Method and system for predicting beach erosion based on kinetic mechanisms and machine learning

Abstract

The present application provides a kind of based on dynamics mechanism and machine learning's beach recession prediction method, comprising: step 1. Collecting the historical period data of the river section where the prediction section is located;Step 2. The basic data is screened and the characteristic variable, beach recession probability target variable and beach recession width target variable are calculated based on the screened basic data;Step 3. Build beach recession probability prediction model and beach recession width prediction model;Step 4. The accuracy of the beach recession probability prediction model and the beach recession width prediction model is evaluated;Step 5. Based on beach recession probability prediction model and beach recession width prediction model, the beach recession probability and beach recession width of future period are predicted respectively.The present application method can automatically learn the law from multi-source historical data, consider the influence of multiple factors, establish the nonlinear mapping relationship between input characteristics and output targets, realize the quantitative prediction of beach recession, and has important significance for reducing disaster risk.

Description

Technical Field

[0001] This invention belongs to the field of river channel and river morphology prediction technology, specifically relating to a method and system for predicting bank collapse based on dynamic mechanisms and machine learning. Background Technology

[0002] Bank erosion is a phenomenon of instability and damage to river, lake, and ocean bank slopes caused by a combination of natural and human factors. It is one of the major challenges urgently needing to be addressed in the field of water conservancy engineering. Bank erosion is characterized by its sudden onset, high destructive power, and wide-ranging impact, seriously threatening the lives and property of people living along the banks and hindering economic development in those areas.

[0003] Traditional methods for predicting bank collapse primarily rely on empirical formulas and mechanical models. Empirical formulas are typically established based on observational data from specific regions and conditions. However, due to significant differences in geological, hydrological, and human activities across different regions, these formulas lack universality and are difficult to extend to other areas or conditions. Furthermore, empirical formulas usually only consider a few influencing factors, failing to comprehensively reflect the complex mechanisms of bank collapse. Mechanical models, on the other hand, are based on theories of soil mechanics and river dynamics, considering multiple factors to establish bank stability analysis models and assessing collapse risk by calculating bank safety factors. However, mechanical models require input of numerous soil physical and mechanical property parameters, such as soil cohesion, internal friction angle, and permeability coefficient. These parameters are often difficult to obtain accurately, especially for complex geological structures and heterogeneous soils, significantly impacting the predictive accuracy of mechanical models.

[0004] In recent years, machine learning technology has demonstrated powerful advantages in data mining and pattern recognition, providing new approaches to solving the aforementioned problems. Machine learning algorithms can automatically learn patterns from existing data and establish a nonlinear mapping relationship between input features and output targets, thereby achieving more accurate predictions. However, current research on quantitative prediction of bank erosion width remains relatively limited. Some models fail to fully incorporate the influence of multiple factors such as water and sediment, riverbed boundaries, and previous riverbed deformation, and lack correlation between feature selection and the model. Summary of the Invention

[0005] This invention combines multi-source historical data and constructs a riverbank collapse prediction model with an embedded feature selection module based on dynamic mechanisms and machine learning algorithms. This model can intelligently predict the probability and width of collapse of specific riverbanks under certain future conditions.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] A method for predicting beach collapse based on dynamic mechanisms and machine learning includes the following steps:

[0008] Step 1. Collect historical basic data for the river section where the prediction section is located;

[0009] Step 2. Filter the basic data and calculate the characteristic variables, the target variable of beach receding probability, and the target variable of beach receding width based on the filtered basic data;

[0010] Step 3. Preprocess the calculated feature variables, construct the beach receding probability prediction model and the beach receding width prediction model, and train the beach receding probability prediction model based on the preprocessed feature variables and the beach receding probability target variable, and train the beach receding width prediction model based on the preprocessed feature variables and the beach receding width target variable.

[0011] Step 4. Evaluate the accuracy of the beach erosion probability prediction model and the beach erosion width prediction model;

[0012] Step 5. Based on the beach erosion probability prediction model and the beach erosion width prediction model, predict the beach erosion probability and beach erosion width for future periods, respectively.

[0013] Furthermore, the historical basic data in step 1 includes the flow, water level, sediment concentration and median particle size of the bed sediment at the hydrological station in the river section where the predicted section is located in the historical period, as well as the measured topographic data of the predicted section.

[0014] Furthermore, in step 2, the objective variable for the probability of beach collapse is: whether the beach has deformed and collapsed, the beach collapse sample is marked as 1, and the stable or silted-up sample is marked as 0.

