Groundwater pollution environmental risk accurate quantification method based on SVM dichotomy model

Abstract

The invention discloses a groundwater pollution environment risk accurate quantification method based on an SVM dichotomy model, and belongs to the technical field of soil-groundwater pollution environment prediction and risk management and control. A statistical substitution model based on numerical simulation and a general SVM dichotomy model are established and optimized for a typical condition framework groundwater environment system, so that the eurytopic groundwater pollution environment risk fine assessment method is further obtained. Reliable prediction or judgment of value domain fine division sections of soil and underground water environment states, evolution processes and related influences in different application scenes can be accurately and efficiently realized on the premise of satisfying a given condition framework; the accuracy, efficiency and practicability of evaluation and pre-judgment of the groundwater pollution environment spatio-temporal evolution state, pollutant spatio-temporal distribution, environmental influence and the like are remarkably improved, and water and soil environment system protection and resource safety utilization are achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of soil-groundwater pollution prediction, environmental assessment and risk control, and specifically relates to a method for accurately quantifying the environmental risk of groundwater pollution based on a SVM binary classification model. Background Art

[0002] At present, in the prediction of groundwater pollution and environmental risk assessment, the application of numerical simulation system to evaluate and quantify groundwater pollution status and environmental risks in specific application scenarios is a common research method. However, for accurate numerical simulation, it is necessary to comprehensively consider and refine many influencing factors and complex evolution processes in specific scenarios, which is time-consuming and highly professional. In addition, there is a large uncertainty in the modeling and prediction of complex medium groundwater systems. The use of Monte Carlo and other methods for uncertainty assessment requires hundreds or thousands of numerical simulation tests. However, the computational cost is huge or even unacceptable, and the efficiency is low. Therefore, there is an urgent need for a practical, efficient, simple and universal method for accurately segmenting the value range of characteristic variables related to the evolution of groundwater pollution environment. Summary of the invention

[0003] The purpose of the present invention is to provide a precise quantification method for groundwater pollution environmental risk based on the SVM binary classification model, so as to solve the technical problems in the prior art of difficulty in predicting and quantifying characteristic variables related to the evolution of groundwater pollution environment, prominent uncertainty problems, high cost, low efficiency and poor real-time performance.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] In the first aspect, the present invention provides a method for accurately quantifying the environmental risk of groundwater pollution based on a SVM binary classification model, comprising:

[0006] Step A: Determine the common feature set of the groundwater pollution environmental system, which is called the system condition framework. Corresponding to the system condition framework, give a conceptual model of the spatiotemporal evolution of groundwater pollution, select the characteristic variables associated with the spatiotemporal evolution of groundwater pollution, obtain the mathematical control equations of the spatiotemporal evolution of groundwater pollution, and obtain the numerical simulation system of the spatiotemporal evolution of groundwater pollution;

[0007] Step B: Determine the input parameter set V = [v 1 ,v 2 ,…,v n ], select the range of each model input parameter variable, and form the model input parameter variable range set U = [u 1 ,u 2 ,…,u n ];

[0008] Step C: Select the change level of each model input parameter variable, and obtain different typical scenarios composed of different model input parameter combinations of the groundwater pollution environment system through orthogonal test method or full combination method of input parameters, which is deterministic scenario sampling;

[0009] Step D: obtaining corresponding simulation values ​​of characteristic variables associated with different typical scenarios based on the numerical simulation system; obtaining a basic statistical modeling data set according to different model input parameter combinations of the different typical scenarios and their corresponding simulation values;

[0010] Step E: based on the given preset number of supplementary scenarios, each model input parameter variable is sampled within its variation range to obtain different supplementary scenarios consisting of different model input parameter combinations of the groundwater environment system; based on the numerical simulation system, corresponding associated characteristic variable simulation values ​​of the different supplementary scenarios are obtained, and the model input parameter combinations of all different supplementary scenarios and their corresponding associated characteristic variable simulation value sets are collected to obtain an enhanced statistical modeling data set;

[0011] Step F: establishing and optimizing a multivariate regression statistical prediction model of the associated characteristic variables and the model input parameter variables based on the basic statistical modeling data set or the basic statistical modeling data set and the enhanced statistical modeling data set;

