KR102151727B1 - Device and method for assessing groundwater contamination vulnerability using adaptive neuro-fuzzy inference system with differential evolution - Google Patents

Abstract

The present invention relates to a device and a method for groundwater contamination vulnerability assessment based on an adaptive neuro-fuzzy inference system combined with a differential evolution technique. According to the present invention, a groundwater contamination vulnerability factor weight is calculated with a differential evolution algorithm, a groundwater contamination vulnerability index calculated by multiplying the groundwater contamination vulnerability factor weight by groundwater contamination vulnerability factor data is input to an input node of an adaptive neuro-fuzzy inference system (ANFIS), a nitrate nitrogen concentration is input to an output node, and machine learning is performed. An ANFIS model is generated as a result, a groundwater contamination vulnerability factor is input to an input node of the generated ANFIS model, and a predicted nitrate nitrogen concentration value is calculated as a result.

Description

Groundwater pollution vulnerability assessment device and method based on an adaptive neuro-fuzzy inference technique combining differential evolution techniques {DEVICE AND METHOD FOR ASSESSING GROUNDWATER CONTAMINATION VULNERABILITY USING ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM WITH DIFFERENTIAL EVOLUTION}

The present invention relates to an apparatus and method for evaluating groundwater pollution vulnerability based on an adaptive neuro-fuzzy inference technique combined with a differential evolution technique.In particular, a weight of a groundwater pollution vulnerability factor is calculated using a differential evolution (DE) algorithm. And, the groundwater pollution vulnerability index calculated by multiplying the groundwater pollution vulnerability factor data by the weight of the groundwater pollution vulnerability factor is input to the input node of the adaptive neuro-fuzzy inference system (ANFIS) network. An adaptive combination of differential evolution technique that generates an ANFIS model by inputting acid nitrogen concentration to the output node and machine learning, and calculating the predicted value of nitrate nitrogen concentration by inputting a groundwater pollution vulnerability factor to the input node of the generated ANFIS model. An apparatus and method for evaluating groundwater contamination vulnerability based on neuro-fuzzy inference techniques.

In general, for the evaluation of groundwater pollution vulnerability, a drag technique is used. This technique is a method of evaluating the pollution vulnerability by assigning a grade according to a range of drag factors to evaluate the pollution vulnerability of groundwater, and has the advantage that the evaluation is simple and easy to understand.

In Korean Patent Registration No. 2006826, land use, which can improve the performance of groundwater pollution vulnerability assessment by applying the evaluation criteria according to the land use of the evaluation area to the method of using the dynamic model used in the evaluation of groundwater pollution vulnerability. An apparatus and method for evaluating the vulnerability of groundwater pollution in consideration of

However, such a conventional groundwater pollution vulnerability assessment device basically uses a method of using a drag model, and this method of using a drag model has a lack of precision because it evaluates pollution vulnerability by subjectively judging the grade and weight of the evaluation factors. There was a problem.

Korean Patent Registration No. 2006826 Publication (Registration Date: 2019.07.29)

Therefore, the present invention has been made to solve the above problems, and an object of the present invention is to evaluate groundwater contamination vulnerability based on an adaptive neuro-fuzzy inference technique combining differential evolution techniques that can increase the precision of groundwater contamination vulnerability assessment It is to provide an apparatus and a method.