[0015] Furthermore, the target variable for the beach receding width is: the beach deformation width. By using the cross-sectional topography of adjacent measurements of the cross-section, the distance between the starting points of the left and right beach lips is determined, and the difference between the starting points of the beach lips between adjacent measurements is calculated as the beach deformation width.

[0016] Furthermore, the characteristic variables include: water and sediment condition characteristic variables, riverbed boundary characteristic variables, and previous riverbed deformation characteristic variables.

[0017] Furthermore, the water and sediment condition characteristic variables include: average flow rate, average sediment concentration, flow rate variation coefficient, average water flow scour intensity of the future period and the previous 3 periods, average cross-sectional velocity, flow rate change rate, average nearshore water depth and average nearshore velocity.

[0018] Riverbed boundary characteristic variables include: flat beach area, flat beach width and river facies coefficient; beach height and slope; relative distance from the thalweg to the bank, distance from the toe of the slope to the bank, relative elevation difference between the lowest point near the bank and the thalweg, and relative elevation difference between the near banks on the same and opposite sides; toe of the slope retreat width; median particle size of bed sand; distance from the bank to the regulation project;

[0019] The characteristic variables of early-stage riverbed deformation include: the width of early-stage thalweg swing and the thickness of scour and sedimentation, the change in bank slope, and the width of bank deformation.

[0020] Furthermore, the preprocessing in step 3 includes:

[0021] Based on the IterativeImputer method, missing or outlier values ​​of features are imputed;

[0022] Test for autocorrelation of characteristic variables and remove highly autocorrelated features;

[0023] Based on the Yeo-Johnson transform, the feature variables are debiased.

[0024] Transform the feature variable data into a distribution with a mean of 0 and a standard deviation of 1;

[0025] The preprocessed feature variables and the target variable of beach collapse probability are randomly or in a certain order divided into training set, validation set and test set, which are used as input data for the beach collapse probability prediction model.

[0026] The preprocessed feature variables and the target variable for beach receding width are used as input data for the beach receding width prediction model in the same format.

[0027] Furthermore, the specific steps in step 3 for constructing the probability prediction and width prediction models for beach erosion are as follows:

[0028] First, Lasso regression was used to screen candidate feature variables and remove low-correlation features. Then, the screened feature set was input into the KNN model for training. The number of feature variables and other hyperparameters of the KNN algorithm were jointly tuned. Grid search was used to determine the hyperparameter combination that makes the model achieve optimal performance on both the training and validation sets. Based on historical basic data, the beach receding probability prediction model and the receding width prediction model were trained respectively.

[0029] Further, step 5 includes:

[0030] Based on the trained beach erosion probability prediction model, the probability of beach erosion under specific conditions is predicted. If the probability prediction model predicts that it is a erosion, the erosion width is further predicted based on the beach erosion width prediction model.

[0031] On the other hand, the present invention provides a beach receding prediction system based on dynamic mechanisms and machine learning, comprising:

[0032] The data collection module is used to collect historical basic data for the river section where the prediction section is located.

[0033] The variable calculation module is used to filter the basic data and calculate the characteristic variables, the target variable of beach receding probability, and the target variable of beach receding width based on the filtered basic data.

[0034] The model building module is used to preprocess the calculated feature variables, build a beach receding probability prediction model and a beach receding width prediction model, and train the beach receding probability prediction model and the beach receding width prediction model based on the preprocessed feature variables and the target variable of beach receding probability.

[0035] Accuracy evaluation module. It is used to evaluate the accuracy of the beach receding probability prediction model and the beach receding width prediction model;

[0036] The prediction module is used to predict the probability and width of beach erosion in future periods based on the beach erosion probability prediction model and the beach erosion width prediction model, respectively.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The method of this invention can automatically learn patterns from multi-source historical data, consider the influence of multiple factors, establish a nonlinear mapping relationship between input features and output targets, and realize quantitative prediction of beach and bank collapse, which is of great significance for reducing disaster risks. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0040] Figure 1 This is a flowchart of the beach receding prediction method of the present invention;

[0041] Figure 2 This is a schematic diagram illustrating the extraction of target variables and the determination of the slope toe and the lowest point near the shore in this invention.

[0042] Figure 3 This is a data distribution diagram before and after the characteristic variables of this invention have been debiased and standardized.

