CN115905963A - A flood forecasting method and system based on support vector machine model - Google Patents

Abstract

The invention discloses a flood forecasting method and system based on a support vector machine model, belonging to the field of runoff forecasting; the method comprises the steps of S1, collecting hydrological meteorological data of a research area; s2, determining a current generation module parameter CN value of the SCS-CN model according to the collected hydrological meteorological data; s3, improving the flow generation module of the SCS-CN model by utilizing the rainfall characteristic factor and the flow generation module parameter CN value; s4, establishing an SCS-CN three-water-source confluence model by coupling the improved production flow module and the improved three-water-source confluence module of the S3; s5, constructing a dynamic convergence time parameter through a support vector machine model to improve an SCS-CN three-water-source convergence model; s6, forecasting a flood process line of the outlet section of the drainage basin by using the actually measured rainfall data; the influence of rainfall characteristic factors is considered, and the prediction capability of the model in a region with a relatively dry climate is improved; factors such as a spatial distribution condition of rainfall in a field and early rainfall are considered, so that the risk of prediction deviation of the hydrological model caused by uneven spatial distribution of rainfall is reduced.

Description

A flood forecasting method and system based on support vector machine model

Technical Field

The invention belongs to the field of runoff forecasting, and in particular relates to a flood forecasting method and system based on a support vector machine model.

Background Art

In recent years, machine learning models have developed rapidly and have been widely used in flood variable prediction and flood sensitivity assessment. Since the construction of machine learning models does not require a large amount of data, machine learning-based methods have certain advantages in predicting flood variables. In previous related studies, machine learning models have been successfully used to predict flood variable-related problems.

The SCS-CN model is a conceptual model developed by the United States Department of Agriculture. It only requires two parameters, the number of curves and the initial loss rate, to predict runoff. Because it requires less input data, has a relatively simple model structure, and has high prediction accuracy, it is widely used to estimate surface runoff from rainfall events in small watersheds. However, this model is a conceptual model developed based on runoff data from U.S. watersheds. When applied to research areas under different climate types, it is necessary to comprehensively consider local vegetation, topography, soil and other factors. After continuous improvements by domestic scholars, the SCS-CN model has achieved good results in domestic watersheds, but the traditional SCS-CN model only has a runoff generation module and does not include a confluence module. Therefore, when rainfall occurs, the changing trend of the flow at the outlet section of the watershed cannot be drawn.

Most conceptual hydrological models treat the basin as a whole and study the operation and changes of the runoff generation mechanism of the entire basin after a rainfall event. Since conceptual hydrological models do not have the ability to calculate basin cells, when the spatial distribution of rainfall is uneven, it is easy to cause a large deviation between the peak flood time predicted by the conceptual hydrological model and the actual peak flood time, especially in the study area where the basin is long and narrow or the basin area is large.

Summary of the invention

In view of the deficiencies of the prior art, the object of the present invention is to provide a flood forecasting method and system based on a support vector machine model.

The purpose of the present invention can be achieved through the following technical solutions:

A flood forecasting method based on a support vector machine model comprises the following steps:

S1, collect hydrological and meteorological data of the study area;

S2, based on the collected hydrological and meteorological data, determine the CN value of the flow generation module parameter of the SCS-CN model;

S3, using rainfall characteristic factors and runoff module parameter CN value to improve the runoff module of SCS-CN model;

S4, by coupling the improved flow generation module of S3 and the three-water source confluence module, the SCS-CN three-water source confluence model is established;

S5, constructing dynamic confluence hysteresis parameters through support vector machine model to improve the SCS-CN three-source confluence model;

S6, use the measured rainfall data to predict the flood process line of the basin outlet section.

Furthermore, the hydrological and meteorological data collected in S1 include: rainfall intensity, runoff, soil type and land use type.