In order to achieve the above object, the apparatus for evaluating groundwater contamination vulnerability based on an adaptive neuro-purge inference technique combining a differential evolution technique according to an embodiment of the present invention includes a plurality of groundwater contamination vulnerability factor data and a plurality of nitrate nitrogen concentrations. A data input unit configured to input data; A vulnerability factor selection unit configured to select at least one groundwater contamination vulnerability factor data from among the plurality of groundwater contamination vulnerability factor data using a Variance Inflation Factor (VIF) and a tolerance limit (TL); A vulnerability factor weight calculation unit configured to calculate a groundwater contamination vulnerability factor weight by using a differential evolution (DE) algorithm based on the selected groundwater contamination vulnerability factor data and nitrate nitrogen concentration data; The selected groundwater pollution vulnerability factor data is multiplied by the weight of the groundwater pollution vulnerability factor to calculate a groundwater pollution vulnerability index, and the calculated groundwater pollution vulnerability index is an adaptive neuro-fuzzy inference system (ANFIS) An adaptive neuro-fuzzy inference system model generator configured to generate an ANFIS model by machine learning by inputting the nitrate nitrogen concentration to the output node and inputting it to the input node of the network; And a vulnerability value predicting unit configured to calculate a predicted value of nitrate nitrogen concentration by inputting a groundwater pollution vulnerability factor to an input node of the generated ANFIS model.

In the groundwater pollution vulnerability evaluation apparatus based on an adaptive neuro-purge inference method combining the differential evolution method, the plurality of groundwater pollution vulnerability factor data include groundwater surface depth, groundwater recharge amount, aquifer medium, soil medium, topographic slope, barrel It may include the impact of vadose zone, hydraulic conductivity and land use.

In order to achieve the above object, the groundwater pollution vulnerability evaluation method based on the adaptive neuro-fuzzy inference method combining the differential evolution method according to another embodiment of the present invention includes a plurality of groundwater pollution vulnerability factor data and a plurality of data by a data input unit. Inputting the nitrate nitrogen concentration data; Selecting one or more groundwater contamination vulnerability factor data from among the plurality of groundwater contamination vulnerability factor data using a Variance Inflation Factor and a Tolerance Limit by a vulnerability factor selection unit; Calculating a groundwater pollution vulnerability factor weight using a differential evolution (DE) algorithm based on the groundwater pollution vulnerability factor data and the nitrate nitrogen concentration data selected by the vulnerability factor weight calculator; Calculating a groundwater contamination vulnerability index by multiplying the groundwater contamination vulnerability factor data selected by an adaptive neuro-fuzzy inference system model generation unit by a weight of the groundwater contamination vulnerability factor; The groundwater pollution vulnerability index calculated by the adaptive neuro-fuzzy inference system model generation unit is input to an input node of an adaptive neuro-fuzzy inference system (ANFIS) network and the nitrate nitrogen concentration Generating an ANFIS model by inputting to the output node and performing machine learning; And calculating a predicted value of nitrate nitrogen concentration by inputting a groundwater pollution vulnerability factor to an input node of the ANFIS model in which the vulnerability value predicting unit is generated.

According to an apparatus and method for evaluating groundwater contamination vulnerability based on an adaptive neuro-purge inference technique combined with a differential evolution technique according to an embodiment of the present invention, a plurality of groundwater contamination vulnerability factor data and a plurality of nitrate nitrogen concentration data are input. Become; Selecting at least one groundwater contamination vulnerability factor data from among the plurality of groundwater contamination vulnerability factor data using a Variance Inflation Factor and a tolerance limit; Calculating a weight of a groundwater contamination vulnerability factor using a differential evolution (DE) algorithm based on the selected groundwater contamination vulnerability factor data and nitrate nitrogen concentration data; The selected groundwater pollution vulnerability factor data is multiplied by the weight of the groundwater pollution vulnerability factor to calculate a groundwater pollution vulnerability index, and the calculated groundwater pollution vulnerability index is an adaptive neuro-fuzzy inference system (ANFIS) Generating an ANFIS model by machine learning by inputting the nitrate nitrogen concentration to the output node as well as inputting it to the input node of the network; Inputting a groundwater contamination vulnerability factor to an input node of the generated ANFIS model to calculate a nitrate nitrogen concentration prediction value; By being configured, there is an excellent effect that it is possible to increase the precision of the groundwater contamination vulnerability assessment.