[0043] Figure 4 This is a schematic diagram illustrating the parameter tuning basis for the beach receding probability prediction model of the present invention;

[0044] Figure 5 This is a schematic diagram illustrating the parameter tuning basis for the beach receding width prediction model of the present invention;

[0045] Figure 6 This is a schematic diagram of the prediction results of the beach receding probability prediction model of the present invention;

[0046] Figure 7 A schematic diagram of the prediction results of the beach receding width prediction model. Detailed Implementation

[0047] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0048] Example 1

[0049] The overall calculation process of this embodiment is shown below. Figure 1 It should be noted that: ① Bank erosion prediction is performed on a single bank, and even the left and right banks of the same cross section need to be predicted separately; ② Bank erosion probability prediction and bank erosion width prediction are implemented based on independent models; ③ Bank erosion width prediction only uses erosion samples.

[0050] Step 1. Collect historical basic data for the river section where the prediction section is located;

[0051] Specifically, this includes collecting data on flow, water level, sediment concentration, and median bed sediment size from hydrological stations along the river section where the prediction section is located, as well as measured topographic data of the prediction section. If there is a hydrological station near the prediction section, its data is used directly. If the prediction section is far from a hydrological station, data at the prediction section is obtained through linear interpolation based on data from upstream and downstream hydrological stations. To ensure sufficient topographic data, the prediction section is generally selected as a fixed section within the river channel. Increasing the amount of data usually helps improve prediction accuracy; therefore, relevant data should be collected as comprehensively as possible.

[0052] Step 2. Filter the basic data and calculate the characteristic variables, the target variable of beach receding probability, and the target variable of beach receding width based on the filtered basic data;

[0053] The filtering of the basic data includes:

[0054] Remove noise samples; delete noise samples that do not fall under the category of beach deformation, such as samples showing changes in the beach lip position caused by measurement errors or water flow creating new channels.

[0055] In this embodiment, the objective variable for the probability of beach erosion is: whether the beach deforms and erodes. Beach erosion samples are marked as 1, and stable or silted-up samples are marked as 0. The objective variable for the width of beach erosion is: the width of beach deformation. By using the cross-sectional topography of adjacent measurements of the overlay prediction section, the distance between the starting points of the left and right beach lips is determined, and the difference between the starting points of the beach lips between adjacent measurements is calculated as the width of beach deformation.

[0056] By predicting the cross-sectional topography of adjacent measurements, the distance between the starting points of the left and right bank lip is determined. The difference in the starting point distance between adjacent measurements is calculated as the bank deformation width. When bank erosion (siltation) occurs, the deformation width is defined as negative (positive). Figure 2 As shown, during one flood season, the left and right banks of the cross-section receded by 226m and 98m respectively. This invention includes a bank receding probability prediction model and a receding width prediction model. The bank receding probability prediction model is trained using all samples (including collapsing and non-collapsed samples), with the target variable being whether the bank deforms and recedes. Bank receding samples are marked as 1, and stable or silted samples are marked as 0. Due to the limitations of river topographic measurement accuracy, only samples where the bank receding width between adjacent measurements is greater than the actual measurement error are marked as collapsing (1). For the wandering section of the lower Yellow River, only samples where the receding width between adjacent measurements exceeds 6m are marked as collapsing (1). The bank receding width prediction model is trained using only receding samples, with the target variable being the bank receding width.

[0057] The characteristic variables calculated in this embodiment include: water and sediment condition characteristic variables, riverbed boundary characteristic variables, and previous riverbed deformation characteristic variables.

[0058] The characteristic variables of water and sediment conditions are calculated. First, based on typical water and sediment characteristic parameters from existing studies, variables characterizing the average water and sediment conditions of a cross-section are calculated, including average flow rate, average sediment concentration, coefficient of variation of flow rate, average flow scour intensity of the future period and the previous three periods, and average flow velocity of the cross-section. Simultaneously, this invention innovatively introduces the rate of change of flow rate to more accurately describe the impact of dynamic changes in flow rate during flood and dry seasons on the stability of bank collapse. Furthermore, variables characterizing nearshore flow conditions are calculated, including average nearshore water depth and average nearshore flow velocity. If there are no hydrological cross-sections near the predicted cross-section, the water and sediment process is obtained by linear interpolation of measured data from upstream and downstream hydrological cross-sections. The calculation methods for the coefficient of variation of flow rate, rate of change of flow rate, and nearshore water depth and flow velocity are mainly introduced below.