Furthermore, in S2, the step of determining the flow generation module parameters is:

S21, determine the hydrological soil group through the soil type data of the study area and divide it according to the infiltration capacity; resample the land use type of the study area on the ArcGIS platform to match it with the grid under each soil type, calculate the CN value in each grid, and then perform weighted average of the CN value of each grid to finally determine the comprehensive CN value of the basin;

S22, based on the total rainfall in the previous five days, the soil moisture conditions were divided into three levels: dry, normal and wet. K-mean clustering was used to re-divide the interval of soil moisture conditions. The total rainfall in the previous five days was the previous influencing rainfall, and the threshold was set to 120 mm.

S23, the CN value is converted according to the actual soil moisture conditions in the study area, and the mutation intervals between CN ₁ , CN ₂ and CN ₃ are interpolated to obtain the CN values under different antecedent rainfall.

Furthermore, in S3, the improvement of the SCS-CN runoff model is to add the influence of rainfall characteristic factors on the basis of the original SCS-CN model. The specific steps are as follows:

S31, in the runoff module, the previous rainfall needs to supplement the initial water shortage of the basin. If the total rainfall does not meet the initial water shortage, the basin will not produce runoff. The calculation formula for the runoff generated by the rainfall is:

Where: P is the total rainfall, mm; Q is the surface runoff, mm; _Ia is the initial loss, mm; S is the potential water storage capacity, mm;

S32, the calculation formula for initial loss and potential water storage capacity is:

I _a = λS (2)

Where: λ is the initial loss rate; CN is a dimensionless parameter;

S33, add characteristic factors such as rainfall intensity in the runoff generation module, and the revised runoff calculation formula is as follows:

为场次降雨的平均雨强，mm/h；β为雨强修正参数；

为降雨特征因子的综合体现。Where: I ₆₀ is the maximum 60-min rainfall intensity during the rainfall event, mm/h;

is the average rainfall intensity of the rainfall event, mm/h; β is the rainfall intensity correction parameter;

It is a comprehensive reflection of rainfall characteristic factors.

Furthermore, in S4, the process curve of the runoff depth in the study area can be obtained through the SCS-CN runoff generation module, and the runoff depth generated in the study area in each period can be obtained by differential processing, and then input into the three water source confluence module for confluence calculation.

Furthermore, the three water sources refer to: surface runoff, subsoil flow and underground runoff, and are divided by parabolic free water storage curves and free water reservoirs.

Furthermore, in the above-mentioned confluence calculation, the surface water confluence calculation adopts the unit line method, using the unit line to simulate the watershed as a series connection of n linear reservoirs; the confluence calculation of subsoil flow and groundwater runoff adopts the linear reservoir method; the river network confluence adopts the hysteresis algorithm, and its calculation formula is:

Q3(I)＝CS×Q3(I-1)+(1-CS)×QT3(I-L) (5)

QT3(I-L)＝QS(I-L)+QG(I-L)+QI(I-L) (6)

Where: Q3(I) is the unit area river network runoff in the Ith period, ^m3 /s; CS is the river network flow recession coefficient; L is the runoff hysteresis parameter, h; QS is the surface runoff, ^m3 /s; QG is the groundrunoff, ^m3 /s; QI is the soil flow, ^m3 /s; QT3(IL) is the sum of the surface runoff, groundrunoff and soil flow in the ILth period, ^m3 /s.

Further, in S5, the regression variables of the dynamic confluence hysteresis parameters include: the length of the rainfall center in the study area from the basin section, the slope of the rainfall center, the total rainfall, the previous impact rainfall and the total river length; the regression function f(x) of the dynamic confluence hysteresis parameters is:

Where: (α _i ,α _i ^* ) is the Lagrange multiplier; b is the bias; K(x, _xi ) is the radial basis function from the input space to the high-order feature space, that is, the kernel function; σ is the expansion constant of the radial basis function, which represents the radial range of the function.

Furthermore, in S6, first, the measured rainfall data and the runoff parameters of S2, S3, and S4 are input into the SCS-CN three-water source confluence model in S4 to obtain the runoff of each calculation grid in the study area, and then the runoff is input into the improved three-water source confluence model in S5 to obtain the flood process line of the basin outlet section.

