KR102614520B1 - Method and Device for Predicting Time-Concentration Curves using Its Reach-Length Dependence - Google Patents

Abstract

According to an embodiment of the present disclosure, there is provided a method for predicting material mixing behavior in a river, comprising: (a) preprocessing the raw time-concentration curve of a tracer obtained from a natural river by the preprocessing module 12; (b) extracting a plurality of concentration curve features from the preprocessed time-concentration curve by the concentration curve feature extraction module 13; (c) calculating a correlation coefficient between the plurality of concentration curve characteristics and a plurality of previously confirmed river characteristics by the correlation coefficient calculation module 14; (d) Among the plurality of river characteristics, the river characteristic with the highest correlation coefficient is called a significant feature, and the regression model construction module 15 determines the optimal characteristics for at least some of the significant feature and the plurality of concentration curve features. A step in which a regression model is constructed by determining a function; and (e) the regression model verification module 16 compares the time-concentration curve predicted by the regression model with the time-concentration curve predicted by one or more comparative prediction models and the preprocessed time-concentration curve. A method for predicting material mixture is provided, including the step of verifying the regression model.

Description

Method and device for predicting material mixing based on flow distance-concentration curve correlation {Method and Device for Predicting Time-Concentration Curves using Its Reach-Length Dependence}

The present disclosure relates to a method and device for predicting material mixing, and specifically, a method of predicting changes in the statistical characteristics of the concentration curve according to the flowing distance of a material in a river section where information on flow characteristics is not given, through regression analysis. and devices.

The content described in this section simply provides background information for the present disclosure and does not constitute prior art.

As industrialization has recently accelerated and the use of chemical raw materials has increased, the potential risk of accidents involving the inflow of hazardous chemicals into rivers is increasing. These accidents cause serious damage to the aquatic ecosystem and river environment, and can be fatal, especially in the country's flood control system, which relies on rivers and reservoirs for most of its irrigation and water supply. Meanwhile, most hazardous chemical spill accidents occur due to human negligence, making it difficult to predict when and where they will occur. Therefore, the best way to minimize damage is to provide rapid response measures. To this end, a prediction technology is needed that can predict the behavior of hazardous chemicals spreading down rivers and provide information such as arrival time or peak concentration of hazardous chemicals at important points such as downstream water intake sites.

In relation to this, a 1D Solute Transport Model (hereinafter referred to as “1D STM”) is proposed as a prediction system. 1D STM has advantages in terms of fast prediction and low computational cost. In particular, among 1D STMs, 1D ADE (Advection-Dispersion Equation) is the most basic model, and predicts the concentration curve based on Fickian diffusion theory with average flow rate and species dispersion coefficient as parameters. On the other hand, the above method has a problem in that it can only be applied to relatively monotonous rivers with little variability in river shape.

Accordingly, in order to increase prediction accuracy for rivers with irregular topography and flow conditions, a storage zone model that takes into account the influence of the river's storage zone was proposed. Here, the storage zone in a river refers to areas where the flow is very slow or stagnant, such as riffles, puddles, vegetation, or near hydraulic structures. Part of the highly concentrated polluted cloud is temporarily trapped and released back into the flow area after a certain residence time, causing contamination. It refers to the structural form of a river that distorts the cloud and concentration curve into a form with a long tail.

Meanwhile, 1D STM considering the reservoir model simplifies the complex three-dimensional river characteristics and models the river spreading behavior with a few parameters, but at this time, it is difficult to determine the exact parameters.

Specifically, when determining parameters, they are empirically predicted from flow characteristics such as river flow speed, or a tracer experiment is conducted in the target section and parameters appropriate for the target section are corrected using the tracer concentration curve. However, in the case of an actual hazardous chemical inflow accident, the time and location cannot be known in advance, so it is virtually impossible to secure optimal parameters by investigating river flow in advance or conducting tracer experiments in advance. there is.

(Non-patent literature) Eun-Jeong Kim, Chang-Min Park, Mi-Jeong Na, Hyun Park, and Bok-Sun Kim, "Analysis of tributary effects and simulation of water pollution accident scenarios in the Han River water source section using a 3D hydraulic model," 「Journal of the Korean Society for Water Environment」 34, no. 4(2018): 363-374

(Patent Document 1) KR 10-2010-0103117 A

(Patent Document 2) KR 10-2387450 B1 B1

(Patent Document 3) KR 10-2009-0059451 A

(Patent Document 4) KR 10-2004-0104008 A

Accordingly, the purpose of the present disclosure is to provide a material mixing prediction method and system that can predict material behavior in rivers with high accuracy.