[0059] ① Flow variation coefficient (Q) Cv )

[0060] The coefficient of variation is used to characterize the dispersion of a flow series.

[0061]

[0062] In the formula: Q is the average flow rate (m³) of the predicted cross-section during the flood season (non-flood season). 3 / s); n is the number of days in the flood season (non-flood season); Q i The average daily flow (m³) on the i-th day during the flood season (non-flood season) 3 / s).

[0063] ② Rate of change in flow (Q) c )

[0064] The main flow path is closely related to the flow rate. Large fluctuations in flow rate lead to instability in the flow path, making the river morphology prone to drastic changes such as main channel swaying and bank erosion. The flow rate variation is defined as: when predicting bank erosion during the flood season, it is the ratio of the average flow rate during the flood season to the average flow rate during the previous non-flood season; when predicting bank erosion during the non-flood season, it is the ratio of the average flow rate during the previous flood season to the average flow rate during the current non-flood season. A larger value for this characteristic variable increases the probability of bank erosion. The calculation formula is as follows:

[0065] Q c =Q f / Q d (2)

[0066] In the formula: Q f Average flow during the flood season (m³) 3 / s); Q d Average flow rate during non-flood season (m³) 3 / s).

[0067] ③ Average water depth near the shore (H) b )

[0068] The average nearshore water depth during the flood season (and non-flood season) is the difference between the average cross-sectional water level and the elevation at the toe of the bank slope.

[0069]

[0070] In the formula: To calculate the average water level (m) of the cross-section during the flood season (non-flood season); Ze is the elevation at the toe of the slope (m). The toe of the slope is defined as... The following are locations where the riverbed slope changes significantly, such as where the riverbed becomes flat or where the slope sign changes, for example... Figure 2 As shown.

[0071] ④ Nearshore average current velocity (U b )

[0072] Assuming the nearshore roughness and gradient are the same as the cross-sectional average, the nearshore average velocity can be approximated by the cross-sectional average velocity according to Manning's formula. The calculation formula is as follows:

[0073]

[0074] In the formula: Let Z be the average water depth at the cross section, and Z be the ratio of the cross-sectional area to the river width.

[0075] The calculation of riverbed boundary characteristic variables includes main channel and bank slope morphology, main channel location, slope toe setback width, and the impact of regulation projects. Main channel morphology includes flat beach area, flat beach width, and river facies coefficient; bank slope morphology includes beach height and slope; the main channel location is characterized by the relative distance between the thalweg and the bank. To improve the accuracy of the description of riverbed boundary characteristics, this invention proposes variables such as the distance between the slope toe and the bank, the relative elevation difference between the lowest point near the bank and the thalweg, and the relative elevation difference between the near-bank and opposite banks. The distance between the slope toe and the bank represents the width of the beach that buffers against bank erosion; a larger distance indicates a relatively safer bank slope. The relative elevation difference between the lowest point near the bank and the thalweg indirectly reflects the near-bank flow intensity; a smaller elevation difference indicates a relatively stronger near-bank erosion intensity. The relative elevation difference between the near-bank and opposite banks reflects the contrast in the concentration of flow on both sides; flow tends to concentrate on the bank with lower elevation, resulting in relatively lower beach stability on that side. Analysis of a large amount of measured data revealed that these variables have a significant impact on the probability and magnitude of bank erosion, and therefore they are introduced into this method. In addition, the slope toe retreat width is calculated based on the average residual shear stress; the bed sand particle size at the predicted cross-section is obtained by linear interpolation of the measured values ​​at the hydrological cross-section; and the impact of the remediation project on bank erosion is quantified by the distance of the project from the bank.

[0076] ① Beach morphology

[0077] The location of the slope toe is as follows Figure 1 As shown, the formulas for calculating the beach height (BH) and slope (BS) are as follows:

[0078] BH = Z b -Z f (5)

[0079] BS = arctan[(Z b -Z f ) / |X f -X b |] (6)

[0080] In the formula: X f Z f These represent the distance from the starting point of the slope toe and the elevation (m); Z b The elevation of the beach lip is shown in meters (m).