A flood forecasting system based on a support vector machine model, comprising:

Data collection unit: collects hydrological and meteorological data of the study area;

Runoff module parameter solving unit: Determine the CN value of the runoff module parameter of the SCS-CN model based on the collected hydrological and meteorological data;

Runoff module improvement unit: Use rainfall characteristic factors and runoff module parameter CN values to improve the runoff module of the SCS-CN model;

SCS-CN three-water source confluence model construction unit: by coupling the improved flow generation module and the three-water source confluence module, the SCS-CN three-water source confluence model is established;

SCS-CN three-water source confluence model improvement unit: Improve the SCS-CN three-water source confluence model by constructing dynamic confluence hysteresis parameters through the support vector machine model;

Flood forecasting unit: forecast flood process lines at the outlet section of the basin using measured rainfall data;

Beneficial effects of the present invention:

1. The present invention re-divides the division interval of the previous soil moisture conditions by using K-mean mean cluster analysis, and uses cubic linear interpolation to obtain the CN value under different previous influencing rainfall. The influence of rainfall characteristic factors such as rainfall intensity and average rainfall intensity is considered, which solves the problem that the rainfall intensity is not considered in the original runoff module, and improves the prediction ability of the model in relatively arid climate areas;

2. The present invention uses the support vector machine model to construct dynamic basin runoff hysteresis parameters, taking into account factors such as the spatial distribution of rainfall events, previous rainfall and underlying surface conditions, so as to reduce the risk of prediction deviation of the hydrological model due to uneven spatial distribution of rainfall, improve the ability of the hydrological model to predict the peak time, and alleviate the deficiencies and defects of the conceptual hydrological model in the prediction process to a certain extent, which can provide a solution for the conceptual hydrological model when it is applied to complex research basins;

3. The present invention obtains the flood process line of the basin outlet section by coupling the SCS-CN flow generation module and the three-water source confluence module. The SCS-CN three-water source confluence model has high accuracy in simulating peak flow and peak time, which expands the application scope of the SCS-CN model.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

FIG1 is a flow chart of the steps of a flood forecasting method;

Figure 2 is a schematic diagram of the distribution of watershed stations in a village;

Figure 3 shows the distribution of land use (left) and soil types (right) in a village watershed;

FIG4 is a simulation diagram of dynamic confluence hysteresis parameters based on the support vector machine model;

Figure 5 shows the simulation results of some events of the SCS-CN three-source confluence model.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

As shown in FIG1 , a flood forecasting method based on a support vector machine model includes the following steps:

S1. Collect hydrological and meteorological data of the study area; including data on rainfall intensity, runoff, soil type and land use type.

S2, determine the runoff module parameters of the SCS-CN model based on relevant data such as soil type and land use type in the study area; the specific steps are:

S21, the hydrological soil group is determined by the soil texture, soil infiltration rate and other parameters in the study area. The hydrological soil group can be divided into four categories: A, B, C and D, and their infiltration capacity decreases in turn. The land use type of the study area is resampled on the ArcGIS platform to match it with the grid under each soil type, and the CN (Curve Number) value in each grid is calculated. The CN value of each grid is then weighted averaged to finally determine the comprehensive CN value of the basin (soil moisture is normal).

S22, the SCS-CN model divides soil moisture conditions (AMC) into three levels, namely, dry (AMC1), normal (AMC2), and wet (AMC3), based on the total rainfall in the previous five days;

The accuracy of determining the previous soil moisture conditions will directly affect the prediction accuracy and stability of the hydrological model. This study statistically analyzed the relationship between the total rainfall in the five days before the flood and the suitable CN value, and used K-mean clustering to re-divide the interval of soil moisture conditions. The total rainfall in the first five days was the previous influencing rainfall, and the threshold was set to 120 mm.