The problems to be solved by the present invention are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood by those skilled in the art from the description below.

According to an embodiment of the present disclosure, there is provided a method for predicting material mixing behavior in a river, comprising: (a) preprocessing the raw time-concentration curve of a tracer obtained from a natural river by the preprocessing module 12; (b) extracting a plurality of concentration curve features from the preprocessed time-concentration curve by the concentration curve feature extraction module 13; (c) calculating a correlation coefficient between the plurality of concentration curve characteristics and a plurality of previously confirmed river characteristics by the correlation coefficient calculation module 14; (d) Among the plurality of river characteristics, the river characteristic with the highest correlation coefficient is called a significant feature, and the regression model construction module 15 determines the optimal characteristics for at least some of the significant feature and the plurality of concentration curve features. A step in which a regression model is constructed by determining a function; and (e) the regression model verification module 16 compares the time-concentration curve predicted by the regression model with the time-concentration curve predicted by one or more comparative prediction models and the preprocessed time-concentration curve. A method for predicting material mixture is provided, including the step of verifying the regression model.

In addition, preferably, step (a) of the present disclosure includes (a1) the concentration at the point where the power-law relationship between concentration and time in the decreasing section of the raw time-concentration curve ends is referred to as the basal concentration, removing the basal concentration from the time-concentration curve; (a2) The mass is corrected by multiplying the time-concentration curve from which the basal concentration has been removed in step (a1) by the ratio of the measured mass (w _m ) of the tracer to the initial injection mass (w ₀ ) of the tracer. step; (a3) The time-concentration curve corrected in step (a2) is subjected to Fourier transformation, a low pass filter is applied to the transformed curve, and the filtered curve is subjected to inverse Fourier transformation. transformation), thereby obtaining a preprocessed time-concentration curve.

In addition, preferably, the concentration curve characteristics of step (b) of the present disclosure include the initial arrival time of the tracer (T _f ), peak concentration arrival time (T _p ), median arrival time (T _med ), and standard deviation ( σ), concentration maintenance time over 75% of the peak concentration (T ₇₅ ), concentration maintenance time over 50% of the peak concentration (T ₅₀ ), concentration maintenance time over 10% of the peak concentration (T ₁₀ ), skewness (SKNS) , kurtosis (KURT), peak concentration (C _max ), average concentration (C _mean ), slope of the rising part of the curve (S _r ), and average slope of the descending part of the curve to the point where it reaches 10% of the peak concentration (C _max ). (S _f ) and the average slope (S _t ) after the point where it reaches 10% of the peak concentration (C _max ) in the descending part of the curve.

In addition, preferably, the plurality of river characteristics already confirmed in step (c) of the present disclosure include flow distance (L), average flow velocity (U), average cross-sectional area (A), average river width (W), and average water depth. Includes (H).

In addition, preferably, when it is confirmed that the significant feature of step (d) of the present disclosure is the flow distance (L), among the plurality of concentration curve features, peak concentration (C _max ) and peak concentration arrival time (T _p ), the optimal function is determined by obtaining the parameters (θ _cp , θ _tp , θ _tf , θ _w ) between the initial arrival time (T _f ) and the concentration maintenance time (T ₅₀ ) above 50% of the peak concentration.

Also, preferably, the parameters (θ _cp , θ _tp , θ _tf , θ _w ) of the present disclosure are obtained using predetermined equations.

Also, preferably, the one or more comparative prediction models of the present disclosure are Advection-Dispersion Equation (ADE) and Transient Storage Model (TSM).

In addition, according to an embodiment of the present disclosure, an experiment result input module 10 in which tracer experiment results in a natural river are input; A river characteristic input module 11 in which a plurality of river characteristics confirmed from natural rivers are input; a preprocessing module 12 configured to preprocess the tracer experiment results and calculate a preprocessed time-concentration curve; a concentration curve feature extraction module 13 configured to extract a plurality of concentration curve features derived from the preprocessed time-concentration curve; a correlation coefficient calculation module 14 configured to calculate correlation coefficients between the plurality of concentration curve characteristics and the plurality of river characteristics; a regression model building module (15) configured to construct a regression model by determining an optimal function between at least some of the plurality of concentration curve characteristics and one river characteristic among the plurality of river characteristics; A material mixture comprising a regression model verification module 16 configured to verify the regression model by comparing the predicted value predicted by the regression model with the predicted value predicted by one or more comparative prediction models and previously acquired actual values. Provides a prediction device.

Also, preferably, the plurality of river characteristics of the present disclosure include flowing distance (L), average flow velocity (U), average cross-sectional area (A), average river width (W), and average water depth (H).

