Vertical sediment transport rate calculation method and system based on support vector machine algorithm

Abstract

The present invention discloses a method and system for calculating vertical sediment transport rate based on a support vector machine algorithm. The method comprises: collecting historical cross-section water and sediment distribution data and dividing it into a test set and a validation set based on data representativeness; calculating distribution parameters based on the historical cross-section water and sediment distribution data, including parameters of each vertical linear coefficient term, wake function term, bottom concentration term, and exponential distribution term; establishing a support vector machine (SVR) algorithm prediction model based on the test set and validation set, and constructing a predictive relationship between single-point flow velocity, sediment content, and water depth data and the distribution parameters; reconstructing the cross-section water and sediment distribution data using the support vector machine (SVR) algorithm prediction model, calculating the vertical sediment transport rate using the reconstructed cross-section water and sediment distribution data, and verifying the accuracy of the calculated vertical sediment transport rate; and calculating the vertical sediment transport rate online. The present invention achieves accurate reconstruction of the vertical sediment transport rate based on single-point data, providing technical support for efficient and real-time hydrological sediment transport monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological research and river management, and in particular to a method and system for calculating vertical sediment transport rate based on a support vector machine algorithm. Background Art

[0002] Accurately measuring vertical sediment transport rates is crucial for analyzing river erosion and deposition, predicting riverbed evolution, and assessing soil erosion. Traditionally, the six-point method has been widely used as a precise method for measuring vertical sediment transport. While this method, which involves sampling and analyzing six specific locations along a vertical line, can yield relatively accurate data, it requires manual sampling at each point, a cumbersome and time-consuming process that consumes significant human and material resources, making it difficult to meet the demands of real-time monitoring and rapid analysis.

[0003] With the development of online measurement technology, vertical sediment transport rate measurement is gradually moving towards automation and real-time, but a core problem has also become prominent. Online measurement requires inferring the vertical sediment transport rate based on the flow velocity and sediment content at a single point. Existing methods are generally based on direct linear relationship fitting. However, due to the complex nonlinear relationship between the vertical flow velocity distribution, sediment content distribution, and the single-point flow velocity and sediment content, the linear fitting accuracy is poor. This also means that online flow velocity and sediment content measurement equipment cannot meet the actual application requirements when used for hydrological analysis. Therefore, how to accurately derive the vertical sediment transport rate from single-point flow velocity and sediment content has become a key problem in improving the effectiveness of online measurement equipment. Summary of the Invention

[0004] In order to solve the problem in the prior art that the vertical sediment transport rate calculated from single-point flow velocity and sediment content has poor accuracy and cannot be used in practice, the present invention proposes a vertical sediment transport rate calculation method and system based on support vector machine algorithm, which can accurately calculate the vertical sediment transport rate online.

[0005] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0006] The method for calculating vertical sediment transport rate based on the support vector machine algorithm of the present invention includes the following operations:

[0007] Collect historical cross-section water and sediment distribution data and divide them into test and validation sets based on data representativeness;

[0008] The distribution parameters are calculated based on the historical cross-section water and sediment distribution data, including the linear coefficient term of each vertical line, the wake function term, the bottom concentration term and the exponential distribution term;

[0009] Based on the test set and validation set, a support vector machine (SVR) algorithm prediction model was established to construct the prediction relationship between single-point flow velocity, sediment content, and water depth data and distribution parameters.

[0010] The support vector machine (SVR) algorithm prediction model was used to reconstruct the cross-section water and sediment distribution data. The vertical sediment transport rate was calculated using the reconstructed cross-section water and sediment distribution data and compared with the measured vertical sediment transport rate to verify the accuracy of the calculated vertical sediment transport rate.

[0011] The flow velocity, sediment content and water depth data of a single point are detected online, and the vertical sediment transport rate is calculated based on the support vector machine (SVR) algorithm prediction model.

[0012] A further improvement of the present invention is that: historical section water and sediment distribution data is collected, and based on the representativeness of the data, a test set and a verification set are divided, including: collecting historical section water and sediment distribution data, dividing the historical section water and sediment distribution data into flood season section water and sediment distribution data and dry season section water and sediment distribution data, uniformly sampling the flood season section water and sediment distribution data and the dry season section water and sediment distribution data according to the vertical water depth of each section to obtain typical data, selecting 70% of the typical data as the test set, and the remaining 30% of the typical data as the verification set, wherein: the historical section water and sediment distribution data includes the section vertical flow velocity distribution, sediment content distribution, and water depth data, wherein the water depth data is the vertical water depth of each section, the sediment content distribution is the six-point sediment content of each vertical line, the section vertical flow velocity distribution is the six-point flow velocity, the six-point sediment content and the six-point flow velocity are the sediment content and flow velocity of the points on the surface, with relative water depths of 0.2H, 0.4H, 0.6H, 0.8H, and the bottom layer, wherein H is the vertical water depth.