S23, the CN value is converted according to the actual soil moisture conditions in the study area; since there are mutation points in the delineation method of soil moisture in the SCS-CN model, the mutation intervals between CN ₁ , CN ₂ and CN ₃ are interpolated. CN ₁ is the CN value when the soil is relatively dry, CN ₂ is the CN value when the soil is normal (that is, the CN value obtained by S21), and CN ₃ is the CN value when the soil is relatively moist.

S3, using characteristic factors such as rainfall intensity and average rainfall intensity to improve the runoff generation module of the SCS-CN model;

The improvement of SCS-CN runoff model is to add the influence of rainfall characteristic factors on the basis of the original SCS-CN model;

The SCS-CN model is an empirical model based on the water balance equation and two basic assumptions: the first assumption is that the ratio of surface runoff to the possible maximum runoff is equal to the actual infiltration volume and the potential water storage capacity; the second assumption is that there is a certain proportional relationship between the initial loss and the potential water storage capacity;

In the runoff module, the previous rainfall needs to supplement the initial water shortage of the basin. If the total rainfall does not meet the initial water shortage, the basin will not produce runoff. The calculation formula for the runoff generated by the rainfall is:

Where: P is the total rainfall; Q is the surface runoff; _Ia is the initial loss; S is the potential water storage capacity; the calculation formula for initial loss and potential water storage capacity is:

I _a = λS (2)

Where: λ is the initial loss rate; CN is a dimensionless parameter that can be obtained from S2. The larger the CN, the greater the runoff capacity of the basin. Its value is related to factors such as the land use type, vegetation cover type and previous rainfall in the study area. At the same time, the CN value is also a relatively sensitive parameter in the runoff generation module. Changes in the CN value will have a greater impact on the prediction results of the study basin.

In semi-arid areas, the soil water level is low and the soil water shortage is large. After a rainfall, it may be difficult for the basin to reach the field water holding capacity. The rainfall intensity will exceed the infiltration intensity before runoff occurs. Therefore, the rainfall intensity in the area will play a major role in the runoff process. In this regard, characteristic factors such as rainfall intensity are added to the runoff module to improve the prediction accuracy of the model in semi-arid areas. The revised runoff calculation formula is as follows:

为场次降雨的平均雨强，mm/h；β为雨强修正参数，需要通过前期洪水场次的降雨过程进行率定；

为总降雨量、降雨强度等降雨特征因子的综合体现。Where: I ₆₀ is the maximum 60-min rainfall intensity during the rainfall event, mm/h;

is the average rainfall intensity of the rainfall event, mm/h; β is the rainfall intensity correction parameter, which needs to be calibrated through the rainfall process of the previous flood events;

It is a comprehensive reflection of rainfall characteristic factors such as total rainfall and rainfall intensity.

S4, by coupling the improved flow generation module of S3 and the three-water source confluence module, the SCS-CN three-water source confluence model is established;

The process curve of the runoff depth in the study area can be obtained through the SCS-CN runoff generation module. The runoff depth generated in the study area in each period can be obtained by differential processing, and then input into the three-water source confluence module for confluence calculation; the three water sources refer to surface runoff, subsoil flow and underground runoff, which are divided by parabolic free storage curves and free water reservoirs.

In the confluence calculation, the unit line method is used for surface water confluence calculation. The unit line is used to simulate the watershed as a series of n linear reservoirs. The confluence calculation of subsurface flow and groundwater runoff adopts the linear reservoir method. The hysteresis algorithm is used for river network confluence. The calculation formula is:

Q3(I)＝CS×Q3(I-1)+(1-CS)×QT3(I-L) (5)

QT3(I-L)＝QS(I-L)+QG(I-L)+QI(I-L) (6)

Where: Q3(I) is the unit area river network runoff in the Ith period, ^m3 /s; CS is the river network flow recession coefficient; L is the runoff hysteresis parameter, h; QS is the surface runoff, ^m3 /s; QG is the groundrunoff, ^m3 /s; QI is the soil flow, ^m3 /s; QT3(IL) is the sum of the surface runoff, groundrunoff and soil flow in the ILth period, ^m3 /s.