In addition, preferably, the plurality of concentration curve characteristics of the present disclosure include the tracer's initial arrival time (T _f ), peak concentration arrival time (T _p ), median arrival time (T _med ), standard deviation (σ), Concentration maintenance time over 75% of the peak concentration (T ₇₅ ), concentration maintenance time over 50% of the peak concentration (T ₅₀ ), concentration maintenance time over 10% of the peak concentration (T ₁₀ ), skewness (SKNS), kurtosis ( KURT), peak concentration (C _max ), average concentration (C _mean ), slope of the rising part of the curve (S _r ), and average slope of the descending part of the curve to the point where it reaches 10% of the peak concentration (C _max ) (S _f ) and the average slope (S _t ) after the point where the curve descends to 10% of the peak concentration (C _max ).

As described above, according to this embodiment, the behavior of materials in an unknown section can be predicted with high accuracy by using river characteristics at already known points.

1 is a block diagram of a material mixing prediction device according to an embodiment of the present disclosure.
2 to 4 are flowcharts of a method for predicting material mixing using a material mixing prediction device according to an embodiment of the present disclosure.
Figure 5 is for explaining a tracer experiment conducted to obtain information input to the experiment result input module and river characteristics input module according to an embodiment of the present disclosure.
Figures 6 and 7 are for explaining a pre-processing method of a pre-processing module according to an embodiment of the present disclosure.
Figure 8 is for explaining concentration curve characteristics according to an embodiment of the present disclosure.
Figure 9 shows the correlation coefficient between river characteristics and concentration curve characteristics obtained through a material mixing prediction device according to an embodiment of the present disclosure.
Figure 10 shows the correlation between significant features and river characteristics obtained through a material mixing prediction device according to an embodiment of the present disclosure.
Figures 11 to 13 show the results of verifying a regression model by the material mixture prediction device according to an embodiment of the present disclosure.

Hereinafter, some embodiments of the present disclosure will be described in detail through illustrative drawings. When adding reference signs to components in each drawing, it should be noted that the same components are given the same reference numerals as much as possible even if they are shown in different drawings. Additionally, in describing the present disclosure, if it is determined that a detailed description of a related known configuration or function may obscure the gist of the present disclosure, the detailed description will be omitted.

In describing the components of the embodiment according to the present disclosure, symbols such as first, second, i), ii), a), and b) may be used. These codes are only used to distinguish the component from other components, and the nature, sequence, or order of the component is not limited by the code. In the specification, when a part is said to 'include' or 'have' a certain element, this means that it does not exclude other elements, but may further include other elements, unless explicitly stated to the contrary. .

In the present disclosure, “module” refers to hardware or software for performing a specific function, and each module may be provided in one volatile/non-volatile memory or separate volatile/non-volatile memory, which can be appropriately designed. Pay attention to

1 is a block diagram of a material mixture prediction device according to an embodiment of the present disclosure.

Referring to FIG. 1, the material mixing prediction device 1 according to an embodiment of the present disclosure includes an experiment result input module 10, a river characteristics input module 11, a preprocessing module 12, and a concentration curve feature extraction module ( 13), it includes all or part of the correlation coefficient calculation module (14), regression model construction module (15), and regression model verification module (16).

2 to 4 are flowcharts of a method for predicting material mixing using a material mixing prediction device according to an embodiment of the present disclosure.

Referring to FIGS. 1, 2, and 4, the material mixing prediction method according to the present disclosure is based on the experimental results (Raw Tracer BTCs) input by the experimental result input module 10 and the river characteristics input module 11. The preprocessing module 12 includes a step of preprocessing the time-concentration curve based on the input channel and flow properties (S200). Meanwhile, detailed information related to preprocessing is described in detail in Table of Contents 2.

The concentration curve feature extraction module 13 extracts a plurality of concentration curve features (BTC features) from the preprocessed time-concentration curve (Refined BTCs) (S210).

The correlation coefficient calculation module 14 analyzes the correlation between a plurality of concentration curve characteristics and a plurality of river characteristics (S220).

The regression model building module 15 extracts significant features and uses them to build regression models (S230).

The regression model verification module 16 compares the results predicted by the regression model with the results predicted by the comparative prediction model (e.g., ADE and TSM) and the preprocessed time-concentration curve in the S200 step to determine the regression model. Verify (S240).

Hereinafter, with reference to FIGS. 2 to 13 of the present disclosure, the material mixing prediction device 1 and a method of using the same according to an embodiment of the present disclosure will be described step by step.

