Hydrological forecasting method, medium and equipment for correcting surface runoff simulation precision

Abstract

The invention discloses a hydrological forecasting method, medium and equipment for correcting surface runoff simulation precision, and relates to the technical field of hydrological forecasting. The method comprises the following steps: acquiring hydrological driving data, and establishing a hydrological model; constructing a machine learning model to calculate the dynamic contribution weight and contribution water volume of the surface runoff simulation value to the sub-basin outlet flow simulation value; calculating a proportion correction coefficient K by applying a proportion correction method according to the observed value and the simulated value of the outlet flow of the sub-basin; according to the grid surface runoff simulation value and the contribution water volume of the grid surface runoff simulation value to the sub-basin outlet flow simulation value, a reverse correction coefficient C is calculated; an error correction factor alpha is introduced, the grid surface runoff simulation value is corrected through the proportion correction coefficient K, the reverse correction coefficient C and the error correction factor alpha, the corrected grid surface runoff simulation value is used for correcting the hydrological model, and the corrected hydrological model is used for hydrological prediction. The simulation precision of the drainage basin outlet flow can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological prediction, and particularly relates to a hydrological forecasting method, medium, and device for correcting the simulation accuracy of surface runoff. Background Art

[0002] Improving the accuracy of hydrological forecasting is of crucial significance for earthquake prevention and disaster reduction, optimal allocation of water resources, water ecological protection, and even regional sustainable development. Process-based distributed hydrological models (such as the distributed VIC model, Variable Infiltration Capacity) are the main tools for realizing hydrological forecasting. However, although process-based hydrological models have strong hydrological forecasting capabilities, there are still many challenges in further improving their hydrological forecasting accuracy. The most important problem is that the simulation accuracy of surface runoff (R) in space is relatively low, which directly affects the simulation accuracy of discharge (Q).

[0003] There is a close dependence relationship between the spatial distribution characteristics of surface runoff and discharge. Specifically, the water volume after attenuation such as interception and infiltration of surface runoff is the main source and direct contributor to the formation of discharge at the basin outlet. Therefore, improving the simulation accuracy of surface runoff will greatly improve the simulation accuracy of discharge, and thus improve the hydrological forecasting accuracy. However, the insufficient simulation accuracy of surface runoff in existing hydrological models stems from multiple factors, including the uncertainty of meteorological driving data, the complexity of model structure, the role of nonlinear hydrological processes, and the intertwined influence of internal model processes. In addition, due to the lack of high-precision spatial surface runoff observation data, it is impossible to effectively verify the accuracy of surface runoff simulated by process-based hydrological models, resulting in the insufficient improvement of the spatial simulation accuracy of discharge.

[0004] In terms of improving the spatial simulation accuracy of discharge in hydrological models, existing methods mainly include direct correction methods and indirect correction methods. The direct correction method adjusts the simulated value of discharge, such as by calibrating parameters for the basin to improve model output. The indirect correction method indirectly improves the simulation accuracy of discharge by correcting other hydrological variables that affect discharge, such as evapotranspiration and soil moisture. However, these methods do not fully consider the core contribution of surface runoff to the simulation accuracy of discharge, especially the lag effect and attenuation effect generated by surface runoff during the confluence process. The lag effect means that surface runoff does not immediately converge to the basin outlet in space, and the attenuation effect is that surface runoff gradually decreases in the basin due to processes such as infiltration and evaporation. Therefore, the contribution of surface runoff to discharge is often lower than the directly simulated value and will decay with changes in time and space. This makes the method of only calibrating model parameters in existing methods ignore the lag and attenuation effects, often resulting in large errors in the correction results. Summary of the Invention

[0005] The object of the present invention is to propose a hydrological forecasting method for correcting the simulation accuracy of surface runoff, which can solve the problem that the lag effect and attenuation effect generated during the confluence process of surface runoff are not fully considered in the existing surface runoff simulation process. The method includes the following steps:

[0006] S1. Obtain hydrological driving data, establish a hydrological model, and collect the observed values of the outlet flow of the sub-basin measured at the hydrological station, the simulated values of the outlet flow of the sub-basin simulated by the hydrological model, and the simulated values of the grid surface runoff.

[0007] S2. According to the simulated values of the outlet flow of the sub-basin and the simulated values of the grid surface runoff simulated by the hydrological model, construct a machine learning model to calculate the dynamic contribution weight and contribution water volume of the simulated values of the grid surface runoff to the simulated values of the outlet flow of the sub-basin.