S5, based on the SCS-CN three-water source confluence model obtained in S4, a dynamic confluence hysteresis parameter is constructed through a support vector machine model to improve the SCS-CN three-water source confluence model;

When the rainfall in the study basin is spatially uneven or the basin is long and narrow, the basin runoff hysteresis often changes. In the past, the runoff hysteresis of most hydrological models remained unchanged, which easily led to a large deviation between the peak flood time predicted by the hydrological model and the actual peak flood time.

Based on this phenomenon, the support vector machine model is used to construct the dynamic confluence hysteresis parameter (L). First, the factors that may affect the confluence hysteresis of the basin are screened out, and then the principal component analysis is performed to determine the main influencing factors. The final regression variables include the length of the rainfall center from the study area to the basin section, the slope of the rainfall center, the total rainfall, the previous influencing rainfall and the total river length;

The basic idea of the support vector machine is to use the kernel function to map the low-dimensional nonlinear space to the high-dimensional feature space, so that the low-dimensional nonlinear hydrological data can be separated in the high-dimensional space. After a series of derivations, its regression function f(x) can finally be expressed as:

Where: (α _i ,α _i ^* ) is the Lagrange multiplier; b is the bias; K(x, _xi ) is the radial basis function from the input space to the high-order feature space, that is, the kernel function; σ is the expansion constant of the radial basis function, which represents the radial range of the function.

After constructing the functional relationship between the confluence lag parameter and the regression variable through the support vector machine model, the confluence lag parameter of the flood is predicted according to the measured site data and the underlying surface data (including slope, river length, etc.), and then it is substituted into equations (5) and (6) to obtain the improved SCS-CN three-source confluence model.

S6, forecast the flood process line of the outlet section of the basin using the measured rainfall data;

Firstly, the measured rainfall data and the runoff parameters of S2, S3 and S4 are input into the SCS-CN three-source confluence model in S4 to obtain the runoff of each calculation grid in the study area. Then the runoff is input into the improved three-source confluence model in S5 to obtain the flood process line of the outlet section of the basin.

A flood forecasting system based on a support vector machine model, comprising: a data acquisition unit, a flow generation module parameter solving unit, a flow generation module improvement unit, an SCS-CN three-water source confluence model construction unit, an SCS-CN three-water source confluence model improvement unit and a flood forecasting unit;

in:

Data collection unit: collects hydrological and meteorological data of the study area;

Runoff module parameter solving unit: Determine the CN value of the runoff module parameter of the SCS-CN model based on the collected hydrological and meteorological data;

Runoff module improvement unit: Use rainfall characteristic factors and runoff module parameter CN values to improve the runoff module of the SCS-CN model;

SCS-CN three-water source confluence model construction unit: by coupling the improved flow generation module and the three-water source confluence module, the SCS-CN three-water source confluence model is established;

SCS-CN three-water source confluence model improvement unit: Improve the SCS-CN three-water source confluence model by constructing dynamic confluence hysteresis parameters through the support vector machine model;

Flood forecasting unit: forecast flood process lines at the outlet section of the basin using measured rainfall data;

Example:

The following is a specific example of applying the above flood forecasting method using a certain place:

The study area is located in a village basin, and the schematic diagram of the study basin is shown in Figure 2; the hydrological station is located downstream of the water area, with a catchment area of 745km2 and an average annual rainfall of about 500mm. There are 5 rainfall stations in the village basin, with a density of about 149km2/station. The average slope of the basin is 9.19%, which belongs to the temperate monsoon climate and semi-arid area, and the land use types are mainly cultivated land and forest land.

The hydrometeorological data of the village basin from 1957 to 2004 (flood season from June to September), including data on rainfall, runoff, evaporation and land use type, were obtained from the Provincial Hydrological Bureau. After reviewing the three properties of the rainfall and runoff data, 25 floods were finally selected, including the largest flood in the history of the village hydrological station (peak flow of 4520m3/s and peak water level of 57.78m).