[0012] Step G: based on the model input parameter variables, a preset number of scenario samples are obtained to obtain a statistical model verification scenario parameter combination, and the model verification scenario simulation value of the associated characteristic variable corresponding to each scenario is obtained through the numerical simulation system, and the model verification scenario parameter combination and the corresponding scenario simulation value are collected as a model verification data set; based on the model verification data set, the associated characteristic variable prediction value is obtained through the multivariate regression statistical prediction model, and the statistical prediction model prediction error value and its distribution characteristics are obtained based on each scenario simulation value and the corresponding statistical prediction model prediction value; according to the prediction error value and its distribution characteristics, the distribution interval of the statistical model prediction error random variable corresponding to the given preset confidence is determined;

[0013] Step H: based on the model input parameter variables, a preset number of scenario samplings are obtained to obtain SVM modeling data reinforcement scenario parameter combinations, and the SVM modeling data reinforcement scenario statistical model prediction values ​​of the associated characteristic variables corresponding to each scenario are calculated according to the statistical prediction model; based on the statistical model prediction values ​​and the prediction error random variable distribution interval corresponding to a given preset confidence, the prediction confidence interval of the statistical prediction model of the associated characteristic variables is determined; based on the model input parameter variables, a preset number of scenario parameter combinations are obtained and the prediction confidence interval is combined to obtain the SVM modeling statistical analysis reinforcement modeling data set;

[0014] Step I: based on the preset threshold value variation interval and the number of preset threshold values ​​for modeling, M1 different thresholds are selected; based on the basic statistical modeling data set or the basic statistical modeling data set and the enhanced statistical modeling data set, the SVM modeling statistical analysis enhanced modeling data set is added, and the model verification data set is superimposed; corresponding to any given environmental threshold, a corresponding output label is obtained, and the environmental threshold is added as a model input parameter variable for SVM classification machine learning modeling, thereby obtaining an SVM binary classification machine learning data set containing threshold parameter variables;

[0015] Step J: Based on the SVM classification machine learning data set containing threshold parameters, the corresponding SVM binary discrimination model containing threshold parameters is obtained based on and applied to the SVM classification model modeling principle and optimization method;

[0016] Step K: Based on the preset threshold value variation interval, M2 thresholds are selected, and an error analysis test set for testing the SVM binary discriminant model is selected from the statistical modeling data set and the model verification data set. Based on the SVM binary discriminant model, the output label changes of each sample in the error analysis test set corresponding to each threshold are obtained, and the upper and lower limits of the predicted value of the associated characteristic variable are determined by the output label changes. According to the true value of the corresponding numerical simulation value, the maximum relative error of the predicted value of the associated characteristic variable of each test sample in the threshold section is calculated, and the maximum relative error average value of the predicted value of the associated characteristic variable of all test samples in the test set is obtained accordingly;

[0017] Step L: Compare whether the average value of the maximum relative error does not exceed the preset value of the average value of the maximum relative error; if the average value of the maximum relative error does not exceed the preset value of the average value of the maximum relative error, then obtain a threshold-containing SVM two-classification model that meets the partition discrimination accuracy and the corresponding required threshold subdivision node number; if the average value of the maximum relative error exceeds the preset value of the average value of the maximum relative error, then adopt an effective method to narrow the gap between the two, and finally meet the condition that the average value of the maximum relative error does not exceed the preset value of the average value of the maximum relative error, that is, the partition discrimination accuracy is achieved, and obtain a threshold-containing SVM two-classification model that meets the partition discrimination accuracy and the corresponding required threshold subdivision node number;

[0018] Step M: For actual application scenarios that meet the system condition framework, based on the actual conditions and collected data of the specific groundwater environmental system, and the number of threshold segment divisions that ultimately meet the accuracy requirements, for each threshold, based on the SVM binary classification discriminant model, obtain the corresponding output label, and determine the upper and lower limits of the predicted value segment based on the label changes.