[0008] S3. Calculate the proportional correction coefficient K according to the observed values and simulated values of the outlet flow of the sub-basin; calculate the reverse correction coefficient C according to the simulated values of the grid surface runoff and the contribution water volume of the simulated values of the grid surface runoff to the simulated values of the outlet flow of the sub-basin; introduce an error correction factor α, and use the proportional correction coefficient K, the reverse correction coefficient C, and the error correction factor α to correct the simulated values of the grid surface runoff.

[0009] S4. Use the corrected simulated values of the grid surface runoff to correct the hydrological model, and apply the corrected hydrological model for hydrological prediction.

[0010] Furthermore, the hydrological driving data includes: precipitation, maximum temperature, minimum temperature, wind speed, vegetation type and corresponding attributes, soil type and corresponding attributes, and global files for runoff generation and confluence of the hydrological model.

[0011] Furthermore, the machine learning model is a random forest model, and the dynamic contribution weight of the simulated values of the grid surface runoff to the simulated values of the outlet flow of the sub-basin is expressed as:

[0012]

[0013] where, ΔMSE(X j ,t) represents the change in the mean squared error at node t when considering feature X j , X j represents the daily surface runoff simulated value of the j-th grid, MSE(t) represents the mean squared error of node t, N 1 and N 2 respectively represent the sample numbers of the sub-nodes t 1 and t 2 after splitting, N t represents the sample number of node t, MSE(t 1 ) and MSE(t 2respectively represent the two child nodes t after splitting 1 and t 2 of the mean square error, ω i represents the dynamic contribution weight of the simulated surface runoff value of the i-th grid to the simulated sub-basin outlet flow value;

[0014] The contribution water volume of the simulated surface runoff value of each grid to the simulated sub-basin outlet flow value is expressed as:

[0015] R Qi = ω i × R i

[0016] where, R Qi represents the contribution water volume of the simulated surface runoff value of the i-th grid to the simulated sub-basin outlet flow value; R i represents the simulated surface runoff value of the i-th grid, and ω i represents the dynamic contribution weight of the simulated surface runoff value of the i-th grid to the simulated sub-basin outlet flow value.

[0017] Furthermore, the calculation formula of the proportional correction coefficient K is:

[0018]

[0019] where, Q 0 represents the observed value of the sub-basin outlet flow, and Q sim represents the simulated value of the sub-basin outlet flow.

[0020] Furthermore, the reverse correction coefficient C is expressed as:

[0021]

[0022] where, R Qi represents the contribution water volume of the simulated surface runoff value of the i-th grid to the simulated sub-basin outlet flow value; R i represents the simulated surface runoff value of the i-th grid.

[0023] Furthermore, the expression of the corrected grid simulated surface runoff value is as follows:

[0024] R ′ oi = K × R i × (1 + α × C)

[0025] where, R ′ oi represents the corrected grid simulated surface runoff value, K is the proportional correction coefficient, R i represents the simulated surface runoff value of the i-th grid, α is the error correction factor, and C is the reverse correction coefficient.

[0026] Furthermore, the error correction factor α is selected based on making the simulated outlet flow value of the sub-basin of the corrected hydrological model closest to the observed value of the outlet flow of the sub-basin.

[0027] The present invention also provides a computer-readable storage medium storing a computer program, which when executed by a processor implements the above hydrological forecasting method for correcting the simulation accuracy of surface runoff.

[0028] The present invention also provides an electronic device, including a processor and a memory, the processor is interconnected with the memory, wherein the memory is used to store a computer program, the computer program includes computer-readable instructions, and the processor is configured to call the computer-readable instructions to execute the above hydrological forecasting method for correcting the simulation accuracy of surface runoff.

[0029] The beneficial effects brought by the technical solution provided by the present invention are as follows:

[0030] Based on the flow observation data and the output results of the model, the present invention combines relevant technologies such as machine learning, introduces a proportional correction coefficient, a reverse correction coefficient and an error correction factor to correct the simulated value of surface runoff. Using the reverse correction coefficient and the error correction factor, dynamically adjust the contribution of each sub-basin and grid cell, reduce the lag time and attenuation effect in the surface runoff concentration process, more accurately reflect the change of surface runoff in the actual basin, and improve the simulation accuracy of the outlet flow of the basin. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a flowchart of the hydrological forecasting method for correcting the simulation accuracy of surface runoff according to an embodiment of the present invention;

[0032] Figure 2 is a block diagram of an electronic device in an exemplary embodiment of Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0033] To make the objectives, technical solutions and advantages of the present invention clearer, the embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

[0034] The flowchart of the hydrological forecasting method for correcting the simulation accuracy of surface runoff according to the embodiment of the present invention is as Figure 1 , and specifically includes the following steps:

[0035] S1. Obtain hydrological driving data, establish a hydrological model, and collect the observed value Q of the outlet flow of the sub-basin measured at the hydrological station 0 , the simulated value Q of the outlet flow of the sub-basin simulated by the hydrological model sim and the simulated value R of the grid surface runoff i .