The DEM data of the village watershed was downloaded from the geospatial data cloud, and the projection, resampling, filling, flow direction extraction, and flow extraction were performed in ArcGIS software to obtain the geospatial distribution map of the study area. The land use types in the study area were then resampled to match the grid under each soil type.

Based on the soil saturated infiltration rate, minimum infiltration rate and soil texture data of the village watershed, the village watershed was determined to be Class B hydrological soil group, and then the initial CN value of the study area was determined by the land use type of the study area (soil moisture is normal). The soil data comes from the FAO World Soil Type Database with a spatial resolution of 1000m*1000m, and the land use data comes from the global land cover product (GlobeLand30) with a spatial resolution of 30m*30m. Their distribution is shown in Figure 3.

The SCS-CN model determines soil moisture conditions based on the total rainfall in the previous five days, and there may be places with mutations. Therefore, this paper re-divides the soil moisture condition division interval of the original model using K-mean mean clustering, and interpolates the mutation intervals between CN ₁ , CN _2, and CN _3. CN ₁ is the CN value when the soil is relatively dry, and CN ₃ is the CN value when the soil is relatively wet. The CN values of the study area under different previous rainfall influencing the area are shown in Table 1:

Table 1. CN value of SCS-CN model

After determining the CN value of different sessions, the session runoff can be calculated by the following formula:

Where: P is the total rainfall, mm; Q is the surface runoff, mm; _Ia is the initial loss, mm; S is the potential water storage capacity, mm.

The initial loss and potential water storage capacity in the runoff module can be obtained using the following formula:

I _a = λS (2)

Where: λ is the initial loss rate.

In semi-arid areas, the soil water level is low and the soil water shortage is large. After a rainfall, it may be difficult for the basin to reach the field water holding capacity. Only when the rainfall intensity exceeds the infiltration intensity will runoff occur. For this reason, it is necessary to add characteristic factors such as rainfall intensity to the runoff module to improve the prediction accuracy of the model in the village basin. The revised runoff calculation formula is:

为场次降雨的平均雨强，mm/h；β为雨强修正参数；

为总降雨量、降雨强度等降雨特征因子的综合体现。Where: I ₆₀ is the maximum 60-min rainfall intensity during the rainfall event, mm/h;

is the average rainfall intensity of the rainfall event, mm/h; β is the rainfall intensity correction parameter;

It is a comprehensive reflection of rainfall characteristic factors such as total rainfall and rainfall intensity.

Through the above steps, the process curve of runoff depth in the study area can be obtained. Differential processing can be performed to obtain the runoff depth generated by the village basin in each period, and then input it into the three water source confluence module for confluence calculation.

The constant term (B) and the initial loss rate of the empirical relationship between potential water storage capacity and CN value have a certain impact on the runoff module. Therefore, in this example, the constant term and the initial loss rate of the empirical relationship between potential water storage capacity and CN value in the runoff module are calibrated. At the same time, the sensitive parameters KI, KG, CS, CI, CG, and SM in the confluence module are also calibrated. The model parameters are calibrated using 16 floods from 1957 to 1979, and the remaining 9 floods are used to verify the model. The parameter calibration results are shown in Table 2, and the simulation results are shown in Table 3:

Table 2. SCS-CN model parameter values

Table 3. Statistical indicators of the SCS-CN model simulation results for the village watershed

The total runoff is divided into three types: surface runoff, subsoil runoff and underground runoff through the parabola free storage curve, and the confluence module calculation is performed separately. At the same time, according to the concept of full storage runoff, runoff can only be generated on the runoff-generating area. All rainfall on the runoff-generating area generates runoff and enters the reservoir, and the rainfall becomes the free water reservoir recharge.