[0019] Furthermore, in step A, the characteristic variables associated with the spatiotemporal evolution of groundwater pollution are selected, including:

[0020] The associated characteristic variables include: groundwater pollutant concentration, saturated zone pollutant cross-section or interface flux, pollution plume migration and diffusion front position, pollution plume stabilization time, stable pollution plume distribution area, stable pollution plume front distance from the pollution source, stable pollution plume average concentration and pollution plume dissipation time; as well as soil pollutant concentration, vadose zone gas phase pollutant concentration, vadose zone cross-section pollutant flux, vadose zone-saturated zone pollutant interface flux, soil-air interface pollutant flux, rock-air interface pollutant flux and NAPL pollutant spatiotemporal distribution.

[0021] Furthermore, in step A, a set of common characteristics of a type of groundwater pollution environment system to be studied is determined, which is called the system condition framework. Corresponding to the system condition framework, a conceptual model of the spatiotemporal evolution of groundwater pollution is given, including:

[0022] Determine the characteristics of atmospheric precipitation infiltration recharge, which include the presence or absence of atmospheric precipitation infiltration recharge and the spatiotemporal variation characteristics of atmospheric precipitation infiltration recharge intensity;

[0023] Determine the characteristics of the aeration zone, wherein the characteristics of the aeration zone include medium type, lithological structure, medium heterogeneity and anisotropy;

[0024] Determine the characteristics of the saturated zone, wherein the characteristics of the saturated zone include medium type, water-bearing system structure, seepage dimension, flow state, medium heterogeneity and anisotropy characteristics;

[0025] Determine pollution source characteristics, including the phase state of the pollutants of interest, the type of pollutants of interest, the spatial distribution characteristics of the pollution sources, source intensity changes and dynamic characteristics;

[0026] Determine the characteristics of groundwater pollution prevention, control and remediation: the characteristics of groundwater pollution prevention, control and remediation include whether there are groundwater pollution prevention, control and remediation measures, the types of groundwater pollution prevention, control and remediation measures, and the spatiotemporal distribution characteristics of groundwater pollution prevention, control and remediation measures;

[0027] Determine the source-sink characteristics of the pumping and injection wells, wherein the source-sink characteristics of the pumping and injection wells include the presence or absence of pumping and injection wells, the spatial distribution characteristics of the pumping and injection wells, and the dynamic characteristics of the water volume of the pumping and injection wells;

[0028] Determine the interaction characteristics between surface water and groundwater: The interaction characteristics between surface water and groundwater include whether there is interaction between the two, the spatial distribution characteristics of surface water bodies, and the dynamic characteristics of surface water levels;

[0029] Determine the types of groundwater pollutant migration and transformation processes, including advection process, diffusion process, adsorption and desorption process, degradation process, volatilization process, etc.;

[0030] Determining boundary condition characteristics, wherein the boundary condition characteristics include hydrodynamic field boundary condition characteristics and hydrochemical field boundary condition characteristics;

[0031] Initial condition characteristics are determined, wherein the initial condition characteristics include initial condition characteristics of the hydrodynamic field and initial condition characteristics of the hydrochemical field.

[0032] Furthermore, the precise quantification method also includes: processing the associated characteristic variables and the model input parameter variables using any one or any combination of preset data processing methods, and the preset data processing methods include: retaining the value as it is, increasing the value by the same multiple, decreasing the value by the same multiple, logarithmic conversion of the value, unifying the positive and negative values, and normalizing the value.

[0033] Furthermore, in step L, the effective method of narrowing the gap between the two includes reducing the preset value of the average value of the maximum relative error, adjusting the number of threshold subdivision nodes during modeling, adjusting the number of threshold subdivision nodes during model prediction, increasing the statistical analysis and strengthening data samples for SVM binary discriminant model modeling, increasing statistical modeling samples, increasing statistical model prediction error samples, increasing the confidence of the statistical model confidence prediction interval and thus increasing the corresponding confidence prediction interval. The above methods can be used alone or in any combination.