[0013] A further improvement of the present invention is that the distribution parameters are calculated based on the historical cross-section water and sediment distribution data, specifically including:

[0014] According to the vertical velocity distribution of the cross section, the linear coefficient term and the wake function term are calculated using linear fitting, and the expression is:

[0015] ;

[0016] ;

[0017] ;

[0018] in, is the friction flow velocity, is the wake function term, is the linear coefficient term, For water depth, is the vertical height, i.e. the distance from the bed surface, is a constant term, is the flow rate;

[0019] According to the sediment content distribution and water depth data, the bottom concentration term and the exponential distribution term are calculated, and the expression is:

[0020] ;

[0021] ;

[0022] ;

[0023] The relationship between the bottom concentration of sediment when it is in vertical equilibrium distribution is expressed as:

[0024] ;

[0025] in, is the suspended matter concentration at the set location, i.e., the sediment content, is the bottom concentration term, is the negative value of the ratio of sediment settling velocity to sediment diffusion coefficient, that is, the exponential distribution term, is the settling velocity of sediment particles, As a reference point, is the coefficient of water turbulence diffusion or energy dissipation, is the height near the water surface or bed surface, is a natural constant.

[0026] A further improvement of the present invention is to establish a support vector machine (SVR) algorithm prediction model based on the test set and the validation set, and to construct a prediction relationship between the flow velocity, sediment content, and water depth data of a single point and the distribution parameters, specifically including:

[0027] According to the divided test set, determine the independent variables, including the flow velocity, sediment content and water depth at 0.6H of each vertical line of the test set;

[0028] Determine the dependent variables, including the linear coefficient term, wake function term, bottom concentration term, and exponential distribution term calculated at 0.6H of each vertical line in the test set;

[0029] Establish a support vector machine (SVR) algorithm prediction model, substitute independent variables and dependent variables to train the support vector machine (SVR) algorithm prediction model, and obtain model parameters;

[0030] Use the validation set to perform validation calculations on the support vector machine (SVR) algorithm prediction model;

[0031] Based on the verification calculation results, the performance evaluation of the support vector machine SVR algorithm prediction model is carried out.

[0032] A further improvement of the present invention is to reconstruct the cross-sectional water-sediment distribution data using the support vector machine (SVR) algorithm prediction model, calculate the vertical sediment transport rate using the reconstructed cross-sectional water-sediment distribution data, and compare it with the measured vertical sediment transport rate to verify the accuracy of the calculated vertical sediment transport rate, specifically including:

[0033] According to the measured cross-section water and sediment distribution data, the vertical average velocity and vertical average sediment concentration are calculated by the following formula, and the measured vertical sediment transport rate is calculated based on the vertical average velocity and vertical average sediment concentration;

[0034] ;

[0035] ;

[0036] The measured vertical sediment transport rate is:

[0037] ;

[0038] Where: is the vertical average velocity, Position on the vertical line The flow rate at is the relative water depth, with values ​​of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, where H is the vertical water depth. 、 、 、 、 、 They are the flow rates at relative water depths of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, respectively. is the vertical average sediment content, Position on the vertical line The sand content at is the layer weight coefficient, for the surface layer and the bottom layer, Take 1, for the rest of the middle layers, Take 2, For point;

[0039] According to the support vector machine (SVR) algorithm prediction model, the distribution parameters are calculated using the flow velocity, sediment content, and water depth data at 0.6H.

[0040] Based on the calculated distribution parameters, the vertical velocity distribution and sediment content distribution of the cross section are reconstructed to determine the six-point sediment content and six-point velocity. The vertical average sediment content and vertical average velocity are determined by vertical integral calculation. The reconstructed vertical sediment transport rate is obtained from the vertical average sediment content and vertical average velocity.

[0041] The root mean square error and Nash-Sutcliffe efficiency coefficient of the difference between the reconstructed vertical sediment transport rate and the measured vertical sediment transport rate were calculated to evaluate the accuracy of the support vector machine (SVR) algorithm prediction model.

[0042] The vertical sediment transport rate calculation system based on the support vector machine algorithm of the present invention includes:

[0043] The acquisition module is used to collect historical cross-section water and sediment distribution data and divide the data into test and validation sets based on the representativeness of the data;

[0044] Parameter calculation module, used to calculate distribution parameters based on historical cross-section water and sediment distribution data, including linear coefficient terms of each vertical line, wake function terms, bottom concentration terms and exponential distribution terms;

[0045] The model building module is used to establish a support vector machine (SVR) algorithm prediction model based on the test set and validation set, and to construct the prediction relationship between single-point flow velocity, sediment content, and water depth data and distribution parameters;

[0046] The model verification module is used to reconstruct the cross-section water and sediment distribution data using the support vector machine (SVR) algorithm prediction model, calculate the vertical sediment transport rate using the reconstructed cross-section water and sediment distribution data, and compare it with the measured vertical sediment transport rate to verify the accuracy of the calculated vertical sediment transport rate;

[0047] The online calculation module is used to detect the flow velocity, sediment content and water depth data of a single point online, and calculate the vertical sediment transport rate based on the support vector machine (SVR) algorithm prediction model.

