establish a spatiotemporal distribution feature set of groundwater pollution risk at different regional scales, hydrogeological conditions and pollution types, and split it into training and validation sets;

[0056] It should be noted that exploratory data analysis mainly includes the use of statistical methods and visualization tools to explore data distribution, correlation, outliers and other characteristics, as well as data cleaning and preprocessing, and splitting the groundwater pollution risk spatiotemporal distribution feature set into a training set and a validation set at a ratio of 4:1;

[0057] On the basis of maintaining the accurate simulation of well-fitted data by the composite model of groundwater pollution risk zoning, we targeted underfitted data samples and continuously improved the generalization ability and accuracy of the composite model of groundwater pollution risk zoning by optimizing the simulation effect of underfitted data sets. We also conducted uncertainty analysis on the simulation results of the model based on classification error, data completion error and model error.

[0058] It should be noted that the classification error:

[0059]

[0060] Where:

[0061] F c is a predefined probability threshold;

[0062] To predict the results;

[0063] Z(x) is the pollution value at x in space, Z c is the threshold value;

[0064] It should be noted that the data completion error is:

[0065] I e (x; Z c )=δ(x 0 ; Z c |(n))

[0066] Where:

[0067] δ(x 0 ; Z c |(n)) is the space x 0 The prediction error of the cumulative probability that the threshold value Zc is not exceeded;

[0068] It should be noted that the model error:

[0069] p(u,∑|D)=NIW(u,∑|m N , K N , V N , S N )

[0070] Where:

[0071] m N : posterior mean;

[0072] K N : total data volume;

[0073] V N : posterior degrees of freedom;

[0074] S N : posterior error;

[0075] Based on the background information of the study area and the sampling point data, a composite model based on meta-learning for groundwater pollution risk zoning is applied to supplement the groundwater pollution data of non-sampling points, characterize the spatiotemporal heterogeneity of regional groundwater pollution risks, and realize the preliminary delineation of groundwater pollution risk prevention and control areas;

[0076] It should be noted that the detailed description of the spatiotemporal heterogeneity of groundwater pollution risk requires a large amount of point pollution data support. However, in actual work, due to the difficulty and high cost of data collection, the number of sampling points is insufficient, the pollution data of non-sampling points is missing, and the description of the spatiotemporal heterogeneity of groundwater pollution risk is not accurate. Figure 3 As shown in the figure, x 1 and x 2 All of them are non-sampling points, lacking groundwater pollution data. The background information of the proposed study area and the sampling point data are input into the composite model for groundwater pollution risk zoning based on meta-learning. First, the key features are extracted through the attention mechanism, and then the non-sampling point x is output based on the "source-flow-sink" prior knowledge in the prototype network, which has continuously improved performance and has high generalization ability. 1 and x 2 The high-precision pollution data of x can be used to accurately describe the spatiotemporal heterogeneity of groundwater pollution risk after supplementing sufficient groundwater pollution data from non-sampling points. 1 and x 2 The probability of belonging to each category is used to divide the groundwater pollution risk prevention and control areas;

[0077] It should be noted that the number of samples in each category is small, and new categories of samples will be introduced during testing. How to distinguish the samples of the new category and correctly classify and accurately simulate the samples of each category is the key to solving the problem. In other words, this is actually an open set recognition problem. The main idea of ​​meta-learning is to obtain a better result model by training a small number of samples. Therefore, with the help of meta-learning theory, the process of "expanding the surface from the point" is more objective and precise. The coupled attention mechanism ignores unnecessary information, and performs feature extraction and prototype generation. Based on the prototype network, the preliminary delineation of groundwater pollution prevention and control areas can be achieved. We have confirmed the applicability and effectiveness of this method in the preliminary identification of groundwater pollution zoning, and achieved a classification accuracy of 91.07%;

[0078] It should be noted that the composite model for groundwater pollution risk zoning based on meta-learning can only preliminarily delineate groundwater pollution risk prevention and control areas. The reason is that Figure 3As shown in the figure, the yellow grid unit is the result of the spatiotemporal heterogeneity characterization of groundwater pollution risk, which can be divided into groundwater pollution risk control areas. However, whether the boundary of the control area is the control area or other sub-area and whether the resolution of the division is too large still needs to be discussed, which requires further refinement of the grid unit.