That is, according to an apparatus and method for evaluating groundwater pollution vulnerability based on an adaptive neuro-fuzzy inference technique combined with a differential evolution technique according to an embodiment of the present invention, the weight of the groundwater pollution vulnerability factor is objectively calculated using a differential evolution algorithm. Then, the groundwater vulnerability index obtained by multiplying the calculated weight of the groundwater pollution vulnerability factor by the data on the groundwater pollution vulnerability factor is applied to the adaptive neuro-fuzzy inference method network, which is highly efficient among artificial intelligence techniques, to evaluate the vulnerability of groundwater pollution. By being configured, it is possible to increase the precision of the groundwater contamination vulnerability assessment.

1 is a block diagram of an apparatus for evaluating groundwater contamination vulnerability based on an adaptive neuro-fuzzy inference technique incorporating a differential evolution technique according to an embodiment of the present invention.
FIG. 2 is a flowchart of a method for evaluating groundwater contamination vulnerability using an apparatus for evaluating groundwater contamination vulnerability based on an adaptive neuro-fuzzy inference technique combined with a differential evolution technique according to an embodiment of the present invention.
3 is a diagram showing groundwater contamination vulnerability maps by nitrate nitrogen, (a) shows the original nitrate nitrogen concentration distribution map, and (b) is groundwater contamination by the original adaptive neuro-purge inference technique (ANFIS) A vulnerability map is shown, and (C) shows a groundwater pollution vulnerability map by an adaptive neuro-fuzzy inference technique (ANFIS-DE) combining the differential evolution technique according to an embodiment of the present invention.
Figure 4 is a diagram showing the response characteristic curve (ROC) derived in the learning process and the verification process of the original ANFIS technique and the ANFIS-DE technique of the present invention, (a) shows the ROC in the learning process, (b) Represents the ROC during the assay.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

1 is a block diagram of an apparatus for evaluating groundwater contamination vulnerability based on an adaptive neuro-fuzzy inference technique combining a differential evolution technique according to an embodiment of the present invention.

An apparatus for evaluating groundwater contamination vulnerability based on an adaptive neuro-fuzzy inference technique combined with a differential evolution technique according to an embodiment of the present invention includes a data input unit 100 and a vulnerability factor selection unit 200, as shown in FIG. 1. ), a vulnerability factor weight calculation unit 300, an adaptive neuro-fuzzy inference system model generation unit 400, and a vulnerability value prediction unit 500.

The data input unit 100 serves to input a plurality of groundwater contamination vulnerability factor data and a plurality of nitrate nitrogen concentration data into the vulnerability factor selection unit 200.

A plurality of nitrate nitrogen concentration data may be collected, for example, from 95 monitoring wells located in Miryang-si, Gyeongsangnam-do. The depth of the well can be around 30m. Some wells have reached bedrock, but most are in shallow aquifers of alluvial sediments. The nitrate nitrogen concentration of 10 mg/L is the critical value for drinking water. The nitrate nitrogen concentration data were naturally log-transformed to ensure the continuity of the training and testing process of the original ANFIS and ANFIS-DE methods, and inversely transformed by exponentiation after processing.

Data on multiple groundwater pollution vulnerability factors include groundwater depth, groundwater recharge, aquifer media, soil media, topographic gradient, impact of vadose zone, hydraulic conductivity, and land use. Data on multiple groundwater pollution vulnerability factors can take the grade value suggested by (Aller et al., 1987b), for example.

Groundwater depth is the distance from the surface to the groundwater level.

Groundwater recharge refers to the amount of water that reaches the groundwater level through infiltration of unsaturated areas. Groundwater recharge is an essential source of groundwater replenishment in the study area. The amount of groundwater recharge can be determined using the Piscopo method.

Aquifer media can be classified according to geology at a depth of 30 m, for example, in 95 well well data.

The soil medium can be classified by, for example, a soil map of Miryang City issued by the National Institute of Agricultural Science.

The topographic slope can be classified by ArcGIS 10.2 using, for example, a digital elevation map of Miryang City.

The impact of vadose zone can be determined, for example, from the aeration zone geological data of 95 well wells.