[0048] A further improvement of the present invention is that: the operations performed by the acquisition module include: collecting historical section water and sediment distribution data, dividing the historical section water and sediment distribution data into two sequences of flood season section water and sediment distribution data and dry season section water and sediment distribution data, uniformly sampling the flood season section water and sediment distribution data and the dry season section water and sediment distribution data according to the vertical water depth of each section to obtain typical data, selecting 70% of the typical data as a test set, and the remaining 30% of the typical data as a verification set, wherein: the historical section water and sediment distribution data includes the section vertical flow velocity distribution, sediment content distribution, and water depth data, wherein the water depth data is the vertical water depth of each section, the sediment content distribution is the six-point sediment content of each vertical line, the section vertical flow velocity distribution is the six-point flow velocity, the six-point sediment content and the six-point flow velocity are the sediment content and flow velocity of the points on the surface, with relative water depths of 0.2H, 0.4H, 0.6H, 0.8H, and the bottom layer, wherein H is the vertical water depth.

[0049] A further improvement of the present invention is that the operations performed by the parameter calculation module specifically include:

[0050] According to the vertical velocity distribution of the cross section, the linear coefficient term and the wake function term are calculated using linear fitting, and the expression is:

[0051] ;

[0052] ;

[0053] ;

[0054] in, is the friction flow velocity, is the wake function term, is the linear coefficient term, For water depth, is the vertical height, i.e. the distance from the bed surface, is a constant term, is the flow rate;

[0055] According to the sediment content distribution and water depth data, the bottom concentration term and the exponential distribution term are calculated, and the expression is:

[0056] ;

[0057] ;

[0058] ;

[0059] The relationship between the bottom concentration of sediment when it is in vertical equilibrium distribution is expressed as:

[0060] ;

[0061] in, is the suspended matter concentration at the set location, i.e., the sediment content, is the bottom concentration parameter, is the negative value of the ratio of sediment settling velocity to sediment diffusion coefficient, that is, the exponential distribution term, is the settling velocity of sediment particles, As a reference point, is the coefficient of water turbulence diffusion or energy dissipation, is the height near the water surface or bed surface, is a natural constant.

[0062] A further improvement of the present invention is that the specific operations performed by the model building module include:

[0063] According to the divided test set, determine the independent variables, including the flow velocity, sediment content and water depth at 0.6H of each vertical line of the test set;

[0064] Determine the dependent variables, including the linear coefficient term, wake function term, bottom concentration term, and exponential distribution term calculated at 0.6H of each vertical line in the test set;

[0065] Establish a support vector machine (SVR) algorithm prediction model, substitute independent variables and dependent variables to train the support vector machine (SVR) algorithm prediction model, and obtain model parameters;

[0066] Use the validation set to perform validation calculations on the support vector machine (SVR) algorithm prediction model;

[0067] Based on the verification calculation results, the performance evaluation of the support vector machine SVR algorithm prediction model is carried out.

[0068] A further improvement of the present invention is that the specific operations performed by the model verification module include:

[0069] According to the measured cross-section water and sediment distribution data, the vertical average velocity and vertical average sediment concentration are calculated by the following formula, and the measured vertical sediment transport rate is calculated based on the vertical average velocity and vertical average sediment concentration;

[0070] ;

[0071] ;

[0072] The measured vertical sediment transport rate is:

[0073] ;

[0074] Where: is the vertical average velocity, Position on the vertical line The flow rate at is the relative water depth, with values ​​of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, where H is the vertical water depth. 、 、 、 、 、 They are the flow rates at relative water depths of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, respectively. is the vertical average sediment content, Position on the vertical line The sand content at is the layer weight coefficient, for the surface layer and the bottom layer, Take 1, for the rest of the middle layers, Take 2, For point;

[0075] According to the support vector machine (SVR) algorithm prediction model, the distribution parameters are calculated using the flow velocity, sediment content, and water depth data at 0.6H.

[0076] Based on the calculated distribution parameters, the vertical velocity distribution and sediment content distribution of the cross section are reconstructed to determine the six-point sediment content and six-point velocity. The vertical average sediment content and vertical average velocity are determined by vertical integral calculation. The reconstructed vertical sediment transport rate is obtained from the vertical average sediment content and vertical average velocity.

[0077] The root mean square error and Nash-Sutcliffe efficiency coefficient of the difference between the reconstructed vertical sediment transport rate and the measured vertical sediment transport rate were calculated to evaluate the accuracy of the support vector machine (SVR) algorithm prediction model.