[0079] Among them, further, S102, in one embodiment of the present invention, by calculating the spatial variation coefficient, according to the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the grid unit size, the spatiotemporal heterogeneity grid of groundwater pollution risk is refined, specifically including:

[0080] Set N initial central starting points on the boundary of the preliminarily delineated groundwater pollution risk prevention and control area, calculate the spatial variation coefficients from the N initial central starting points to other points with a distance of h in all directions around them, and then increase the value of h by Δh in sequence, and continue to calculate the spatial variation coefficient;

[0081] It should be noted that the spatial variation coefficient Among them, ω(h) is the spatial variation coefficient from x to x+h, N(h) is the number of sample points with an interval distance of h, and Z(x) is a regional variable, which is the degree of influence of a spatial entity on the surrounding environment;

[0082] Take the minimum value among the calculated multiple spatial variation coefficients, and partition the feature space from the marked point to the corresponding initial center starting point as a separate grid unit until all points in the space are divided into corresponding partitions;

[0083] The root mean square error (RMSE), mean absolute error (MAE) and correlation coefficient (R 2 ) is used as the accuracy index to evaluate each feature space grid unit;

[0084] It should be noted that:

[0085]

[0086] MAE=|x 1 -y 1 |+|x 2 -y 2 |+…+|x n -y n | / N

[0087] R 2 =SSR / SST

[0088] It should be noted that if Figure 4 As shown in the figure, the root mean square error (RMSE), mean absolute error (MAE) and correlation coefficient (R 2) is the accuracy evaluation index. After testing, the evaluation results of the grid units divided by this method (blue column) are better than those of the traditional method (gray column);

[0089] Based on the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the size of the grid unit, the spatiotemporal heterogeneity of groundwater pollution risk is refined based on the requirements of spatial resolution.

[0090] It should be noted that there is an inverse proportional relationship between the grid unit size and the spatial variation coefficient of groundwater pollution. Among them, y is the grid unit size, ω(h) is the spatial variation coefficient of groundwater pollution, and k is the groundwater pollution grid refinement coefficient, which is related to the regional hydrogeological conditions. According to this formula, the spatiotemporal heterogeneity grid refinement of groundwater pollution risk that meets the spatial resolution requirements can be achieved;

[0091] It should be noted that the precise definition of the boundary cannot be completed through a single division. Instead, a rough possible boundary range is first divided through the spatiotemporal heterogeneity of groundwater pollution. We call the "possible boundary range" a "grid unit", and then further characterize it in this "grid unit" to form a new "grid unit". After an infinite number of times of this process, it gradually approaches the real boundary range, that is, the idealized "optimal boundary". This requires relying on multiple multi-scale spatial variation coefficient calculations and partitioning processes to identify spatial variation characteristics in each "grid unit", thereby obtaining a smaller "grid unit", until the "minimum grid unit" is output, such as Figure 5 In actual work, "spatial resolution" can be considered as the "minimum grid unit".

[0092] Among them, further, S103, in one embodiment of the present invention, the groundwater water richness, vulnerability, and source load factors are coupled to accurately define the boundary of the groundwater pollution risk prevention and control area, specifically including:

[0093] With the help of ArcGIS software, spatial analysis methods are applied to couple the spatiotemporal heterogeneity grid of groundwater pollution risk and multiple factor information such as groundwater richness, vulnerability, and source load through a multi-factor superposition model;

[0094] It should be noted that the multi-factor superposition model: In the formula, S is the result of multi-factor superposition, W i is the weight of a single factor, X i Assign scores to factors;

[0095] Comprehensively assess the boundaries of groundwater pollution risk prevention and control areas to achieve accurate demarcation of groundwater pollution risk prevention and control areas;

[0096] It should be noted that a meta-learning-based groundwater pollution zoning composite model is built to preliminarily divide the groundwater pollution risk control zone, effectively fill in high-precision pollution data, and characterize the spatiotemporal heterogeneity of groundwater pollution; by calculating the spatial variation coefficient, the spatiotemporal heterogeneity of groundwater pollution risk grid is refined, and the "minimum grid unit" is targeted; groundwater richness, vulnerability, and source load elements are coupled to accurately define the boundaries of groundwater pollution risk control zones; and drilling verification based on the spatial resolution of the control zone is carried out to improve the accuracy of boundary division from the theoretical effective level (about 60%) to the application level (about 80%), which is more practical. In addition, the accuracy of the groundwater pollution risk control zone boundaries divided by this method is about 30% higher than that of traditional methods, such as Figure 6 shown.

[0097] like Figure 2 As shown, the second aspect of the present invention provides a system for accurately dividing the boundaries of groundwater pollution risk prevention and control areas, the system comprising:

[0098] The first meta-learning data filling module S201 is used to preliminarily delineate groundwater pollution risk prevention and control areas;

[0099] The second coefficient of variation extraction module S202 is used to refine the spatiotemporal heterogeneity grid of groundwater pollution risk;

[0100] The third multi-factor boundary analysis module S203 is used to accurately define the boundary of the groundwater pollution risk prevention and control area;

[0101] The fourth boundary drilling verification module S204 is used to ensure high precision and accuracy in the demarcation of the groundwater pollution risk prevention and control zone boundaries.