1. Preparation steps for material mixing prediction method

1.1 Raw data acquisition stage through tracer experiment

The tracer experiment described in this disclosure is an experiment conducted by applying a fluorescent material to an actual measurement section and measuring the concentration of the fluorescent material using an optical sensor.

Figure 5 is for explaining a tracer experiment conducted to obtain information input to the experiment result input module and river characteristics input module according to an embodiment of the present disclosure.

Figure 5(a) shows the location and method of the tracer experiment in Cheongmi Stream located in Korea, and Figure 5(b) shows the location and method of the tracer experiment in Gamcheon located in Korea. Meanwhile, the experiment for the preparation stage of the material mixing prediction method according to the present disclosure does not necessarily have to be performed in the above-mentioned stream, and may be performed in any stream. Note that in the following, embodiments of the present disclosure will be described based on data obtained through experiments conducted in Cheongmicheon and Gamcheon.

Through tracer experiments, each point (A-IP, A-S1, A-S2, A-S3, A-S4, B-IP, B-S1, B-S2,B-S3, B-S4, B-S5 , C,D,-IP, C-S1, C-S2, C-S3, C-S4, D-S1, D-S2, D-S3, D-S4), the detected fluorescence It is configured to measure the concentration of a substance over time, and the concentration information over time at each point is input to the experiment result input module 10 as a dataset. Using these datasets, raw time-concentration curves can be obtained.

1.2 River characteristics acquisition stage

The river characteristics described in this disclosure refer to variables representing the flow within a river section, and are representative of flow distance (L), average flow velocity (U), average cross-sectional area (A), average river width (W), and average water depth ( Includes all or part of H).

River characteristics at each point are actually measured and input into the river characteristic input module (11).

2. Preprocessing step of time-concentration curve

Tracer experiments are widely used as a method to indirectly determine river flow. Meanwhile, the raw data from the tracer experiment is accompanied by various noises, and a method for removing such noise will be described below.

2.1. Basal concentration removal step

Referring to FIG. 3, the preprocessing step includes removing the basal concentration from the raw time-concentration curve (S300).

During tracer experiments, the concentration of fluorescent substances is detected using an optical sensor. Meanwhile, in the case of general rivers, the concentration of fluorescent substances tends to be measured higher than the actual concentration due to suspended matter present in the water. The concentration that increases due to suspended matter present in water is called basal concentration.

It has been revealed by empirical formulas that the distribution of residence time of substances affected by the storage zone of substances follows a power function. Using this principle, it can be seen that the basal concentration is the concentration in the section where the value of concentration versus time no longer follows the power function in the decreasing part of the original time-concentration curve.

In other words, the concentration at the point where the power law relationship of concentration with respect to time begins to end during the concentration reduction section of the raw time-concentration curve is called the basal concentration, and the concentration equal to the basal concentration is subtracted from the raw time-concentration curve. This will cause the raw time-concentration curve to shift overall down with respect to the concentration axis.

2.2. Mass correction step

Referring to FIG. 3, the preprocessing step includes correcting the mass of the time-concentration curve from which the basal concentration has been removed (S310).

One or more optical sensors are disposed at each point shown in Figure 5. For example, at points A-S1, a plurality of optical sensors are arranged in a row along a cross section perpendicular to the direction of river flow. The concentration of the fluorescent substance acquired in the experiment is the sum of all the concentrations measured by a plurality of optical sensors arranged in a row. Therefore, the concentration value obtained in the experiment will be measured to be smaller than the actual tracer concentration. To correct this, the mass (w _m ) of the tracer measured at the downstream point (A-S4, B-S5, C-S4, D-S4 in Figure 5) is measured and compared with the initial injection mass (w ₀ ). Correct the time-concentration curve with the basal concentration removed. For example, there may be a method of multiplying the ratio of the measured mass of the tracer (w _m ) to the initial injection mass (w ₀ ) by the concentration axis of the time-concentration curve from which the basal concentration has been removed.

Meanwhile, mass correction according to the material mixture prediction method of the present disclosure is based on the assumption that the injected tracer is a conservative material and that the mass of the injected tracer is all detected downstream.

2.3. Noise removal stage

Referring to FIG. 3, the preprocessing step includes removing noise by applying a filter to the mass-corrected time-concentration curve (S320). Meanwhile, noise removal must be performed in the frequency domain. In step S310, the corrected time-concentration curve is Fourier transformed, and a low pass filter is applied to the Fourier transformed curve. Then, by inverse Fourier transformation again, the time-concentration curve with the preprocessing completed will be finally obtained.

At this time, an ideal filter, a Gaussian filter, or a Butterworth filter may be used as the low-pass filter. In the present disclosure, a response function (H) according to [Equation 1] is used. It is preferred that a Butterworth filter is applied.