[0036] In the embodiment of the present invention, the established hydrological model is a large-scale distributed VIC hydrological model. Specifically, the obtained hydrological driving data includes: precipitation, maximum temperature, minimum temperature, wind speed, vegetation type and corresponding attributes, soil type and corresponding attributes, and global files for runoff generation and confluence in the hydrological model.

[0037] S2. According to the simulated values of the sub-basin outlet flow and the grid surface runoff simulated by the hydrological model, construct a machine learning model to calculate the dynamic contribution weight of each grid surface runoff simulated value to the sub-basin outlet flow simulated value, and estimate the contribution water volume of each grid surface runoff simulated value to the sub-basin outlet flow simulated value.

[0038] The machine learning model in the embodiment of the present invention can select a support vector machine, a decision tree model, or a random forest model. When the random forest model is selected, a functional relationship between the grid surface runoff simulated value and the sub-basin outlet flow simulated value is established based on this model, and the contribution of the surface runoff simulated value of each grid to the sub-basin outlet flow simulated value is calculated. The random forest model uses the daily surface runoff simulated values of each grid as the independent variable X and the daily sub-basin outlet flow simulated value as the dependent variable Y for training, with a monthly training period, and outputs the contribution weight ω of each grid. i 。

[0039] The contribution weight ω of each grid point to the sub-basin outlet flow simulated value i is represented by the feature importance function of the random forest. The dynamic contribution weight of each grid surface runoff simulated value to the sub-basin outlet flow simulated value is expressed as:

[0040]

[0041] where, ΔMSE(X j ,t) represents the change in the mean squared error of node t when considering feature X j , X j represents the daily surface runoff simulated value of the jth grid, MSE(t) represents the mean squared error of node t, N 1 and N 2 respectively represent the sample numbers of the child nodes t 1 and t 2 after splitting, N t represents the sample number of node t, MSE(t 1 ) and MSE(t 2 ) respectively represent the mean squared errors of the two child nodes t 1 and t 2 after splitting, ω i represents the dynamic contribution weight of the surface runoff simulated value of the ith grid to the sub-basin outlet flow simulated value, and T represents the number of nodes.

[0042] According to the obtained weight ω i Calculate the amount of water contributed by the surface runoff simulation value of each grid to the sub - basin outlet flow simulation value. The amount of water contributed by the surface runoff simulation value of each grid to the sub - basin outlet flow simulation value is expressed as:

[0043] R Qi = ω i ×R i

[0044] where R Qi represents the amount of water contributed by the surface runoff simulation value of the i - th grid to the sub - basin outlet flow simulation value; R i represents the surface runoff simulation value of the i - th grid, and ω i represents the dynamic contribution weight of the surface runoff simulation value of the i - th grid to the sub - basin outlet flow simulation value.

[0045] S3. According to the observed value and simulation value of the sub - basin outlet flow, calculate the proportional correction coefficient K using the proportional correction method; according to the surface runoff simulation value of the grid and the amount of water contributed by the surface runoff simulation value of the grid to the sub - basin outlet flow simulation value, calculate the reverse correction coefficient C; introduce the error correction factor α, and use the proportional correction coefficient K, the reverse correction coefficient C, and the error correction factor α to correct the surface runoff simulation value of the grid.

[0046] As a preferred solution of the present invention, the calculation formula of the proportional correction coefficient K is:

[0047]

[0048] where Q 0 represents the observed value of the sub - basin outlet flow, and Q sim represents the simulation value of the sub - basin outlet flow.

[0049] Calculate the amount of water contributed by the preliminarily corrected surface runoff simulation value of the grid to the sub - basin outlet flow simulation value according to the proportional correction coefficient K, and calculate the preliminarily corrected surface runoff simulation value of the grid.

[0050] The amount of water contributed by the preliminarily corrected surface runoff simulation value of the grid to the sub - basin outlet flow simulation value calculated according to the proportional correction coefficient K is expressed as:

[0051]

[0052] In order to make the estimated value close to R oi , it is necessary to amplify the value of R Qi to make it close to R i , so the following basic relational expression is obtained:

[0053]

[0054] Obtain the preliminarily corrected grid surface runoff simulation value: R oi = K × R i .