[0102] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0103] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for accurately dividing the boundaries of groundwater pollution risk prevention and control areas, characterized in that: The following steps are involved: Build a meta-learning-based composite model for groundwater pollution risk zoning, fill in high-precision data on groundwater pollution distribution in non-sampling areas, and preliminarily delineate groundwater pollution risk prevention and control areas; By calculating the spatial variation coefficient and based on the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the size of the grid unit, the spatiotemporal heterogeneity of groundwater pollution risk can be refined. Couple groundwater richness, vulnerability, and source load factors to accurately define the boundaries of groundwater pollution risk prevention and control areas; The meta-learning-based composite model for groundwater pollution risk zoning is constructed to fill in high-precision data on groundwater pollution distribution in non-sampling areas and preliminarily delineate groundwater pollution risk prevention and control areas. The specific steps are as follows: By selecting and cleaning no less than 10 classic groundwater migration mechanism models, a groundwater migration composite model covering the "source-flow-sink" relationship of 8 typical industries and various typical groundwater hydrogeological characteristics was integrated; Based on the groundwater migration composite model, a prototype network in the meta-learning field is constructed, and the attention mechanism is coupled to build a composite model for groundwater pollution risk zoning based on meta-learning. Integrate multi-year, multi-regional groundwater historical monitoring data, new well sampling data and online monitoring data, conduct exploratory data analysis, establish a spatiotemporal distribution feature set of groundwater pollution risk at different regional scales, hydrogeological conditions and pollution types, and split it into training and validation sets; On the basis of maintaining the accurate simulation of well-fitted data by the composite model of groundwater pollution risk zoning, we targeted underfitted data samples and continuously improved the generalization ability and accuracy of the composite model of groundwater pollution risk zoning by optimizing the simulation effect of underfitted data sets. We also conducted uncertainty analysis on the simulation results of the model based on classification error, data completion error and model error. Based on the background information of the study area and the sampling point data, a composite model based on meta-learning for groundwater pollution risk zoning is applied to supplement the groundwater pollution data of non-sampling points, characterize the spatiotemporal heterogeneity of regional groundwater pollution risks, and achieve the preliminary delineation of groundwater pollution risk prevention and control areas; The method of calculating the spatial variation coefficient and refining the spatiotemporal heterogeneity grid of groundwater pollution risk based on the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the size of the grid unit includes: Set N initial central starting points on the boundary of the preliminarily delineated groundwater pollution risk prevention and control area, calculate the spatial variation coefficients from the N initial central starting points to other points with a distance of h in all directions around them, and then increase the value of h by Δh in sequence, and continue to calculate the spatial variation coefficient; Take the minimum value among the calculated multiple spatial variation coefficients, and partition the feature space from the marked point to the corresponding initial center starting point as a separate grid unit until all points in the space are divided into corresponding partitions; The root mean square error RMSE, mean absolute error MAE and correlation coefficient R 2 Each divided feature space grid unit is evaluated as an accuracy index; According to the quantitative relationship between the spatial variation coefficient of groundwater pollution risk and the size of the grid unit, and based on the requirements of spatial resolution, the spatiotemporal heterogeneity of groundwater pollution risk grid is refined.

2. According to claim 1, a method for accurately dividing the boundaries of groundwater pollution risk prevention and control areas is characterized in that: The coupling of groundwater richness, vulnerability, and source load factors to accurately define the boundaries of groundwater pollution risk prevention and control areas includes: With the help of ArcGIS software, based on spatial analysis methods, multi-factor overlay analysis was performed to couple the spatiotemporal heterogeneity grid of groundwater pollution risk and the multi-factor information of groundwater richness, vulnerability, and source load; Analyze and determine the boundaries of groundwater pollution risk prevention and control areas, and achieve accurate demarcation of the boundaries of groundwater pollution risk prevention and control areas; Carry out drilling verification based on the spatial resolution of the prevention and control zone to ensure high precision and accuracy in the demarcation of the boundaries of the groundwater pollution risk prevention and control zone.

3. A system for accurately dividing the boundaries of groundwater pollution risk prevention and control areas based on the method of claim 1 or 2, characterized in that: The system includes: a first meta-learning data filling module, which is used to preliminarily delineate groundwater pollution risk prevention and control areas; a second coefficient of variation extraction module, which is used to refine the spatiotemporal heterogeneity grid of groundwater pollution risk; a third multi-factor boundary analysis module, which is used to accurately define the boundaries of groundwater pollution risk prevention and control areas; and a fourth boundary drilling verification module, which is used to ensure the high precision and accuracy of the boundary demarcation of groundwater pollution risk prevention and control areas.

Images