Here, the cutoff frequency (w _c ) and filtering order (n) can be appropriately changed by the designer. Figure 6 shows the response of the Butterworth filter according to the size of the order (n). Specifically, this is a graph comparing the response functions of the ideal filter and Gaussian filter and the response of the Butterworth filter. The slope of the transition section between the pass band and the cutoff band can be adjusted through the order n value of the filter, which has the advantage of separating unnecessary high frequency components from low frequency components that are to be preserved.

Figure 7 shows a graph comparing the noise removal performance of a Gaussian filter and a Butterworth filter. Specifically, the noise processing results of applying a Gaussian filter (blue markings of each graph) under the condition that the distortion of the raw time-concentration curve at each point (gray markings of each graph) is limited to 0.1%, and the noise of applying the Butterworth filter This shows the processing results (red markings on each graph). Referring to the time-concentration curve results at each point in Figure 7, it can be seen that the results of applying the Butterworth filter are better than the results of applying the Gaussian filter. Here, superior means showing a smoother trend in the time-concentration curve.

3. Concentration curve feature extraction step

Figure 8 is for explaining concentration curve characteristics according to an embodiment of the present disclosure.

Referring to FIG. 8, concentration curve characteristics refer to statistical characteristics necessary to determine the shape of the curve in the time-concentration curve. Concentration curve features largely include time-related features, material diffusion-related features, curve shape-related features, concentration-related features, and curve slope-related features.

Time-related features include the initial arrival time of the polluted cloud (T _f ), peak concentration arrival time (T _p ), and median arrival time (T _med ).

Diffusion-related characteristics of substances include standard deviation (σ) and retention time (T ₇₅ , T ₅₀ , T ₁₀ ) above a certain concentration (75%, 50%, 10% of the peak concentration).

Curve shape-related features include skewness (SKNS) and kurtosis (KURT). The more a substance is affected by the storage zone, the longer the tail of the acid-concentration curve will be, which means the greater the skewness (SKNS). Therefore, as the degree of diffusion of a substance increases, the kurtosis (KURT) will decrease.

Concentration-related characteristics include peak concentration (C _max ) and average concentration (C _mean ).

The slope-related characteristics of the curve are the slope of the rising part of the curve (S _r ), the average slope up to 10% of the peak concentration (C _max ) in the descending part of the curve (S _f ), and the peak concentration (C Includes the average slope (S _t ) after the point where it drops to 10% of _max ).

In summary, the 14 concentration curve characteristics described above correspond to statistical characteristics that can determine the approximate shape of time-concentration. Therefore, if the numerical values of the concentration curve characteristics can be determined, conversely, the approximate shape of the time-concentration curve can be inferred. To this end, the concentration curve feature extraction module 13 is configured to extract concentration curve features from the preprocessed time-concentration curve.

4. Regression model construction stage

4.1. Correlation analysis step between concentration curve characteristics and river characteristics

Next, the correlation between river characteristics and concentration curve characteristics is analyzed. Illustratively, the correlation may be determined by calculating the correlation coefficient according to [Equation 2] by the correlation coefficient calculation module 14. Meanwhile, in the present disclosure, the characteristic with the highest correlation with the concentration curve characteristic among river characteristics is referred to as a “significant feature.”

The correlation coefficients of 5 river characteristics with respect to 14 concentration curve characteristics calculated by the correlation coefficient calculation module 14 are shown in FIG. 9. Referring to Figure 9, it is confirmed that among river characteristics, flow distance (L, red bar graph in Figure 9) shows the highest correlation with concentration curve characteristics. In particular, it is understood that the flow distance (L) shows a significantly higher correlation than other river characteristics, so it will be possible to infer the concentration curve characteristics to some extent only with the flow distance (L). That is, the material mixing prediction device 1 according to an embodiment of the present disclosure determines the flow distance (L) as a significant feature.

Referring to Figure 10, among the concentration curve characteristics, the peak concentration (C _max ), peak concentration arrival time (T _p ), initial arrival time (T _f ), and concentration maintenance time above 50% of the peak concentration (T ₅₀ ) are the flow distance. It is confirmed to have a high correlation coefficient with (L).

4.2. Steps to build a regression model using parameters

The regression model building module 15 builds a regression model using the significant features and four concentration curve features identified in Table of Contents 4.1. For this purpose, significant characteristics, namely, flow distance (L), peak concentration (C _max ), peak concentration arrival time (T _p ), initial arrival time (T _f ), and concentration maintenance time above 50% of the peak concentration (T ₅₀ ), each parameter (θ _cp , θ _tp , θ _tf , θ _w ) must be determined.