[0034] Based on the above technical solution, the embodiments of the present invention can at least produce the following technical effects:

[0035] The invention provides a method for accurately quantifying the environmental risk of groundwater pollution based on the SVM binary classification model, selects associated characteristic variables related to the spatiotemporal evolution of the groundwater pollution environment, establishes and optimizes a statistical substitution model based on numerical simulation and a general SVM binary classification model for a groundwater environmental system in a typical conditional framework, and further obtains a method for accurately assessing the environmental risk of groundwater pollution with wide adaptability, which can accurately and efficiently realize the reliable prediction or determination of the precise segmentation of the value range of the soil and groundwater environmental status, evolution process and related impacts in different application scenarios under the premise of meeting a given conditional framework, and significantly improves the accuracy, efficiency and practicality of the assessment and prediction of the spatiotemporal evolution state of the groundwater pollution environment, the spatiotemporal distribution of pollutants and their environmental impacts, and other aspects, providing strong support for the protection of the soil and water environment system and the safe utilization of resources, especially for the rapid prediction and early warning of soil and groundwater pollution and reliable risk control. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in this field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0037] In a specific embodiment, polycyclic aromatic hydrocarbons (PAHs), as organic matter widely present in the environment, are highly valued by countries around the world due to their persistence, carcinogenicity and mutagenicity. The carcinogenic concentration of PAHs in groundwater published by the United States is in the range of 0.2-6.9 ng / L, but its corresponding concentration in surface water is 0.1-800 ng / L, most of which are in the range of 2-50 ng / L, and the solubility of PAHs in water is mostly above 1000 ng / L. The polluted surface organic polluted water body continues to infiltrate, and will undergo seepage, adsorption and interception, hydrodynamic dispersion and biodegradation in the aeration zone, and enter the shallow groundwater along the aeration zone soil with relatively weak anti-fouling performance, thereby causing the PAHs concentration in the groundwater to continue to increase. According to the analysis of the solubility and carcinogenic concentration range of PAHs, PAHs in groundwater are prone to exceed the carcinogenic standard. Therefore, it is necessary to pay attention to the migration and transformation of polycyclic aromatic hydrocarbon pollutants in the aeration zone, including quantitative research on the flux changes of polycyclic aromatic hydrocarbon pollutants at the aeration zone-diving interface under typical conditions. This will provide an important basis for the temporal and spatial distribution of relevant organic pollutants in the water and soil environment of the vadose zone under typical conditions and for the risk assessment of groundwater pollution. This embodiment uses limited numerical simulation data to establish an efficient and reliable vadose zone pollutant flux prediction model, demonstrating the important role of the present invention in environmental protection.

[0038] In view of the above problems, the precise quantification method of groundwater pollution environmental risk based on SVM binary classification model is combined. The specific steps are as follows:

[0039] 1. Select the stable flux of pollutants at the interface of the vadose zone and saturated zone as the characteristic variable associated with the groundwater environment, and construct a conceptual model of the spatiotemporal evolution of the corresponding groundwater pollution environmental risk under the framework of typical conditions, that is, a conceptual model of the flux change of the vadose zone-phreatic interface under the framework of layered heterogeneous typical conditions, and further obtain mathematical control equations and numerical models suitable for characterizing the spatiotemporal evolution of the corresponding groundwater pollution environmental risk, and based on this, screen the numerical simulation system that can quantify the spatiotemporal evolution of the groundwater environment-associated characteristic variables.

[0040] Specifically, there are polycyclic aromatic hydrocarbon pollution problems in coking plant sites, petrochemical sites, power plant pollution sites, etc. in my country. When precipitation forms surface water in these sites, the water infiltrates steadily to form a stable seepage field. In this case, the concentration of organic matter in the infiltrating water is high, even close to the solubility, which can be regarded as a constant flow and constant concentration infiltration process. Due to construction excavation or filling, ground compaction or hardening, horizontal drainage and other engineering measures and the natural sedimentary characteristics of the strata, the stratigraphic structure of actual contaminated sites, landfills, etc. usually has horizontal layered heterogeneous characteristics. This study sets the simulation area as a vertical layered heterogeneous profile, and the entire research range is divided into upper and lower layers with different lithologies. The surface polycyclic aromatic hydrocarbon organic pollutant water enters the groundwater through the double-layer heterogeneous vadose zone soil, one-dimensional vertical flow, the upper boundary is a given flow and a given concentration condition, and the stable flux of pollutants at the vadose zone-groundwater interface is used as the associated characteristic variable.