The hydraulic conductivity data can be obtained, for example, from a water permeability test or a slug test.

Land use was created through map classification of Landsat images, and land uses can be identified and classified according to distribution in the research area.

The vulnerability factor selection unit 200 selects at least one groundwater contamination vulnerability factor data from among the eight plurality of groundwater contamination vulnerability factor data above using a Variance Inflation Factor and a tolerance limit. It serves to input to the weight calculation unit 300 and the adaptive neuro-fuzzy inference system (ANFIS) model generation unit 400.

The vulnerability factor weight calculation unit 300 uses a differential evolution (DE) algorithm based on the groundwater contamination vulnerability factor data selected by the vulnerability factor selection unit 200 and the corresponding nitrate nitrogen concentration data. It serves to calculate the vulnerability factor weight.

The DE algorithm does not use differential information and uses a self-organized vector-based search technique. The DE algorithm is characterized by its simplicity, robustness, fast convergence speed, and few control variables. Therefore, this algorithm is suitable for solving the non-differentiable problem and nonlinear optimization problem. The DE algorithm is a technique that requires a large number of population solutions to construct a population, and each solution constructs a certain number of parameters according to the dimension of the problem. The basic concept of the DE algorithm lies in how to create a variable solution that decides to choose the best variable (parent).

The strategy applied to the DE algorithm is to generate a new solution using the difference between two vectors randomly selected from the population. For each solution, a trial solution is generated from the original population by performing mutation, mating, and selection processes. The existing and new solutions are compared, and the optimal solution is delivered to the next generation. For this, an initial population consisting of as many vectors or solutions as the number of populations is randomly generated between the two boundaries of the D-dimensional space of the problem. The notation for each solution X _i in all generations g can be expressed as [Equation 1]

[Equation 1]

는 D 차원 공간의 모집단의 개체수를 나타내며,

이다] [here,

Represents the population of the population in the D-dimensional space,

to be]

), 공여자 벡터 또는 돌연변이 벡터(

)는 다음 [수학식 2]를 사용하여 생성된다.In the mutation stage, a single solution of the population in all generations (

), donor vector or mutant vector (

) Is generated using the following [Equation 2].

[Equation 2]

는 최상의 적합 함수를 달성하는 솔루션이고, F는 수렴 속도를 제어하는 돌연변이 인자이며, [1, 0] 사이의 범위 값이고,

는 현재 세대에서 선택한 무작위 솔루션 벡터이다][here,

Is the solution to achieve the best fit function, F is the mutant factor controlling the rate of convergence, and is a value ranging between [1, 0],

Is the random solution vector chosen by the current generation]

는 원 모집단의 돌연변이 벡터(

) 및 표적 벡터(

)에 의해 생성된다. 이 과정은 다음의 [수학식 3]과 같이 표현될 수 있다.In the mating stage, the test vector

Is the mutant vector of the original population (

) And target vector (

). This process can be expressed as the following [Equation 3].

[Equation 3]

(시험 벡터)와 원 모집단

(표적 벡터)의 비교에 기초하여 차세대 벡터

로 전달된다. 이 프로세스는 다음의 [수학식 4]와 같이 표현될 수 있다.In the selection phase, the new population is in the mating phase

(Test vector) and original population

Next-generation vector based on comparison of (target vector)

Is delivered to. This process can be expressed as the following [Equation 4].

[Equation 4]

The adaptive neuro-fuzzy inference model generation unit 400 multiplies the groundwater contamination vulnerability factor data selected by the vulnerability factor selection unit 200 by the groundwater contamination vulnerability factor weight calculated by the vulnerability factor weight calculation unit 300 to groundwater. Calculate the pollution vulnerability index, and input the calculated groundwater pollution vulnerability index into the input node of the adaptive neuro-fuzzy inference system (ANFIS), and nitrate nitrogen corresponding to the groundwater pollution vulnerability factor. It plays a role of creating an ANFIS model by inputting the concentration to the output node of the adaptive neuro-fuzzy inference system (ANFIS) and performing machine learning.