To determine the parameters, the regression model building module 15 finds the optimal function among distribution functions with a single parameter in [Equation 3] to [Equation 6].

At this time, UBC and DBC in [Equation 3] and [Equation 6] mean the upstream boundary condition and the downstream boundary condition. Additionally, in [Equation 3] is the maximum value of concentration (C) at the downstream boundary condition, in [Equation 3] is the maximum value of concentration (C) at the upstream boundary condition, in [Equation 4] is the change in peak concentration arrival time, [Equation 5] is the change in initial arrival time, in [Equation 6] is the time to maintain more than 50% of the maximum concentration at the downstream boundary, in [Equation 6] means the time to maintain more than 50% of the maximum concentration at the upstream boundary.

The parameters of each expression ( , , , ) can be determined as the optimal value from the concentration curve measured upstream. The parameters determined for the target site of this example are as shown in [Table 1].

Parameter (R ² ) Test A 0.00109 4.696 3.931 0.00040 (0.657) (0.999) (0.995) (0.746) Test B 0.00112 4.601 2.768 0.00059 (0.532) (0.744) (0.490) (0.505) Test C 0.00184 2.284 1.966 0.00066 (0.910) (0.918) (0.963) (0.982) Test D 0.00231 2.692 2.248 0.00073 (0.985) (0.958) (0.980) 0.984) overall 0.00150 3.464 2.567 0.00066 (0.626) (0.538) (0.559) (0.800)

Using the material mixing prediction method according to the present disclosure, it can be applied to predicting the concentration curve by correcting the parameters of the model from tracer experimental data that can be measured upstream and applying it to the unknown river section downstream, It has the advantage of having high adaptability to unknown sections.

5. Regression model verification step

The material mixing prediction method according to the present disclosure compares the predicted value using a regression model based on the correlation between flow distance (L)-concentration curve characteristics with the predicted value using the comparative prediction model and the preprocessed time-concentration value, Accuracy can be verified. At this time, the comparative prediction model may preferably be an Advection-Dispersion Equation (ADE) model and a Transient Storage Model (TSM).

5.1. Parameter calibration steps

The material mixture prediction method may further include the step of calibrating model parameters using data from the remaining observation points, assuming that data from the most downstream measurement point is not known from the tracer experiment results. The parameter correction case is as shown in [Table 2], and the parameter correction case and verification case are shown in Figure 11.

Correction case destination Application model Concentration curve measurement position A Cheongmi Stream RP S1, S2, S3 A-M1 ADE & TSM S1, S3 A-M2 S2, S3 B Cheongmicheon Stream RP S1, S2, S3, S4 B-M1 ADE & TSM S1, S4 B-M2 S2, S4 B-M3 S3, S4 C Gamcheon RP S1, S2, S3 C-M1 ADE & TSM S1, S3 C-M2 S2, S3 D Gamcheon RP S1, S2, S3 D-M1 ADE & TSM S1, S3 D-M2 S2, S3

The parameters corresponding to each correction case of the model applied in this example are determined as shown in [Table 3].

Model Para-meter Unit Calibration cases Test A Test B Test C Test D M1 M2 M1 M2 M3 M1 M2 M1 M2 ADE m s-1 0.177 0.193 0.279 0.241 0.256 0.564 0.611 0.395 0.384 m2 s-1 18.15 13.87 23.32 13.82 5.681 14.08 4.5 07 6.650 8.090 TSM m s-1 0.241 0.250 0.274 0.308 0.256 0.625 0.645 0.395 0.420 m2 s-1 1.702 1.214 23.14 2.175 5.681 0.547 0.591 6.650 1.730 m2 3.728 2.959 90.62 2.723 ~10-6 4.088 30.72 50.71 0.860 10-4 s-1 1.51 1.25 ~10-5 2.518 0.199 3.63 2.92 ~10-5 2.01 RP m-1 0.00132 0.00123 0.00182 0.00234 s m-1 4.745 3.671 1.653 2.425 s m-1 3.993 1.911 1.622 2.095 m-1 0.00045 0.00068 0.00070 0.00079

5.2. Accuracy verification step of regression model

12 to 13 compare the time-concentration curve predicted by the regression model generated and corrected by the material mixing prediction device 1, the preprocessed time-concentration curve, and the time-concentration curve predicted by the prediction model. This shows the comparison results. Specifically, Figure 12 shows the curve model itself, and Figure 13 shows the coefficient of determination (R ² ) calculated using the results shown in Figure 12. At this time, the coefficient of determination (R ² ) is calculated by the regression model verification module 16.