[0041] Specifically, the selected organic pollutants are polycyclic aromatic hydrocarbons represented by benzo[a]pyrene. Such pollutants are semi-volatile pollutants with limited volatilization effect in a relatively closed underground environment, so the volatilization effect of pollutants is not considered in this study.

[0042] Specifically, CHEMFLO-2000 software is used. CHEMFLO-2000 is a numerical simulation system used to simulate water flow and solute transport in the unsaturated zone. The water flow process is described by the Richard equation, see formula (1); and the solute migration and transformation process is characterized by the convection diffusion and adsorption degradation equations, see formula (2). The two are coupled and solved by the finite difference method.

[0043]

[0044] Where: h is the matrix potential; t is the time; x is the distance from the surface downward; K(h) is the hydraulic conductivity coefficient under the matrix potential condition.

[0045] The partial differential governing equation for solute transport considering linear adsorption and first-order degradation has the general form:

[0046]

[0047] In the formula, t is time; θ is soil moisture content; R is the hysteresis coefficient of solute in soil, which is expressed as: ρ=ρ(x) is soil density, k is the distribution coefficient; C is the concentration of solute in water; x is the distance from the surface downward; D is the hydrodynamic diffusion coefficient (including molecular diffusion and mechanical dispersion); q=q(x,t) is the water flow rate; α=α(x) is the first-order degradation coefficient of microorganisms in water.

[0048] The Van Genuchten model, see formula (3), is used to describe the soil moisture characteristic curve, which represents the functional relationship between water content and matrix potential and other hydraulic parameters.

[0049]

[0050] In the formula: θ is soil moisture content; h is matrix potential; θs is saturated moisture content; θr is residual moisture content; α, m, n are VG model empirical coefficients, where m = 1-1 / n.

[0051] 2. Based on the established conceptual model, the numerical model and the numerical simulation system, determine the parameters that affect the spatiotemporal evolution of the selected environmental correlation characteristic variables (see Table 1) and form a parameter set X = [X 1 ,X 2 ,…,X n ], n is the total number of parameters considered, n = 16; considering the variable range of each parameter, the variable interval to be considered is selected (see Table 1), forming the parameter interval set V = [V 1 ,V 2 ,…,V 16 ].

[0052] 3. Select the change level of each parameter (see Table 1 below for the input parameter level and value table), apply the orthogonal experimental design method, and obtain typical scenarios of different parameter combinations of the groundwater environmental system, where the number of orthogonal experimental design scenarios meets the non-overfitting condition of statistical modeling;

[0053] Table 1 Model input parameter levels and values

[0054]

[0055] Note: F represents pollutant flux; p represents the percentage of upper layer thickness; T represents the thickness of the vadose zone; K upper and K lower denote the permeability coefficients of the upper and lower layers, respectively; and represent the saturated water content of the upper and lower layers respectively; and Represent the residual moisture content of the upper and lower layers respectively; α upper and α lower Represent the distribution coefficients of the upper layer and the lower layer respectively; n upper and n lower represent the exponential parameters in the upper and lower layers of the Van Genuchten model, respectively; Q represents the pollutant flow; C represents the pollutant concentration; D represents the dispersion; K d Represents the degradation coefficient.

[0056] 4. For the characteristic variables associated with groundwater pollution environmental risks, the simulated values ​​Y of the characteristic variables associated with groundwater pollution environmental risks in typical scenarios of orthogonal experimental design are obtained based on numerical simulation to form a statistical modeling data set for deterministic scenario sampling.

[0057] 5. Add 20 different parameter combination scenario samplings, apply the numerical simulation system to obtain the corresponding characteristic variable simulation values, and integrate them with the statistical modeling data set of the aforementioned deterministic scenario sampling to form a statistical modeling data set. Take the natural logarithm of the characteristic variable after being divided by 100, and select appropriate data processing methods (no conversion, proportional scaling, logarithmic conversion, etc.) for other model input parameter variables. Establish and optimize the multivariate regression statistical prediction model of Y and X as follows:

[0058]

[0059] Where: F represents the pollutant flux; p represents the proportion of upper layer thickness; T represents the thickness of the vadose zone.