[0081] ② Relative distance of the main current offshore

[0082] The distance from the toe of the slope to the shore (EB) can be combined with the relative distance from the thalweg to the shore (TB) to approximate the degree of the main current's contact with the shore. The calculation methods for both are as follows:

[0083] EB = |X e -Xb | (7)

[0084] TB=|X t -X b | / B bf (8)

[0085] In the formula: X t X e X b These are the distances (m) from the starting points of the thalweg, the nearshore scour point, and the left or right shoal lip, respectively; B bf The width of the Pingtan River is (m).

[0086] ③ Relative elevation difference (LT) between the lowest point near the shore and the thalweg

[0087] like Figure 2 As shown, the nearshore lowest point is the point with the lowest elevation within the nearshore bed area, with one on each of the left and right banks. Sometimes, the nearshore lowest point on one bank may coincide with the thalweg point in the cross-section. The relative elevation difference between the nearshore lowest point and the thalweg can approximately characterize the nearshore scouring intensity, and its calculation formula is as follows:

[0088] LT=(Z l -Z t ) / H bf (9)

[0089] In the formula: Z l Z t H represents the elevation (m) of the lowest point near the shore and the thalweg, respectively; bf The depth of the flat beach is (m).

[0090] ④ Relative elevation difference between the riverbeds on the left and right banks (△Z)

[0091] Water flow is generally concentrated on the lower banks near the shore. The relative elevation difference between the left and right banks near the shore was calculated to characterize the comparison of the degree of water flow concentration on the left and right banks. The smaller the value of this variable, the greater the probability of bank instability and collapse. The calculation formula is as follows:

[0092]

[0093] In the formula: The average elevation (m) within the near-shore riverbed area on the left and right banks, respectively.

[0094] ⑤ Slope toe setback width (ΔW)

[0095] The erosion of the slope toe by the water flow causes the bank slope to become steeper, creating conditions for bank collapse. The following formula is used here to calculate the width of the slope toe retreating due to water erosion:

[0096]

[0097] In the formula: kd m is the scouring coefficient. 3 / (Ns), is related to the soil's inherent properties and the initial shear stress, k d =2×10 -7 τ c -0.5 λ1 is the erosion index, typically taken as 1.0; τ c The initial shear stress during soil erosion, N / m 2 ;τ f The nearshore water flow shear stress is N / m. 2 And it is assumed that it is proportional to the water depth, i.e., τ f =γ w hJ, where γ w The specific weight of water, N / m³ 3 h represents the nearshore water depth in meters; J represents the longitudinal slope of the water surface.

[0098] The soil on the banks of the meandering section of the lower Yellow River has relatively large porosity and is not yet fully compacted. The following formula is used to calculate the initial shear stress of the soil:

[0099] τ c =6.68×10 2 ×d+3.67×10 -6 / d (12)

[0100] In the formula: d is the soil particle size in meters, which is approximately replaced by the median particle size of the bed sand at the slope toe.

[0101] ⑥ Impact of the remediation project

[0102] The impact of the remediation project is reflected by calculating the distance from the project to the shore (WB), which is the distance from the beach lip to the project. The width of the beach retreat will not exceed this value.

[0103] WB = |X b -X w | (13)

[0104] In the formula: X w The distance from the starting point of the project is (m).

[0105] Calculate characteristic variables of early-stage riverbed deformation. Given the lag in riverbed evolution, in addition to the current riverbed boundary conditions, this method supplements the calculation of key variables from the early-stage riverbed evolution, including the width and thickness of the thalweg swing, changes in bank slope, and the width of bank deformation, etc. The calculation formulas are as follows:

[0106] ① Width of the deep-water swing in the first stage (ΔB) t )

[0107] ΔB t =r×|X t -X't | (14)

[0108] In the formula: X is used when predicting changes in the riverbanks during the flood season. t 、X' t X represents the distance (m) from the starting point of the thalweg before the flood season of the current year and after the flood season of the previous year; when predicting changes in the shoal during the non-flood season, X t 、X' t These represent the distance (m) from the starting point of the thalweg after and before the flood season of the current year, respectively; r is a control symbol parameter, where r = -1 if the thalweg swings towards the bank in the previous period, and r = 1 if it swings towards the center of the river.

[0109] ② Average scouring and silting thickness of the first stage and the first two stages (ΔZ) t )

[0110]

[0111]

[0112] In the formula: Z is used when predicting changes in the riverbanks during the flood season. t Z' t Z” t Z represents the thalweg elevation (m) before the current flood season, after the previous flood season, and before the previous flood season, respectively; when predicting changes in the shoal during the non-flood season, Z... t Z' t Z” t The elevations (m) of the thalween after the flood season of the current year, before the flood season of the current year, and after the flood season of the previous year are respectively; r is a control symbol parameter, r = -1 when the thalween is saturated, and r = 1 when it is silted up.