[0078] The present invention has the following beneficial effects: By establishing a support vector machine (SVR) algorithm prediction model, the present invention performs fitting and optimization of relevant distribution parameters, determines the optimal measurement point based on the distribution parameter fitting, and reconstructs cross-sectional water and sediment distribution data to achieve accurate calculation of vertical sediment transport rate. This invention solves the key issue of linking vertical point selection and sediment transport rate in online measurement, achieving accurate reconstruction of vertical sediment transport rate based on single-point data, and providing technical support for efficient and real-time hydrological sediment transport monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention;

[0080] Figure 2 1 is a schematic diagram comparing the actual value and the predicted value of the linear coefficient term in an embodiment of the present invention;

[0081] Figure 3 3 is a comparison chart of the reconstructed vertical sediment transport rate and the measured vertical sediment transport rate in an embodiment of the present invention. DETAILED DESCRIPTION

[0082] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0083] like Figure 1 As shown, the vertical sediment transport rate calculation method based on the support vector machine algorithm of this embodiment includes the following steps:

[0084] Step 1: Collect historical cross-section water and sediment distribution data and divide it into test and validation sets based on data representativeness. The historical cross-section water and sediment distribution data includes cross-section vertical velocity distribution, sediment content distribution, and water depth data. The water depth data is the water depth of each vertical line of the cross-section, the sediment content distribution is the six-point sediment content of each vertical line, and the cross-section vertical velocity distribution is the six-point velocity. The six-point sediment content and six-point velocity are the sediment content and velocity of the surface point at relative water depths of 0.2H, 0.4H, 0.6H, 0.8H, and the bottom layer, where H is the vertical water depth.

[0085] Historical cross-section water and sediment distribution data were divided into two sets: flood season cross-section water and sediment distribution data and dry season cross-section water and sediment distribution data. Typical data were uniformly sampled based on the water depth of each vertical line of the cross-section. 70% of the typical data was selected as the test set, and the remaining 30% as the validation set. Tables 1 and 2 show the cross-section water and sediment distribution data for 2018 and 2019.

[0086] Table 1: Cross-section water and sediment distribution data in 2018

[0087]

[0088] Table 2: Cross-section water and sediment distribution data in 2019

[0089]

[0090] The samples selected in this embodiment are shown in Table 3 and Table 4:

[0091] Table 3: Dry season sampling data

[0092]

[0093] Table 4: Sampling data during flood season

[0094]

[0095] Step 2: Calculate the distribution parameters based on the historical cross-section water and sediment distribution data, including the linear coefficient term of each vertical line, the wake function term, the bottom concentration term, and the exponential distribution term. The specific steps include:

[0096] Step 2.1, determine the linear coefficient term and the wake function term. This includes giving the calculation expression of the cross-section vertical velocity distribution in logarithmic form:

[0097] ;

[0098] in, is the friction flow velocity, is the flow velocity, is the equivalent roughness height of the bed surface, which is 0.01 times the water depth, i.e. 0.01h. For water depth, is the vertical height, i.e. the distance from the bed surface, is a constant term.

[0099] After conversion:

[0100] ;

[0101] ;

[0102] ;

[0103] in, is the wake function term, is the linear coefficient term.

[0104] Step 2.2, determine the bottom concentration term and exponential distribution term.

[0105] ;

[0106] ;

[0107] ;

[0108] The relationship between the bottom concentration of sediment when it is in vertical equilibrium distribution is expressed as:

[0109] ;

[0110] in, is the suspended matter concentration at the set location, i.e., the sediment content, is the bottom concentration term, is the negative value of the ratio of sediment settling velocity to sediment diffusion coefficient, that is, the exponential distribution term, is the settling velocity of sediment particles, As a reference point, is the coefficient of water turbulence diffusion or energy dissipation, is the height near the water surface or bed surface, is a natural constant.

[0111] In step 2.3, based on the historical cross-section water and sediment distribution data, the linear coefficient term, tail function term, bottom concentration term, and exponential distribution term of each vertical line are calculated.

[0112] According to the vertical velocity distribution of the cross section, the corresponding linear coefficient term and wake function term are obtained:

[0113] According to the velocity data of six points on the vertical line, namely the surface layer, 0.2H, 0.4H, 0.6H, 0.8H and the bottom layer, the , use linear fitting to find the linear coefficients of each vertical line of the cross section , wake function term .

[0114] According to the sediment content of the six points of surface layer, 0.2H, 0.4H, 0.6H, 0.8H and bottom layer water depth on the vertical line, the sediment content of the six points is respectively , use linear fitting to find the bottom concentration of each vertical line of the cross section and exponential distribution terms .

[0115] Step 3: Based on the test set and validation set, a support vector machine (SVR) algorithm prediction model is established to construct the prediction relationship between single-point flow velocity, sediment content, and water depth data and distribution parameters. The specific steps are:

[0116] Step 3.1, select a representative point on the vertical line. The representative point on the vertical line can be taken at the middle position of the vertical water depth. In this embodiment, the point at a water depth of 0.6H is selected, and the flow velocity, sediment content and water depth value at 0.6H are used as independent variables, and substituted into the support vector machine SVR algorithm for fitting calculation.