In the confluence calculation, the unit line method is used for surface water confluence calculation. The unit line is used to simulate the watershed as a series of n linear reservoirs. The confluence calculation of subsurface flow and groundwater runoff adopts the linear reservoir method. The hysteresis algorithm is used for river network confluence. The calculation formula is:

Q3(I)＝CS×Q3(I-1)+(1-CS)×QT3(I-L) (5)

QT3(I-L)＝QS(I-L)+QG(I-L)+QI(I-L) (6)

Where: Q3(I) is the unit area river network runoff in the Ith period, ^m3 /s; CS is the river network flow recession coefficient; L is the runoff hysteresis parameter, h; QS is the surface runoff, ^m3 /s; QG is the groundrunoff, ^m3 /s; QI is the soil flow, ^m3 /s; QT3(IL) is the sum of the surface runoff, groundrunoff and soil flow in the ILth period, ^m3 /s.

If the study area is set in a narrow and long basin or the basin area is large, it is often necessary to consider the spatial distribution of rainfall. In this example, the dynamic convergence hysteresis parameters are constructed by using the support vector machine model. The relevant variables include the length of the rainfall center from the basin section, the slope of the rainfall center, the total river length, the total rainfall, and the previous rainfall and other hydrological and meteorological characteristic parameters. After a series of derivations, the regression function can finally be expressed as:

Where: (α _i ,α _i ^* ) is the Lagrange multiplier; b is the bias; K(x, _xi ) is the radial basis function from the input space to the high-order feature space, that is, the kernel function; σ is the expansion constant of the radial basis function, which represents the radial range of the function.

The support vector machine model is often used to solve machine learning problems with small samples, and can simplify problems such as classification and regression. Since the support vector machine model is a supervised machine learning method, it requires certain previous data to build a database. Therefore, some previous flood data are used to build the model. The actual value of the model is the confluence hysteresis parameter when the peak time simulated by the hydrological model is the same as the actual peak time. The closer the predicted value is to the 1:1 straight line, the better the model prediction result. The prediction results are shown in Figure 4.

In summary, the SCS-CN three-source confluence model can be constructed through the above steps. The simulation results of some events are shown in Figure 5.

In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims (10)

1. A flood forecasting method based on a support vector machine model, characterized in that it comprises the following steps: S1, collect hydrometeorological data of the study area; S2, based on the collected hydrological and meteorological data, determine the CN value of the flow generation module parameter of the SCS-CN model; S3, using rainfall characteristic factors and runoff module parameter CN value to improve the runoff module of SCS-CN model; S4, by coupling the improved flow generation module of S3 and the three-water source confluence module, the SCS-CN three-water source confluence model is established; S5, constructing dynamic confluence hysteresis parameters through support vector machine model to improve the SCS-CN three-source confluence model; S6, use the measured rainfall data to predict the flood process line of the basin outlet section. 2. A flood forecasting method based on a support vector machine model according to claim 1, characterized in that the hydrological and meteorological data collected in S1 include: rainfall intensity, runoff, soil type and land use type. 3. A flood forecasting method based on a support vector machine model according to claim 2, characterized in that, in said S2, the step of determining the parameters of the flow generation module is: S21, determine the hydrological soil group through the soil type data of the study area and divide it according to the infiltration capacity; resample the land use type of the study area on the ArcGIS platform to match it with the grid under each soil type, calculate the CN value in each grid, and then perform weighted average of the CN value of each grid to finally determine the comprehensive CN value of the basin; S22, based on the total rainfall in the previous five days, the soil moisture conditions were divided into three levels: dry, normal and wet. K-mean clustering was used to re-divide the interval of soil moisture conditions. The total rainfall in the previous five days was the previous influencing rainfall, and the threshold was set to 120 mm. S23, the CN value is converted according to the actual soil moisture conditions in the study area, and the mutation intervals between CN ₁ , CN ₂ and CN ₃ are interpolated to obtain the CN values under different antecedent rainfall. 4. A flood forecasting method based on a support vector machine model according to claim 3, characterized in that, in S3, the improvement of the SCS-CN runoff model is to add the influence of rainfall characteristic factors on the basis of the original SCS-CN model, and the specific steps are: S31, in the runoff module, the previous rainfall needs to supplement the initial water shortage of the basin. If the total rainfall does not meet the initial water shortage, the basin will not produce runoff. The calculation formula for the runoff generated by the rainfall is:

Where: P is the total rainfall, mm; Q is the surface runoff, mm; _Ia is the initial loss, mm; S is the potential water storage capacity, mm; S32, the calculation formula for initial loss and potential water storage capacity is: I _a = λS (2)

Where: λ is the initial loss rate; CN is a dimensionless parameter; S33, add characteristic factors such as rainfall intensity in the runoff generation module, and the revised runoff calculation formula is as follows:

Where: I ₆₀ is the maximum 60-min rainfall intensity during the rainfall event, mm/h;

is the average rainfall intensity of the rainfall event, mm/h; β is the rainfall intensity correction parameter;

It is a comprehensive reflection of rainfall characteristic factors. 5. A flood forecasting method based on a support vector machine model according to claim 1, characterized in that, in S4, the process curve of the runoff depth in the study area can be obtained through the SCS-CN runoff generation module, and the runoff depth generated in the study area in each time period can be obtained by differential processing, and then it is input into the three water source confluence module for confluence calculation. 6. A flood forecasting method based on a support vector machine model according to claim 5, characterized in that the three water sources refer to: surface runoff, subsoil flow and underground runoff, and are divided by a parabolic free water storage curve and a free water reservoir. 7. A flood forecasting method based on a support vector machine model according to claim 5, characterized in that, in the runoff calculation, the surface water runoff calculation adopts the unit line method, using the unit line to simulate the watershed as a series connection of n linear reservoirs; the runoff calculation of subsurface flow and groundwater runoff adopts the linear reservoir method; the river network runoff adopts the hysteresis algorithm, and its calculation formula is: Q3(I)＝CS×Q3(I-1)+(1-CS)×QT3(I-L) (5) QT3(I-L)＝QS(I-L)+QG(I-L)+QI(I-L) (6) Where: Q3(I) is the unit area river network runoff in the Ith period, ^m3 /s; CS is the river network flow recession coefficient; L is the runoff hysteresis parameter, h; QS is the surface runoff, ^m3 /s; QG is the groundrunoff, ^m3 /s; QI is the soil flow, ^m3 /s; QT3(IL) is the sum of the surface runoff, groundrunoff and soil flow in the ILth period, ^m3 /s. 8. A flood forecasting method based on a support vector machine model according to claim 1, characterized in that the regression variables of the dynamic confluence hysteresis parameter include: the length of the rainfall center of the study area from the basin section, the slope of the rainfall center, the total rainfall, the previous impact rainfall and the total river length; the regression function f(x) of the dynamic confluence hysteresis parameter is:

Where: (α _i ,α _i ^* ) is the Lagrange multiplier; b is the bias; K(x, _xi ) is the radial basis function from the input space to the high-order feature space, that is, the kernel function; σ is the expansion constant of the radial basis function, which represents the radial range of the function. 9. A flood forecasting method based on a support vector machine model according to claim 1, characterized in that, in S6, first, the measured rainfall data and the runoff parameters of S2, S3, and S4 are input into the SCS-CN three-water source confluence model in S4 to obtain the runoff of each calculation grid in the study area, and then the runoff is input into the improved three-water source confluence model in S5 to obtain the flood process line of the basin outlet section. 10. A flood forecasting system based on a support vector machine model, characterized by comprising: Data collection unit: collects hydrological and meteorological data of the study area; Runoff module parameter solving unit: Determine the CN value of the runoff module parameter of the SCS-CN model based on the collected hydrological and meteorological data; Runoff module improvement unit: Use rainfall characteristic factors and runoff module parameter CN values to improve the runoff module of the SCS-CN model; SCS-CN three-water source confluence model construction unit: by coupling the improved flow generation module and the three-water source confluence module, the SCS-CN three-water source confluence model is established; SCS-CN three-water source confluence model improvement unit: Improve the SCS-CN three-water source confluence model by constructing dynamic confluence hysteresis parameters through the support vector machine model; Flood forecasting unit: Use measured rainfall data to forecast flood process lines at the outlet section of the basin.

Images