[0113] ③ Changes in beach slope in the first stage (△S)

[0114] ΔS=BS'-BS (17)

[0115] In the formula: when predicting changes in the riverbank during the flood season, BS and BS' represent the riverbank slopes before the flood season of the current year and after the flood season of the previous year, respectively; when predicting changes in the riverbank during the non-flood season, BS and BS' represent the riverbank slopes after the flood season of the current year and before the flood season, respectively. If the slope becomes steeper, △S<0; otherwise, △S≥0.

[0116] Step 3. Preprocess the calculated feature variables, construct the beach receding probability prediction model and the beach receding width prediction model, and train the beach receding probability prediction model based on the preprocessed feature variables and the beach receding probability target variable, and train the beach receding width prediction model based on the preprocessed feature variables and the beach receding width target variable.

[0117] Preprocessing of the calculated feature variables includes:

[0118] Based on the IterativeImputer method, missing or outlier feature values ​​are imputed. However, due to limitations in the interpolation accuracy of topographic surveys and hydro-sediment processes, some feature values ​​are missing or numerically anomalous. To address this issue, this method utilizes IterativeImputer for multiple interpolation.

[0119] Unlike simple mean or mode imputation, the IterativeImputer method uses an iterative regression model to predict missing values, resulting in more accurate imputation. For each missing value, IterativeImputer uses other features as input to train a regression model to predict that missing value. Initial imputation typically employs a simple strategy (such as mean imputation). Subsequent iterations are performed, using the imputed features to predict the missing values ​​of other features in each iteration, until convergence or a predetermined number of iterations is reached. Its advantage lies in its ability to leverage the relationships between all features, thereby improving imputation accuracy.

[0120] To test the autocorrelation of feature variables, highly autocorrelated features are removed. High correlation between features may lead to multicollinearity, affecting the stability and accuracy of the model. Although the feature variables in step 3 have different physical meanings, they may still show high correlation in the data. For example, when the interannual fluctuations in water and sediment conditions are small, the water and sediment characteristic parameters of the current year may be correlated with the average water and sediment characteristics of previous years; when the thalweg is always close to a certain bank, the relative distance of the thalweg from the bank may be correlated with the distance of the nearshore scour point from the bank. Therefore, it is necessary to test the feature correlation and remove highly autocorrelated features. Specifically, first, the correlation coefficient matrix of all features is calculated; second, a threshold (0.9) is set, and feature pairs above this threshold are considered highly correlated; finally, one of the highly correlated features is deleted.

[0121] Based on the Yeo-Johnson transform, the feature variables are de-biased. In real-world data, feature variables may exhibit significant skewness, with the data distribution tilted to the left or right, failing to present a symmetrical normal distribution. Figure 4 As shown in figure a, this skewness can negatively impact model training. De-skewing reduces skewness by transforming the data, making the data distribution closer to a normal distribution, reducing the impact of outliers, and helping the model more accurately capture the relationship between features and the target variable. Considering that the feature variables in beach receding prediction can be both positive and negative, this method uses the Yeo-Johnson transform to de-skew the features.

[0122] Standardization transforms feature variable data into a distribution with a mean of 0 and a standard deviation of 1. Standardization balances the influence of each feature, allowing the model to utilize all feature information more fairly. If features are not standardized, their wide range of values ​​may affect similarity measurements or distance calculations, thus reducing model performance. Figure 3 Taking the river facies coefficient as an example, the data distribution before and after debiasing and standardization is shown. Before processing, the river facies coefficient ranged from 9.83 to 50.63, with the data mainly concentrated in the range of 10 to 15, showing a significant left skew. After processing, the variable ranged from -1.62 to 1.82, and the data distribution became more even.

[0123] The preprocessed feature variables and the target variable of beach collapse probability are randomly or in a certain order divided into training set, validation set and test set, which are used as input data for the beach collapse probability prediction model.

[0124] The preprocessed feature variables and the target variable for beach receding width are used as input data for the beach receding width prediction model in the same format.