[0117] In step 3.2, a small-sample machine learning algorithm (SVR) is used to establish a predictive relationship between the velocity and sediment concentration at 0.6 H and the linear coefficient, wake function, bottom concentration, and exponential distribution parameters. In practice, the main parameters to be adjusted are the regularization parameter (BoxConstraint), kernel function scaling factor (KernelScale), and maximum tolerance error (Epsilon).

[0118] Given a training dataset , is the independent variable, is the dependent variable, in this invention Including water depth, sediment content, flow velocity, It includes the linear coefficient terms of each vertical line, the wake function term, the bottom concentration term and the exponential distribution term.

[0119] The goal of the support vector machine (SVR) algorithm prediction model is to find the function ,in, For the sample, is the weight vector, is a bias term, satisfying:

[0120] Predicted value The deviation from the true value does not exceed , is the maximum tolerable error.

[0121] Through slack variables and Processing Exceeded Sample.

[0122] Objective function and constraints:

[0123] ;

[0124] ;

[0125] in: is the weight vector, which controls the direction of the regression hyperplane. is the regularization parameter, balancing the weight of model complexity and training error, 、 is a slack variable that allows samples to exceed the interval band ( is the maximum tolerable error).

[0126] Introducing Lagrange multipliers 、 (corresponding to the constraints) and 、 (corresponding to the non-negativity of the slack variables), construct the Lagrangian function:

[0127] ;

[0128] in: is the sample, that is, the independent variable, is the dependent variable, is the sample ordinal number, is the bias term, is the total number of samples.

[0129] Substituting the above conditions into the Lagrangian function, we get the dual problem:

[0130] ;

[0131] in: 、 、 、 is the Lagrange multiplier, For the sample, is the sample ordinal number.

[0132] Constraints:

[0133] ;

[0134] in: Is the kernel function, used to handle nonlinear regression. When the corresponding sample is the support vector.

[0135] The final regression function is:

[0136] ;

[0137] The bias term Calculated by support vector:

[0138] 1. Choose to satisfy or Sample .

[0139] 2. Utilize Solve .

[0140] The support vector machine (SVR) algorithm prediction model can flexibly handle linear and nonlinear regression problems while ensuring the generalization ability of the model.

[0141] According to the divided test set, the independent variables are the velocity, sediment content and water depth at 0.6H of each vertical line of the test set, and the dependent variables are the linear coefficient term parameters, wake function term parameters, bottom concentration term parameters and exponential distribution term calculated at 0.6H of each vertical line, and the support vector machine SVR prediction model is established. Substitute the independent variables and dependent variables to train the support vector machine SVR algorithm prediction model and obtain the regularization parameter , kernel function scaling factor K and maximum tolerance error Three parameters. Use the validation set to perform validation calculations on the support vector machine (SVR) algorithm prediction model. Based on the validation calculation results, calculate the performance indicators of the validation set and test set results, including the correlation index R. 2 And the mean square error indicator MSE.

[0142] R 2 The calculation expression is:

[0143] ;

[0144] in: is the residual sum of squares, which represents the deviation between the model prediction value and the true value; is the total sum of squares, which represents the total variability of the true values ​​relative to their mean; is the mean of the true values.

[0145] The MSE formula is:

[0146] ;

[0147] in: Indicates the The true value of the sample, Indicates the The predicted value of samples, N represents the total number of samples.

[0148] If the correlation is less than the set value or the mean square error is greater than the set value, adjust the regularization parameter , kernel function scaling factor K and maximum tolerance error After taking the value, recalculate until the correlation index R 2 And the mean square error index MSE meets the requirements.

[0149] like Figure 2 As shown, the linear coefficient term For example, get the optimal parameters: , , .

[0150] Step 4: Use the support vector machine (SVR) algorithm prediction model to reconstruct the cross-sectional water and sediment distribution data, calculate the vertical sediment transport rate using the reconstructed cross-sectional water and sediment distribution data, and compare it with the measured vertical sediment transport rate to verify the accuracy of the calculated vertical sediment transport rate.

[0151] Step 4.1: Calculate the vertical average velocity and vertical average sediment concentration based on the measured cross-section water and sediment distribution data, and calculate the measured vertical sediment transport rate based on the vertical average velocity and vertical average sediment concentration;

[0152] ;

[0153] ;

[0154] The measured vertical sediment transport rate is:

[0155] ;

[0156] Where: is the vertical average velocity, Position on the vertical line The flow rate at is the relative water depth, with values ​​of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, where 0 represents the surface layer, 1H represents the bottom layer, and H is the vertical water depth. 、 、 、 、 、 They are the flow rates at relative water depths of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, respectively. is the vertical average sediment content, Position on the vertical line The sand content at is the layer weight coefficient, for the surface layer and the bottom layer, Take 1, for the rest of the middle layers, Take 2, It is a point position with a value of 1 to 6, corresponding to the relative water depth: 0, 0.2H, 0.4H, 0.6H, 0.8H, 1H.