The vulnerability value predictor 500 serves to calculate a nitrate nitrogen concentration predicted value by inputting a groundwater contamination vulnerability factor to an input node of the ANFIS model generated by the adaptive neuro-fuzzy inference model generating unit 400.

Hereinafter, a method for evaluating groundwater pollution vulnerability using a groundwater pollution vulnerability evaluation apparatus based on an adaptive neuro-fuzzy inference method combining the differential evolution method according to an embodiment of the present invention configured as described above will be described.

2 is a flowchart of a groundwater contamination vulnerability assessment method using a groundwater contamination vulnerability assessment apparatus based on an adaptive neuro-fuzzy inference technique combined with a differential evolution technique according to an embodiment of the present invention, where S is a step Means.

First, when a plurality of groundwater contamination vulnerability factor data and a plurality of nitrate nitrogen concentration data are inputted by the data input unit 100 (S100), the vulnerability factor selection unit 200 performs a Variance Inflation Factor and a tolerance limit. At least one groundwater contamination vulnerability factor data is selected from among a plurality of groundwater contamination vulnerability factor data using (Tolerance Limit) (S200).

Next, the vulnerability factor weight calculation unit 300 uses a differential evolution (DE) algorithm based on the groundwater contamination vulnerability factor data selected in step S200 and the corresponding nitrate nitrogen concentration data. The weight is calculated (S300).

Subsequently, the adaptive neuro-fuzzy inference system model generation unit 400 multiplies the groundwater pollution vulnerability factor data selected in step S200 by the weight of the groundwater pollution vulnerability factor calculated in step S300 to calculate the groundwater pollution vulnerability index. (S400).

Next, the adaptive neuro-fuzzy inference system model generation unit 400 applies the groundwater pollution vulnerability index calculated in step S400 to the input node of the adaptive neuro-fuzzy inference system (ANFIS) network. In addition to inputting, the ANFIS model is generated by machine learning by inputting the nitrate nitrogen concentration to the output node (S410).

Next, the vulnerability value predictor 500 calculates a nitrate nitrogen concentration predicted value by inputting a groundwater pollution vulnerability factor to the input node of the ANFIS model generated in step S410 (S500).

On the other hand, a nitrate nitrogen map (target: Miryang-si, Gyeongsangnam-do), which is a standard for evaluating groundwater pollution vulnerability, was generated using raw nitrate nitrogen data, original ANFIS model, and ANFIS-DE model, as shown in FIG. .

The original nitrate nitrogen map (Fig. 3a) is similar to the nitrate nitrogen map of the original ANFIS model (Fig. 3b) and the nitrate nitrogen map of the ANFIS-DE model of the present invention (Fig. 3c). The nitrate nitrogen map of the original ANFIS model (Fig. 3b) is similar to the original nitrate nitrogen map and the nitrate nitrogen map of the ANFIS-DE model (Fig. 3c), but is different near the Nakdong River in the southern region of Miryang City. The very high vulnerability index of the original ANFIS model near the Nakdong River is much higher than that of the ANFIS-DE model. This fact indicates that the ANFIS-DE model of the present invention is very similar to the original nitrate nitrogen map and is far superior to the original ANFIS model.

Meanwhile, Table 1 below shows the results of comparing the error and correlation between the nitrate nitrogen values calculated by the original ANFIS method and the ANFIS-DE method of the present invention and the original nitrate nitrogen value.

As a result of comparing the error values, the ANFIS-DE method of the present invention has a smaller absolute error mean value (MAE) and the square root value of the square error mean value (RMSE) than the original ANFIS method, and the correlation coefficient is the ANFIS-DE method of the present invention. Since it appears larger than the ANFIS technique, it can be seen that the ANFIS-DE technique of the present invention yields better results.