[0125] This embodiment constructs a beach receding prediction model based on the K-Nearest Neighbors (KNN) algorithm and innovatively incorporates Lasso regression into the model for feature selection. Specifically, Lasso regression is first used to screen candidate feature variables, eliminating low-relevance features. The filtered feature set is then input into the KNN model for training to improve the model's prediction accuracy and generalization performance. During model optimization, the number of feature variables and other hyperparameters of the KNN algorithm are jointly tuned, and a grid search is used to determine the hyperparameter combination that achieves optimal performance on both the training and validation sets. Based on historical data, both the beach receding probability prediction model and the receding width prediction model are trained.

[0126] The parameter optimization basis for the beach erosion probability prediction model is as follows: Figure 4 As shown. By plotting learning curves of model accuracy versus the number of neighboring points for different numbers of features, a corresponding analysis graph is generated for each number of features. The parameter optimization basis for the beach receding width prediction model is as follows. Figure 5 As shown, the model lists the actual feature variables used by the model under different numbers of features in the training and validation sets, as well as the goodness-of-fit indices corresponding to different numbers of neighbor points. Based on the above results, the optimal number of features and neighbor points can be selected to ensure that the model has both high prediction accuracy and generalization ability.

[0127] Step 4. Evaluate the accuracy of the beach erosion probability prediction model and the beach erosion width prediction model;

[0128] Evaluate the model's accuracy on the test set. If the accuracy on the test set is also high, the trained model can be considered to have a certain generalization ability, and its prediction results have certain reference value. The model can be used for prediction (step 5). If it does not meet the requirements, the model needs to be retrained (step 4).

[0129] Step 5. Based on the beach erosion probability prediction model and the beach erosion width prediction model, predict the beach erosion probability and beach erosion width for future periods, respectively.

[0130] Input the sample data to be predicted, and based on the model trained in step 5, predict the probability and width of beach erosion under specific conditions. This method first applies the erosion probability prediction model, outputting the erosion probability (0-100%) and prediction category (erosion or no erosion) for each predicted sample, such as... Figure 6 As shown. If the prediction result is collapse, then the collapse width prediction model will output the collapse width of each predicted sample, as shown. Figure 7 As shown.

[0131] Example 2

[0132] This embodiment provides a beach collapse prediction system based on dynamic mechanisms and machine learning, including:

[0133] The data collection module is used to collect historical basic data for the river section where the prediction section is located.

[0134] The variable calculation module is used to filter the basic data and calculate the characteristic variables, the target variable of beach receding probability, and the target variable of beach receding width based on the filtered basic data.

[0135] The model building module is used to preprocess the calculated feature variables, build a beach receding probability prediction model and a beach receding width prediction model, and train the beach receding probability prediction model and the beach receding width prediction model based on the preprocessed feature variables and the target variable of beach receding probability.

[0136] Accuracy evaluation module. It is used to evaluate the accuracy of the beach receding probability prediction model and the beach receding width prediction model;

[0137] The prediction module is used to predict the probability and width of beach erosion in future periods based on the beach erosion probability prediction model and the beach erosion width prediction model, respectively.

[0138] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

[0139] It should be understood that any parts not described in detail in this specification belong to the prior art.

[0140] It should be understood that the above description of the preferred embodiments is quite detailed, but it should not be considered as a limitation on the scope of protection of this invention. Those skilled in the art, under the guidance of this invention, can make substitutions or modifications without departing from the scope of protection of the claims of this invention, and all such substitutions or modifications fall within the scope of protection of this invention. The scope of protection of this invention should be determined by the appended claims.