[0157] Step 4.2: Calculate the linear coefficient term based on the support vector machine (SVR) algorithm prediction model using the velocity, sediment content, and water depth data at 0.6H. , wake function term , bottom concentration item and exponential distribution terms .

[0158] According to steps 2.1 and 2.2, reconstruct the vertical velocity distribution and sediment content distribution, determine the velocity and sediment content at each vertical water depth point, calculate and determine the vertical average sediment content and vertical average velocity, and obtain the reconstructed vertical sediment transport rate from the vertical average sediment content and vertical average velocity.

[0159] Reconstructed results of flow velocity and sediment concentration distribution along the water depth direction:

[0160] ;

[0161] ;

[0162] Reconstructed vertical average velocity and vertical average sediment content for:

[0163] ;

[0164] ;

[0165] Reconstructed vertical sediment transport rate for:

[0166] .

[0167] Step 4.3: Verify the accuracy of the SVR prediction model based on the reconstructed vertical sediment transport rate. This includes calculating the root mean square error (RRMSE) and Nash-Sutcliffe efficiency (NSE) of the difference between the reconstructed vertical sediment transport rate and the measured vertical sediment transport rate to evaluate the accuracy of the SVR prediction model.

[0168] The calculation expression of root mean square error (RMSE) is:

[0169] ;

[0170] Where N is the total number of samples, Indicates the The true value of the sample, that is, the measured vertical sediment transport rate, Indicates the The predicted value of the sample is the reconstructed vertical sediment transport rate.

[0171] ;

[0172] in, is the mean of the true values, that is, the mean of the measured vertical sediment transport rates.

[0173] The calculation expression of Nash-Sutcliffe efficiency coefficient (NSE) is:

[0174] ;

[0175] in, is the mean of the true values, that is, the mean of the measured vertical sediment transport rates.

[0176] The calculation results of the sediment transport rate (reconstructed vertical sediment transport rate) and the verification sediment transport rate (measured vertical sediment transport rate) are shown in the following table:

[0177] Table 5: Reconstructed vertical line and calculation results of verified sediment transport rate

[0178]

[0179] The calculation results of MSE and NSE are shown in Figure 3 In the figure, the horizontal coordinate of each data point represents the calculated sediment transport rate, and the vertical coordinate represents the verified sediment transport rate. When the data point basically falls on the 45° line, it means that the calculated sediment transport rate is equal to the verified sediment transport rate. The data point deviates and falls below the 45° line, which means that the calculated sediment transport rate is larger than the verified sediment transport rate. The data point deviates and falls above the 45° line, which means that the calculated sediment transport rate is smaller than the verified sediment transport rate. When RRMSE<0.3 and NSE>0.75, the support vector machine SVR algorithm prediction model is evaluated to have a higher accuracy. The calculation results of this embodiment are: MSE=0.7955, RRMSE=0.23, NSE=0.954.

[0180] Step 5: Detect the velocity, sediment content, and water depth data of a single point online, and calculate the vertical sediment transport rate based on the support vector machine (SVR) algorithm prediction model.

[0181] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art and, unless defined similarly as herein, will not be interpreted in an idealized or overly formal sense.

[0182] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for calculating vertical sediment transport rate based on support vector machine algorithm, characterized by: The following operations are included: Collect historical cross-section water and sediment distribution data and divide them into test and validation sets based on data representativeness; The distribution parameters are calculated based on the historical cross-section water and sediment distribution data, including the linear coefficient term of each vertical line, the wake function term, the bottom concentration term and the exponential distribution term; Based on the test set and validation set, a support vector machine (SVR) algorithm prediction model was established to construct the prediction relationship between single-point flow velocity, sediment content, and water depth data and distribution parameters. The support vector machine (SVR) algorithm prediction model was used to reconstruct the cross-section water and sediment distribution data. The vertical sediment transport rate was calculated using the reconstructed cross-section water and sediment distribution data and compared with the measured vertical sediment transport rate to verify the accuracy of the calculated vertical sediment transport rate. Online detection of single-point flow velocity, sediment content and water depth data, and calculation of vertical sediment transport rate based on the support vector machine (SVR) algorithm prediction model; The support vector machine (SVR) algorithm prediction model was used to reconstruct the cross-sectional water and sediment distribution data. The vertical sediment transport rate was calculated using the reconstructed cross-sectional water and sediment distribution data. The vertical sediment transport rate was compared with the measured vertical sediment transport rate to verify the accuracy of the calculated vertical sediment transport rate. Specifically, the following methods were used: According to the measured cross-section water and sediment distribution data, the vertical average velocity and vertical average sediment concentration are calculated by the following formula, and the measured vertical sediment transport rate is calculated based on the vertical average velocity and vertical average sediment concentration; ； ； The measured vertical sediment transport rate is: ； Where: is the vertical average velocity, Position on the vertical line The flow rate at is the relative water depth, with values ​​of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, where H is the vertical water depth. 、 、 、 、 、 They are the flow rates at relative water depths of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, respectively. is the vertical average sediment content, Position on the vertical line The sand content at is the layer weight coefficient, for the surface layer and the bottom layer, Take 1, for the rest of the middle layers, Take 2, For point; According to the support vector machine (SVR) algorithm prediction model, the distribution parameters are calculated using the flow velocity, sediment content, and water depth data at 0.6H. Based on the calculated distribution parameters, the vertical velocity distribution and sediment content distribution of the cross section are reconstructed to determine the six-point sediment content and six-point velocity. The vertical average sediment content and vertical average velocity are determined by vertical integral calculation. The reconstructed vertical sediment transport rate is obtained from the vertical average sediment content and vertical average velocity. The root mean square error and Nash-Sutcliffe efficiency coefficient of the difference between the reconstructed vertical sediment transport rate and the measured vertical sediment transport rate were calculated to evaluate the accuracy of the support vector machine (SVR) algorithm prediction model.