[Table 1]

On the other hand, Figure 4 is a diagram showing the response characteristic curve (ROC) derived in the learning process and the verification process of the original ANFIS technique and the ANFIS-DE technique of the present invention, (a) shows the ROC in the learning process, ( b) represents the ROC during the assay.

ROC is determined by the following [Equation 5] and [Equation 6].

[Equation 5]

[Here, TPR (True Positive Rate) represents the ratio of the true amount, TP represents the number of cases where groundwater contamination is predicted, and FN represents the number of cases where groundwater contamination is not predicted. Indicate the number of observed cases]

[Equation 6]

[Here, FPR (False Positive Ratio) represents the ratio of false positives, FP represents the number of cases where groundwater contamination is not observed where groundwater contamination is predicted, and TN is groundwater contamination where groundwater contamination is not predicted. Represents the number of unobserved cases]

After calculating the TPR and FPR, the area under the curve created by displaying the FPR value on the X-axis and the TPR value on the Y-axis is called AUC (Area Under Curve), and the technique that makes the area larger is better than the technique with less. I can.

As shown in Figure 4, the AUC value in the learning process and the verification process of the original ANFIS technique is (0.879, 0.643), and the AUC value in the learning process and the verification process of the ANFIS-DE technique of the present invention is (0.975, 0.857). Therefore, since the values of the ANFIS-DE technique of the present invention are all larger than that of the original ANFIS technique, the ANFIS-DE technique of the present invention yielded better results than the original ANFIS technique. This means that the nitrate nitrogen value predicted by the ANFIS-DE technique of the present invention has little error from the original nitrate nitrogen value.

According to an apparatus and method for evaluating groundwater contamination vulnerability based on an adaptive neuro-purge inference technique combined with a differential evolution technique according to an embodiment of the present invention, a plurality of groundwater contamination vulnerability factor data and a plurality of nitrate nitrogen concentration data are input. Become; Selecting at least one groundwater contamination vulnerability factor data from among the plurality of groundwater contamination vulnerability factor data using a Variance Inflation Factor and a tolerance limit; Calculating a weight of a groundwater contamination vulnerability factor using a differential evolution (DE) algorithm based on the selected groundwater contamination vulnerability factor data and nitrate nitrogen concentration data; The selected groundwater pollution vulnerability factor data is multiplied by the weight of the groundwater pollution vulnerability factor to calculate a groundwater pollution vulnerability index, and the calculated groundwater pollution vulnerability index is an adaptive neuro-fuzzy inference system (ANFIS) Generating an ANFIS model by machine learning by inputting the nitrate nitrogen concentration to the output node as well as inputting it to the input node of the network; Inputting a groundwater contamination vulnerability factor to an input node of the generated ANFIS model to calculate a nitrate nitrogen concentration prediction value; By being configured, it is possible to increase the precision of the groundwater contamination vulnerability assessment.

That is, according to an apparatus and method for evaluating groundwater pollution vulnerability based on an adaptive neuro-fuzzy inference technique combined with a differential evolution technique according to an embodiment of the present invention, the weight of the groundwater pollution vulnerability factor is objectively calculated using a differential evolution algorithm. Then, the groundwater vulnerability index obtained by multiplying the calculated weight of the groundwater pollution vulnerability factor by the data on the groundwater pollution vulnerability factor is applied to the adaptive neuro-fuzzy inference method network, which is highly efficient among artificial intelligence techniques, to evaluate the vulnerability of groundwater pollution. By being configured, it is possible to increase the precision of the groundwater contamination vulnerability assessment.

In the drawings and specification, the best embodiments have been disclosed, and specific terms are used, but these are only used for the purpose of describing the embodiments of the present invention, and are used to limit the meaning or the scope of the present invention described in the claims. Was not done. Therefore, those of ordinary skill in the art will appreciate that various modifications and equivalent other embodiments are possible therefrom. Therefore, the true technical scope of the present invention should be determined by the technical spirit of the appended claims.