Claims

1. A method for predicting beach collapse based on dynamic mechanisms and machine learning, characterized in that, Includes the following steps: Step 1. Collect historical basic data for the river section where the prediction section is located; Step 2. Filter the basic data and calculate the characteristic variables, the target variable of beach receding probability, and the target variable of beach receding width based on the filtered basic data; The characteristic variables include: water and sediment condition characteristic variables, riverbed boundary characteristic variables, and previous riverbed deformation characteristic variables; the water and sediment condition characteristic variables include: average flow rate, average sediment concentration, flow rate variation coefficient, average flow scour intensity of the future period and the previous 3 periods, average cross-sectional velocity, flow rate change rate, average nearshore water depth, and average nearshore flow velocity. Riverbed boundary characteristic variables include: flat beach area, flat beach width and river facies coefficient; beach height and slope; relative distance from the thalweg to the bank, distance from the toe of the slope to the bank, relative elevation difference between the lowest point near the bank and the thalweg, and relative elevation difference between the near banks on the same and opposite sides; toe of the slope retreat width; median particle size of bed sand; distance from the bank to the regulation project; The characteristic variables of early-stage riverbed deformation include: the width and thickness of early-stage thalweg swing, changes in bank slope, and the width of bank deformation. Step 3. Preprocess the calculated feature variables to construct a beach erosion probability prediction model and a beach erosion width prediction model. Train the beach erosion probability prediction model based on the preprocessed feature variables and the beach erosion probability target variable, and train the beach erosion width prediction model based on the preprocessed feature variables and the beach erosion width target variable. The specific steps for constructing the beach erosion probability prediction and beach erosion width prediction models are as follows: First, Lasso regression was used to screen candidate feature variables and remove low-correlation features. Then, the screened feature set was input into the KNN model for training. The number of feature variables and other hyperparameters of the KNN algorithm were jointly tuned. Grid search was used to determine the hyperparameter combination that makes the model achieve optimal performance on both the training and validation sets. Based on historical basic data, the beach receding probability prediction model and the receding width prediction model were trained respectively. Step 4. Evaluate the accuracy of the beach erosion probability prediction model and the beach erosion width prediction model; Step 5. Based on the beach erosion probability prediction model and the beach erosion width prediction model, predict the beach erosion probability and beach erosion width for future periods, respectively.

2. The method for predicting beach collapse based on dynamic mechanisms and machine learning according to claim 1, characterized in that, The historical basic data in step 1 includes the flow, water level, sediment concentration and median particle size of bed sediment at the hydrological station in the river section where the predicted section is located, as well as the measured topographic data of the predicted section.

3. The method for predicting beach collapse based on dynamic mechanisms and machine learning according to claim 1, characterized in that, In step 2, the objective variable for the probability of beach collapse is: whether the beach has deformed and collapsed. Beach collapse samples are marked as 1, and stable or silted samples are marked as 0.

4. The method for predicting beach collapse based on dynamic mechanisms and machine learning according to claim 1, characterized in that, The target variable for the width of the beach receding is the width of the beach deformation. It is determined by using the cross-sectional topography of adjacent measurements to predict the cross-section, the distance between the starting points of the left and right beach lips, and the difference between the starting points of the beach lips between adjacent measurements, which is used as the width of the beach deformation.

5. The method for predicting beach collapse based on dynamic mechanisms and machine learning according to claim 1, characterized in that, The preprocessing in step 3 includes: Based on the IterativeImputer method, missing or outlier values ​​of features are imputed; Test for autocorrelation of characteristic variables and remove highly autocorrelated features; Based on the Yeo-Johnson transform, the feature variables are debiased. Transform the feature variable data into a distribution with a mean of 0 and a standard deviation of 1; The preprocessed feature variables and the target variable of beach receding probability are randomly or in a certain order divided into training set, validation set and test set, which are used as input data for the beach receding probability prediction model. The preprocessed feature variables and the target variable for beach receding width are used as input data for the beach receding width prediction model in the same format.

6. The method for predicting beach collapse based on dynamic mechanisms and machine learning according to claim 1, characterized in that, Step 5 includes: Based on the trained beach erosion probability prediction model, the probability of beach erosion under specific conditions is predicted. If the probability prediction model predicts that it is a erosion, the erosion width is further predicted based on the beach erosion width prediction model.

7. A shoreline collapse prediction system based on dynamic mechanisms and machine learning, characterized in that, include: The data collection module is used to collect historical basic data for the river section where the prediction section is located. The variable calculation module is used to filter the basic data and calculate the characteristic variables, the target variable of beach receding probability, and the target variable of beach receding width based on the filtered basic data. The model building module is used to preprocess the calculated feature variables, build a beach receding probability prediction model and a beach receding width prediction model, and train the beach receding probability prediction model and the beach receding width prediction model based on the preprocessed feature variables and the target variable of beach receding probability. Accuracy evaluation module. It is used to evaluate the accuracy of the beach receding probability prediction model and the beach receding width prediction model; Prediction module. It is used to predict the probability of beach collapse and the width of beach collapse in future periods based on the beach collapse probability prediction model and the beach collapse width prediction model, respectively. The beach collapse prediction system based on dynamic mechanisms and machine learning is used to perform the steps in the beach collapse prediction method based on dynamic mechanisms and machine learning as described in any one of claims 1-6.

Images