2. The method for calculating vertical sediment transport rate based on the support vector machine algorithm according to claim 1, wherein: The method of collecting historical section water and sediment distribution data and dividing the data into a test set and a validation set based on data representativeness includes: collecting historical section water and sediment distribution data, dividing the historical section water and sediment distribution data into flood season section water and sediment distribution data and dry season section water and sediment distribution data, uniformly sampling the flood season section water and sediment distribution data and the dry season section water and sediment distribution data according to the vertical water depth of each section to obtain typical data, selecting 70% of the typical data as the test set, and the remaining 30% of the typical data as the validation set, wherein: the historical section water and sediment distribution data includes the section vertical flow velocity distribution, sediment content distribution, and water depth data, wherein the water depth data is the vertical water depth of each section, the sediment content distribution is the six-point sediment content of each vertical line, the section vertical flow velocity distribution is the six-point flow velocity, the six-point sediment content and the six-point flow velocity are the sediment content and flow velocity of the surface point with relative water depths of 0.2H, 0.4H, 0.6H, 0.8H, and the bottom layer, wherein H is the vertical water depth.

3. The method for calculating vertical sediment transport rate based on support vector machine algorithm according to claim 2, characterized in that: The calculation of distribution parameters based on historical cross-section water and sediment distribution data specifically includes: According to the vertical velocity distribution of the cross section, the linear coefficient term and the wake function term are calculated using linear fitting, and the expression is: ； ； ； in, is the friction flow velocity, is the wake function term, is the linear coefficient term, For water depth, is the vertical height, i.e. the distance from the bed surface, is a constant term, is the flow rate; According to the sediment content distribution and water depth data, the bottom concentration term and the exponential distribution term are calculated, and the expression is: ； ； ； The relationship between the bottom concentration of sediment when it is in vertical equilibrium distribution is expressed as: ； in, is the suspended matter concentration at the set location, i.e., the sediment content, is the bottom concentration term, is the negative value of the ratio of sediment settling velocity to sediment diffusion coefficient, that is, the exponential distribution term, is the settling velocity of sediment particles, As a reference point, is the coefficient of water turbulence diffusion or energy dissipation, is the height near the water surface or bed surface, is a natural constant.

4. The method for calculating vertical sediment transport rate based on support vector machine algorithm according to claim 2, wherein: Based on the test set and validation set, a support vector machine (SVR) algorithm prediction model was established to construct the prediction relationship between single-point flow velocity, sediment content, and water depth data and distribution parameters, including: According to the divided test set, determine the independent variables, including the flow velocity, sediment content and water depth at 0.6H of each vertical line of the test set; Determine the dependent variables, including the linear coefficient term, wake function term, bottom concentration term, and exponential distribution term calculated at 0.6H of each vertical line in the test set; Establish a support vector machine (SVR) algorithm prediction model, substitute independent variables and dependent variables to train the support vector machine (SVR) algorithm prediction model, and obtain model parameters; Use the validation set to perform validation calculations on the support vector machine (SVR) algorithm prediction model; Based on the verification calculation results, the performance evaluation of the support vector machine SVR algorithm prediction model is carried out.