[0060] 6. Further select the number of random sampling 35, generate random numbers of parameters again and form corresponding new scenarios with more parameter combinations. Use the numerical simulation system to obtain the simulation values ​​of the corresponding groundwater pollution environmental risk associated characteristic variables, integrate them to form a model verification data set, and use the established regression statistical prediction model to obtain the prediction error value and its distribution characteristics. Based on this, further determine the distribution interval of the prediction error random variable corresponding to the given confidence level of the groundwater environment associated characteristic variable statistical prediction model (-0.41, 0.40).

[0061] 7. Further select the random sampling number of 10,000, generate random numbers of parameters again and form corresponding new more parameter combination scenarios, and apply the aforementioned multivariate regression statistical prediction model to obtain the corresponding groundwater environment associated characteristic variable prediction value Y S Based on the distribution interval of the prediction error random variable corresponding to the given confidence level of the statistical prediction model of the groundwater environment-related characteristic variables determined above (-0.41, 0.40), the upper and lower limits of the true value of the groundwater environment-related characteristic variables under the given confidence level, namely the prediction confidence interval (Y P1 ,Y P2 ),in:

[0062] Y P1 =Y S / (1+0.4)

[0063] Y P2 =Y S / (1-0.41)

[0064] This random parameter combination scenario and the characteristic variable prediction interval together constitute the statistical analysis enhanced modeling data set.

[0065] 8. The preset value of the maximum relative error average is 0.09; according to the threshold change range, 12 typical thresholds V are selected. C-k (k=1,2,……12)=(-8,-9,…,-19), for each V C-k , obtain its statistical analysis and enhanced modeling data set, and convert each V C-i The corresponding statistical analysis enhanced modeling data set adds a new input parameter V C-k The sets are then combined to form an SVM machine learning training set containing threshold parameters, and the corresponding SVM machine learning test set is obtained based on the model verification data set.

[0066] 9. Based on the SVM classification machine learning data set containing threshold parameters, and based on and applying the SVM classification model modeling principle and optimization method, obtain the corresponding SVM binary classification discrimination model containing threshold parameters;

[0067] 10. Based on the preset threshold value variation range, first select 12 thresholds.

[0068] 11. The model validation data set is used as the error analysis test set for the SVM binary classification discriminant model. Based on the SVM binary classification discriminant model, the output label changes corresponding to the corresponding thresholds of each sample in the error analysis test set are obtained. The upper and lower limits of the predicted values ​​of the associated characteristic variables are determined by the output label changes. According to the true value of the corresponding numerical simulation value, the maximum relative error of the threshold section of the predicted value of the associated characteristic variable of each test sample is calculated. The average value of the maximum relative error of the predicted values ​​of the associated characteristic variables of all test samples in the test set is 0.124, which does not meet the preset accuracy requirements (see Table 2).

[0069] 12. Based on the preset threshold value variation range, 23 thresholds are further selected and step 11 is repeated. The maximum relative error average is 0.102, which does not meet the preset accuracy requirement (see Table 2).

[0070] 13. Based on the preset threshold value variation range, 23 thresholds are further selected, and steps 8 to 9 and step 11 are repeated. The maximum relative error average is 0.096, which still does not meet the preset accuracy requirement (see Table 2).

[0071] 14. Based on the preset threshold value variation range, 46 thresholds are further selected and step 11 is repeated. The maximum relative error average is 0.084, which meets the preset accuracy requirement (see Table 2).

[0072] Table 2 Analysis of the statistical results of the corresponding errors of SVM modeling with different threshold values ​​and model threshold judgment points

[0073]

[0074] Note:

[0075] Error analysis scenario 1: SVM binary classification model with threshold parameter was built using data from 12 different thresholds (from -8 to -19), and 12 node thresholds were selected for error analysis.

[0076] Error analysis scenario 2: SVM binary classification model with threshold parameter was built using data from 12 different thresholds (from -8 to -19), and 23 node thresholds were selected for error analysis.

[0077] Error analysis scenario 3: SVM binary classification model with threshold parameter was built using data from 23 different thresholds (from -8 to -19), and 23 node thresholds were selected for error analysis.

[0078] Error analysis scenario 3: SVM binary classification model with threshold parameter was built using data from 23 different thresholds (from -8 to -19), and 46 node thresholds were selected for error analysis.