100: data input unit
200: vulnerability factor selection unit
300: vulnerability factor weight calculation unit
400: Adaptive Neuro-Fuzzy Inference System (ANFIS) model generator
500: vulnerability value prediction unit

Claims (4)

A data input unit configured to input a plurality of groundwater pollution vulnerability factor data and a plurality of nitrate nitrogen concentration data;
A vulnerability factor selection unit configured to select at least one groundwater contamination vulnerability factor data from among the plurality of groundwater contamination vulnerability factor data using a Variance Inflation Factor and a tolerance limit;
A vulnerability factor weight calculation unit configured to calculate a groundwater contamination vulnerability factor weight by using a differential evolution (DE) algorithm based on the selected groundwater contamination vulnerability factor data and nitrate nitrogen concentration data;
The selected groundwater pollution vulnerability factor data is multiplied by the weight of the groundwater pollution vulnerability factor to calculate a groundwater pollution vulnerability index, and the calculated groundwater pollution vulnerability index is an adaptive neuro-fuzzy inference system (ANFIS) An adaptive neuro-fuzzy inference system model generation unit configured to generate an ANFIS model by machine learning by inputting the nitrate nitrogen concentration to the output node and inputting it to the input node of the network; And
Including; a vulnerability value prediction unit configured to calculate a nitrate nitrogen concentration predicted value by inputting a groundwater pollution vulnerability factor to the generated input node of the ANFIS model,
After inputting the groundwater pollution vulnerability index to the input node of the adaptive neuro-fuzzy inference method network and inputting the nitrate nitrogen concentration to the output node to generate an ANFIS model by machine learning, the generated ANFIS model is input Groundwater pollution vulnerability based on an adaptive neuro-fuzzy inference technique that combines the differential evolution technique, which is configured to calculate the predicted value of the nitrate nitrogen concentration by inputting the groundwater pollution vulnerability factor to the node. Evaluation device.
The method of claim 1,
The plurality of groundwater pollution vulnerability factor data includes a differential evolution technique, including groundwater depth, groundwater recharge, aquifer medium, soil medium, topographic gradient, impact of vadose zone, hydraulic conductivity, and land use. Groundwater pollution vulnerability assessment device based on the combined adaptive neuro-fuzzy inference technique.
As a groundwater contamination vulnerability assessment method using a groundwater contamination vulnerability assessment device based on an adaptive neuro-fuzzy inference technique combining the differential evolution technique of claim 1,
Inputting a plurality of groundwater pollution vulnerability factor data and a plurality of nitrate nitrogen concentration data by a data input unit;
Selecting one or more groundwater contamination vulnerability factor data from among the plurality of groundwater contamination vulnerability factor data using a Variance Inflation Factor and a Tolerance Limit by a vulnerability factor selection unit;
Calculating a groundwater pollution vulnerability factor weight using a differential evolution (DE) algorithm based on the groundwater pollution vulnerability factor data and the nitrate nitrogen concentration data selected by the vulnerability factor weight calculator;
Calculating a groundwater pollution vulnerability index by multiplying the groundwater pollution vulnerability factor data selected by the adaptive neuro-fuzzy inference system model generator by the weight of the groundwater pollution vulnerability factor;
The groundwater pollution vulnerability index calculated by the adaptive neuro-fuzzy inference system model generation unit is input to the input node of the adaptive neuro-fuzzy inference system (ANFIS) network, and the nitrate nitrogen concentration Generating an ANFIS model by inputting to the output node and performing machine learning; And
And calculating a nitrate nitrogen concentration prediction value by inputting a groundwater contamination vulnerability factor to an input node of the ANFIS model in which the vulnerability value prediction unit is generated.
The method of claim 3,
The plurality of groundwater pollution vulnerability factor data is a method for evaluating groundwater pollution vulnerability including groundwater depth, groundwater recharge, aquifer medium, soil medium, topographic gradient, impact of vadose zone, hydraulic conductivity and land use. .

Images