In the case of an experiment performed according to an embodiment of the present disclosure, all river characteristics are not given and the results are based on only the flowing distance (L). Therefore, the results predicted by the models may not completely match the actual measurements (i.e., preprocessed time-concentration curves). In particular, referring to Figure 13, the simulation results of ADE and TSM show that the prediction accuracy for peak concentration decreases in the order of tracer experiment cases C, D, B, and A. In addition, it can be seen that the calculated coefficient of determination (R ² ), which indicates the accuracy of these prediction results, decreases as the flowing distance increases.

In contrast, as shown in FIG. 13, the regression model according to an embodiment of the present disclosure (red bar graph in FIG. 13) is relatively more accurate than the ADE model and the TSM model in the case of tests A, C and D. Show results. It is believed that the low accuracy in Test B is due to the low correlation in the correlation analysis results at the target site. However, referring to the red dotted line in (b) of Figure 12, it is confirmed that the prediction accuracy is still acceptable.

In particular, it is confirmed that the regression model according to an embodiment of the present disclosure shows high accuracy in prediction for test C, which is the longest-distance prediction (see FIG. 13). Specifically, while the coefficient of determination values of the ADE model and the TSM model indicate prediction failure, the technique of the present disclosure shows relative excellence in predicting long-range concentration curves.

Considering the long-distance prediction performance and computational cost in an unknown section of a river, the prediction technique presented in this disclosure is advantageous for quickly and accurately predicting concentration curves when an actual hazardous chemical inflow accident occurs.

The above description is merely an illustrative explanation of the technical idea of this embodiment, and those skilled in the art will be able to make various modifications and variations without departing from the essential characteristics of this embodiment. Accordingly, the present embodiments are not intended to limit the technical idea of the present embodiment, but rather to explain it, and the scope of the technical idea of the present embodiment is not limited by these examples. The scope of protection of this embodiment should be interpreted in accordance with the claims below, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of rights of this embodiment.

1: Material mixing prediction device
10: Experiment result input module
11: River characteristics input module
12: Preprocessing module
13: Concentration curve feature extraction module
14: Correlation coefficient calculation module
15: Regression model construction module
16: Regression model verification module

Claims (12)