[0079] 15. There is an actual application scenario that meets the system condition framework. According to the actual conditions and collected data of the specific groundwater environmental system, and the number of threshold segment divisions that ultimately meet the accuracy requirements is 46, for each threshold, the corresponding output label is obtained based on the SVM two-classification discrimination model. According to the label change, the upper and lower limits of the predicted value segment are determined to be -15.25 and -15.5 respectively.

[0080] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0081] It should also be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0082] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements fall within the scope of the present invention to be protected. The scope of the present invention to be protected is defined by the attached claims and their equivalents.

Claims

1. A precise quantification method for groundwater pollution environmental risk based on the SVM binary classification model, characterized by: include: Step A: Determine the common feature set of the groundwater pollution environmental system, which is called the system condition framework. Corresponding to the system condition framework, give a conceptual model of the spatiotemporal evolution of groundwater pollution, select the characteristic variables associated with the spatiotemporal evolution of groundwater pollution, obtain the mathematical control equations of the spatiotemporal evolution of groundwater pollution, and obtain the numerical simulation system of the spatiotemporal evolution of groundwater pollution; Step B: Determine the input parameter set V = [v1, v2, …, v n ], select the range of each model input parameter variable to form the model input parameter variable range set U = [u1, u2, ..., u n ]; Step C: Select the change level of each model input parameter variable, and obtain different typical scenarios composed of different model input parameter combinations of the groundwater pollution environment system through orthogonal test method or full combination method of input parameters, which is deterministic scenario sampling; Step D: obtaining corresponding simulation values ​​of characteristic variables associated with different typical scenarios based on the numerical simulation system; obtaining a basic statistical modeling data set according to different model input parameter combinations of the different typical scenarios and their corresponding simulation values; Step E: based on the given preset number of supplementary scenarios, each model input parameter variable is sampled within its variation range to obtain different supplementary scenarios consisting of different model input parameter combinations of the groundwater environment system; based on the numerical simulation system, corresponding associated characteristic variable simulation values ​​of the different supplementary scenarios are obtained, and the model input parameter combinations of all different supplementary scenarios and their corresponding associated characteristic variable simulation value sets are collected to obtain an enhanced statistical modeling data set; Step F: establishing and optimizing a multivariate regression statistical prediction model of the associated characteristic variables and the model input parameter variables based on the basic statistical modeling data set or the basic statistical modeling data set and the enhanced statistical modeling data set; Step G: based on the model input parameter variables, a preset number of scenario samples are obtained to obtain a statistical model verification scenario parameter combination, and the model verification scenario simulation value of the associated characteristic variable corresponding to each scenario is obtained through the numerical simulation system, and the model verification scenario parameter combination and the corresponding scenario simulation value are collected as a model verification data set; based on the model verification data set, the associated characteristic variable prediction value is obtained through the multivariate regression statistical prediction model, and the statistical prediction model prediction error value and its distribution characteristics are obtained based on each scenario simulation value and the corresponding statistical prediction model prediction value; according to the prediction error value and its distribution characteristics, the distribution interval of the statistical model prediction error random variable corresponding to the given preset confidence is determined; Step H: obtaining a preset number of scenario samples based on the model input parameter variables, obtaining a combination of SVM modeling data reinforcement scenario parameters, and calculating the SVM modeling data reinforcement scenario statistical model prediction value of the associated characteristic variables corresponding to each scenario according to the statistical prediction model; determining the prediction confidence interval of the associated characteristic variable statistical prediction model according to the statistical model prediction value and the prediction error random variable distribution interval corresponding to a given preset confidence; Based on the model input parameter variables, a preset number of scenario parameter combinations and the prediction confidence interval are obtained and then combined to obtain an SVM modeling statistical analysis enhanced modeling data set; Step I: based on the preset threshold value variation interval and the number of preset threshold values ​​for modeling, M1 different thresholds are selected; based on the basic statistical modeling data set or the basic statistical modeling data set and the enhanced statistical modeling data set, the SVM modeling statistical analysis enhanced modeling data set is added, and the model verification data set is superimposed; corresponding to any given environmental threshold, a corresponding output label is obtained, and the environmental threshold is added as a model input parameter variable for SVM classification machine learning modeling, thereby obtaining an SVM binary classification machine learning data set containing threshold parameter variables; Step J: Based on the SVM classification machine learning data set containing threshold parameters, the corresponding SVM binary discrimination model containing threshold parameters is obtained based on and applied to the SVM classification model modeling principle and optimization method; Step K: Based on the preset threshold value variation interval, M2 thresholds are selected, and an error analysis test set for testing the SVM binary discriminant model is selected from the statistical modeling data set and the model verification data set. Based on the SVM binary discriminant model, the output label changes of each sample in the error analysis test set corresponding to each threshold are obtained, and the upper and lower limits of the predicted value of the associated characteristic variable are determined by the output label changes. According to the true value of the corresponding numerical simulation value, the maximum relative error of the predicted value of the associated characteristic variable of each test sample in the threshold section is calculated, and the maximum relative error average value of the predicted value of the associated characteristic variable of all test samples in the test set is obtained accordingly; Step L: Compare whether the average value of the maximum relative error does not exceed the preset value of the average value of the maximum relative error; if the average value of the maximum relative error does not exceed the preset value of the average value of the maximum relative error, then obtain a threshold-containing SVM two-classification model that meets the partition discrimination accuracy and the corresponding required threshold subdivision node number; if the average value of the maximum relative error exceeds the preset value of the average value of the maximum relative error, then adopt an effective method to narrow the gap between the two, and finally meet the condition that the average value of the maximum relative error does not exceed the preset value of the average value of the maximum relative error, that is, the partition discrimination accuracy is achieved, and obtain a threshold-containing SVM two-classification model that meets the partition discrimination accuracy and the corresponding required threshold subdivision node number; Step M: For actual application scenarios that meet the system condition framework, based on the actual conditions and collected data of the specific groundwater environmental system, and the number of threshold segment divisions that ultimately meet the accuracy requirements, for each threshold, based on the SVM binary classification discriminant model, obtain the corresponding output label, and determine the upper and lower limits of the predicted value segment based on the label changes.