5. The vertical sediment transport rate calculation system based on the support vector machine algorithm is characterized by: include: The acquisition module is used to collect historical cross-section water and sediment distribution data and divide the data into test and validation sets based on the representativeness of the data; Parameter calculation module, used to calculate distribution parameters based on historical cross-section water and sediment distribution data, including linear coefficient terms of each vertical line, wake function terms, bottom concentration terms and exponential distribution terms; The model building module is used to establish a support vector machine (SVR) algorithm prediction model based on the test set and validation set, and to construct the prediction relationship between single-point flow velocity, sediment content, and water depth data and distribution parameters; The model verification module is used to reconstruct the cross-section water and sediment distribution data using the support vector machine (SVR) algorithm prediction model, calculate the vertical sediment transport rate using the reconstructed cross-section water and sediment distribution data, and compare it with the measured vertical sediment transport rate to verify the accuracy of the calculated vertical sediment transport rate; Online calculation module, used to detect the flow velocity, sediment content and water depth data of a single point online, and calculate the vertical sediment transport rate based on the support vector machine (SVR) algorithm prediction model; The specific operations performed by the model validation module include: According to the measured cross-section water and sediment distribution data, the vertical average velocity and vertical average sediment concentration are calculated by the following formula, and the measured vertical sediment transport rate is calculated based on the vertical average velocity and vertical average sediment concentration; ； ； The measured vertical sediment transport rate is: ； Where: is the vertical average velocity, Position on the vertical line The flow rate at is the relative water depth, with values ​​of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, where H is the vertical water depth. 、 、 、 、 、 They are the flow rates at relative water depths of 0, 0.2H, 0.4H, 0.6H, 0.8H, and 1H, respectively. is the average sediment content of the vertical line, is the position on the vertical line The sand content at is the layer weight coefficient, for the surface layer and the bottom layer, Take 1, for the rest of the middle layers, Take 2, For point; According to the support vector machine (SVR) algorithm prediction model, the distribution parameters are calculated using the flow velocity, sediment content, and water depth data at 0.6H. Based on the calculated distribution parameters, the vertical velocity distribution and sediment content distribution of the cross section are reconstructed to determine the six-point sediment content and six-point velocity. The vertical average sediment content and vertical average velocity are determined by vertical integral calculation. The reconstructed vertical sediment transport rate is obtained from the vertical average sediment content and vertical average velocity. The root mean square error and Nash-Sutcliffe efficiency coefficient of the difference between the reconstructed vertical sediment transport rate and the measured vertical sediment transport rate were calculated to evaluate the accuracy of the support vector machine (SVR) algorithm prediction model.

6. The vertical sediment transport rate calculation system based on the support vector machine algorithm according to claim 5 is characterized in that: The operations performed by the acquisition module include: collecting historical section water and sediment distribution data, dividing the historical section water and sediment distribution data into two sequences of flood season section water and sediment distribution data and dry season section water and sediment distribution data, uniformly sampling the flood season section water and sediment distribution data and the dry season section water and sediment distribution data according to the vertical water depth of each section to obtain typical data, selecting 70% of the typical data as a test set, and the remaining 30% of the typical data as a validation set, wherein: the historical section water and sediment distribution data includes the section vertical flow velocity distribution, sediment content distribution, and water depth data, wherein the water depth data is the vertical water depth of each section, the sediment content distribution is the six-point sediment content of each vertical line, the section vertical flow velocity distribution is the six-point flow velocity, the six-point sediment content and the six-point flow velocity are the sediment content and flow velocity of the points on the surface, with relative water depths of 0.2H, 0.4H, 0.6H, 0.8H, and the bottom layer, wherein H is the vertical water depth.

7. The vertical sediment transport rate calculation system based on the support vector machine algorithm according to claim 6 is characterized in that: The operations performed by the parameter calculation module specifically include: According to the vertical velocity distribution of the cross section, the linear coefficient term and the wake function term are calculated using linear fitting, and the expression is: ； ； ； in, is the friction flow velocity, is the wake function term, is the linear coefficient term, For water depth, is the vertical height, i.e. the distance from the bed surface, is a constant term, is the flow rate; According to the sediment content distribution and water depth data, the bottom concentration term and the exponential distribution term are calculated, and the expression is: ； ； ； The relationship between the bottom concentration of sediment when it is in vertical equilibrium distribution is expressed as: ； in, is the suspended matter concentration at the set location, i.e., the sediment content, is the bottom concentration term, is the negative value of the ratio of sediment settling velocity to sediment diffusion coefficient, that is, the exponential distribution term, is the settling velocity of sediment particles, As a reference point, is the coefficient of water turbulence diffusion or energy dissipation, is the height near the water surface or bed surface, is a natural constant.

8. The vertical sediment transport rate calculation system based on the support vector machine algorithm according to claim 6 is characterized in that: The specific operations performed by the model building module include: According to the divided test set, determine the independent variables, including the flow velocity, sediment content and water depth at 0.6H of each vertical line of the test set; Determine the dependent variables, including the linear coefficient term, wake function term, bottom concentration term, and exponential distribution term calculated at 0.6H of each vertical line in the test set; Establish a support vector machine (SVR) algorithm prediction model, substitute independent variables and dependent variables to train the support vector machine (SVR) algorithm prediction model, and obtain model parameters; Use the validation set to perform validation calculations on the support vector machine (SVR) algorithm prediction model; Based on the verification calculation results, the performance evaluation of the support vector machine SVR algorithm prediction model is carried out.

Images