As a method for predicting material mixing behavior in rivers,
(a) preprocessing the raw time-concentration curve of the tracer obtained from a natural river by the preprocessing module 12;
(b) extracting a plurality of concentration curve features from the preprocessed time-concentration curve by the concentration curve feature extraction module 13;
(c) calculating a correlation coefficient between the plurality of concentration curve characteristics and a plurality of previously confirmed river characteristics by the correlation coefficient calculation module 14;
(d) Among the plurality of river characteristics, the river characteristic with the highest correlation coefficient is called a significant feature, and the regression model construction module 15 determines the optimal characteristics for at least some of the significant feature and the plurality of concentration curve features. A step in which a regression model is constructed by determining a function; and
(e) The regression model verification module 16 compares the time-concentration curve predicted by the regression model with the time-concentration curve predicted by one or more comparative prediction models and the preprocessed time-concentration curve. Including the step of verifying the regression model,
Material mixing prediction method.
According to paragraph 1,
In step (a),
(a1) the concentration at the point where the power law relationship between concentration and time in the decreasing section of the original time-concentration curve ends is referred to as the basal concentration, and the basal concentration is removed from the time-concentration curve;
(a2) The mass is corrected by multiplying the time-concentration curve from which the basal concentration has been removed in step (a1) by the ratio of the measured mass (w _m ) of the tracer to the initial injection mass (w ₀ ) of the tracer. step;
(a3) The time-concentration curve corrected in step (a2) is subjected to Fourier transformation, a low pass filter is applied to the transformed curve, and the filtered curve is subjected to inverse Fourier transformation. Transformation), which includes the step of obtaining a preprocessed time-concentration curve,
Material mixing prediction method.
According to paragraph 2,
The concentration curve characteristics of step (b) include the tracer's initial arrival time (T _f ), peak concentration arrival time (T _p ), median arrival time (T _med ), standard deviation (σ), and 75% or more of the peak concentration. Concentration holding time (T ₇₅ ), concentration holding time over 50% of the peak concentration (T ₅₀ ), concentration holding time over 10% of the peak concentration (T ₁₀ ), skewness (SKNS), kurtosis (KURT), peak concentration ( C _max ), average concentration (C _mean ), slope of the rising part of the curve (S _r ), average slope (S _f ) of the descending part of the curve to the point where it reaches 10% of the peak concentration (C _max ), and of the descending part of the curve Including the average slope (S _t ) after the point where it falls to 10% of the peak concentration (C _max ),
Material mixing prediction method.
According to paragraph 3,
The plurality of river characteristics already confirmed in step (c) include flowing distance (L), average flow velocity (U), average cross-sectional area (A), average river width (W), and average water depth (H),
Material mixing prediction method.
According to paragraph 4,
If the significant feature of step (d) is confirmed to be the flow distance (L), among the plurality of concentration curve features, peak concentration (C _max ), peak concentration arrival time (T _p ), and first arrival time (T _f ) and the optimal function is determined by obtaining the parameters (θ _cp , θ _tp , θ _tf , θ _w ) between the concentration retention time (T ₅₀ ) of 50% or more of the peak concentration,
Material mixing prediction method.
According to clause 5,
The parameters (θ _cp , θ _tp , θ _tf , θ _w ) are obtained using Equations 1 to 4 below,
[Equation 1]
[Equation 2]
[Equation 3]
[Equation 4]
In [Equation 1] above, is the maximum value of concentration (C) at the downstream boundary condition, is the maximum value of concentration (C) at the upstream boundary condition,
In [Equation 2] above, is the change in time to reach peak concentration,
In [Equation 3] above, is the change in initial arrival time,
In [Equation 4] above, is the time to maintain more than 50% of the maximum concentration at the downstream boundary, is the time to maintain more than 50% of the maximum concentration at the upstream boundary,
Material mixing prediction method.
According to paragraph 1,
The one or more comparative prediction models are Advection-Dispersion Equation (ADE) and Transient Storage Model (TSM),
Material mixing prediction method.
An experiment result input module (10) in which tracer experiment results in natural rivers are input;
A river characteristic input module 11 in which a plurality of river characteristics confirmed from natural rivers are input;
a preprocessing module 12 configured to preprocess the tracer experiment results and calculate a preprocessed time-concentration curve;
a concentration curve feature extraction module 13 configured to extract a plurality of concentration curve features derived from the preprocessed time-concentration curve;
a correlation coefficient calculation module 14 configured to calculate correlation coefficients between the plurality of concentration curve characteristics and the plurality of river characteristics;
a regression model building module (15) configured to construct a regression model by determining an optimal function between at least some of the plurality of concentration curve characteristics and one river characteristic among the plurality of river characteristics; and
Comprising a regression model verification module 16 configured to verify the regression model by comparing the predicted value predicted by the regression model with the predicted value predicted by one or more comparative prediction models and previously acquired actual values,
Material mixing prediction device.
According to clause 8,
The plurality of river characteristics include flow distance (L), average flow velocity (U), average cross-sectional area (A), average river width (W), and average water depth (H),
Material mixing prediction device.
According to clause 9,
The plurality of concentration curve characteristics include the tracer's initial arrival time (T _f ), peak concentration arrival time (T _p ), median arrival time (T _med ), standard deviation (σ), and concentration maintenance time of 75% or more of the peak concentration. (T ₇₅ ), concentration maintenance time over 50% of the peak concentration (T ₅₀ ), concentration maintenance time over 10% of the peak concentration (T ₁₀ ), skewness (SKNS), kurtosis (KURT), peak concentration (C _max ) , average concentration (C _mean ), slope of the rising part of the curve (S _r ), average slope up to 10% of the peak concentration (C _max ) in the descending part of the curve (S _f ), and peak concentration in the descending part of the curve ( Containing the average slope (S _t ) after the point where it falls to 10% of C _max ),
Material mixing prediction device.
According to clause 10,
One river characteristic among the plurality of river characteristics is the flow distance (L),
If at least some of the plurality of concentration curve characteristics are peak concentration (C _max ), peak concentration arrival time (T _p ), first arrival time (T _f ), and concentration maintenance time (T ₅₀ ) of 50% or more of the peak concentration, The regression model building module 15 constructs a regression model by calculating parameters (θ _cp , θ _tp , θ _tf , θ _w ) using Equations 1 to 4 below,
[Equation 1]
[Equation 2]
[Equation 3]
[Equation 4]
In [Equation 1] above, is the maximum value of concentration (C) at the downstream boundary condition, is the maximum value of concentration (C) at the upstream boundary condition,
In [Equation 2] above, is the change in time to reach peak concentration,
In [Equation 3] above, is the change in initial arrival time,
In [Equation 4] above, is the time to maintain more than 50% of the maximum concentration at the downstream boundary, is the time to maintain more than 50% of the maximum concentration at the upstream boundary,
Material mixing prediction device.
According to clause 8,
The one or more comparative prediction models are Advection-Dispersion Equation (ADE) and Transient Storage Model (TSM),
Material mixing prediction device.

Images