2. The method according to claim 1, characterized in that In step A, the common feature set of the groundwater pollution environmental system is determined, specifically including: determining the atmospheric precipitation infiltration and recharge characteristics, determining the vadose zone characteristics, determining the saturated zone characteristics, determining the pollution source characteristics, determining the groundwater pollution prevention, control and remediation characteristics, determining the source and sink characteristics of the injection wells, determining the interaction characteristics between surface water and groundwater, determining the type of groundwater pollutant migration and transformation process, determining the boundary condition characteristics and determining the initial condition characteristics.

3. The method according to claim 1, characterized in that In step A, the associated characteristic variables include: groundwater pollutant concentration, saturated zone pollutant cross-section or interface flux, pollution plume migration and diffusion front position, pollution plume stabilization time, stable pollution plume distribution area, stable pollution plume front distance from the pollution source, stable pollution plume average concentration and pollution plume dissipation time; as well as soil pollutant concentration, vadose zone gas phase pollutant concentration, vadose zone cross-section pollutant flux, vadose zone-saturated zone pollutant interface flux, soil-air interface pollutant flux, rock-air interface pollutant flux and NAPL pollutant spatiotemporal distribution.

4. The method according to claim 1, characterized in that: In step H, obtaining a preset number of scenario samples based on the model input parameters specifically includes: obtaining a preset number of random scenario samples based on random values ​​of the model input parameters according to the variation range of the model input parameters, and deterministic scenario samples including selecting the variation level of each model input parameter and then combining the model input parameters.

5. The method according to claim 1, characterized in that In step L, the effective method of narrowing the gap between the two includes reducing the preset value of the average value of the maximum relative error, adjusting the number of threshold subdivision nodes during modeling, adjusting the number of threshold subdivision nodes during model prediction, increasing the statistical analysis and strengthening data samples for SVM binary discriminant model modeling, increasing statistical modeling samples, increasing statistical model prediction error samples, increasing the confidence of the statistical model confidence prediction interval and thus increasing the corresponding confidence prediction interval. The above methods can be used alone or in any combination.

Images