[0055] Among them, represents the preliminary correction value of the contribution water volume of the surface runoff simulation value of the i-th grid to the sub-basin outlet flow simulation value; R Qi represents the contribution water volume of the surface runoff simulation value of the i-th grid to the sub-basin outlet flow simulation value; R oi represents the surface runoff simulation value of the i-th grid after preliminary correction; R i represents the surface runoff simulation value of the i-th grid.

[0056] According to the estimated contribution water volume of the surface runoff simulation value of each grid to the sub-basin outlet flow simulation value, introduce the reverse correction coefficient C to solve the error caused by the lag time and attenuation effect during the process of surface runoff reaching the sub-basin outlet, and correct the grid surface runoff simulation value after preliminary correction again to obtain the second corrected grid surface runoff simulation value.

[0057] The reverse correction coefficient C is expressed as:

[0058]

[0059] Among them, R Qi represents the contribution water volume of the surface runoff simulation value of the i-th grid to the sub-basin outlet flow simulation value; R i represents the surface runoff simulation value of the i-th grid.

[0060] The second corrected grid surface runoff simulation value is expressed as:

[0061] S4. The reverse correction coefficient C may not be able to completely offset the attenuation difference caused by the different distances of the surface runoff simulation values of each grid flowing to the sub-basin outlet. Therefore, introduce the error correction factor α to control the contribution degree of each sub-basin to the total flow, reflect the differences during the runoff simulation process, optimize the grid surface runoff value after the second correction to obtain the finally corrected grid surface runoff value, and evaluate the improvement effect of the finally corrected grid surface runoff on the simulation accuracy of the hydrological model through the two-way verification method to obtain the corrected hydrological model, and finally apply the corrected hydrological model for hydrological prediction.

[0062] The expression of the finally corrected grid surface runoff value simulation is as follows:

[0063] R ′oi = K × R i × (1 + α × C)

[0064] Wherein, R ′ oi represents the finally corrected grid surface runoff simulation value, K is the proportional correction coefficient, and R i represents the surface runoff simulation value of the i-th grid, α is the error correction factor, and C is the reverse correction coefficient.

[0065] The selection of the error correction factor α is based on making the simulated value of the sub-basin outlet flow closest to the observed value of the sub-basin outlet flow after being corrected by α. And the correction results are verified in a top-down and bottom-up manner. The top-down verification method is to correct the upstream sites and use the uncorrected downstream sites to verify the model accuracy. The bottom-up verification method is to correct the downstream hydrological sites and use the uncorrected upstream sites to verify the model accuracy.

[0066] In an exemplary embodiment, there is provided a computer-readable storage medium storing a computer program, which when executed by a processor implements the above-mentioned hydrological forecasting method for correcting the simulation accuracy of surface runoff.

[0067] Please refer to Figure 2 , in an exemplary embodiment, there is further provided an electronic device including at least one processor, at least one memory, and at least one communication bus.

[0068] Wherein, a computer program is stored on the memory, and the computer program includes computer-readable instructions. The processor calls the computer-readable instructions stored in the memory through the communication bus to execute the above-mentioned hydrological forecasting method for correcting the simulation accuracy of surface runoff.

[0069] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A hydrological forecasting method for correcting surface runoff simulation accuracy, characterized in that: The following steps are involved: S1. Obtain hydrological driving data, establish a hydrological model, and collect the observed values ​​of sub-basin outlet flow measured by hydrological stations, the simulated values ​​of sub-basin outlet flow simulated by the hydrological model, and the simulated values ​​of grid surface runoff; S2. Based on the simulated values ​​of the sub-basin outlet flow and the grid surface runoff simulated values ​​simulated by the hydrological model, a machine learning model is constructed to calculate the dynamic contribution weight and contribution water volume of the grid surface runoff simulated value to the sub-basin outlet flow simulated value; S3. Calculate the proportional correction coefficient K by using the proportional correction method according to the observed and simulated values ​​of the outlet flow of the sub-basin; calculate the inverse correction coefficient C according to the grid surface runoff simulation value and the contribution of the grid surface runoff simulation value to the sub-basin outlet flow simulation value; The error correction factor α is introduced, and the grid surface runoff simulation value is corrected by the proportional correction coefficient K, the inverse correction coefficient C and the error correction factor α; S4. Use the corrected grid surface runoff simulation values ​​to correct the hydrological model, and use the corrected hydrological model to perform hydrological prediction.

2. A hydrological forecasting method for correcting surface runoff simulation accuracy according to claim 1, characterized in that: Hydrological driving data include: precipitation, maximum temperature, minimum temperature, wind speed, vegetation type and corresponding attributes, soil type and corresponding attributes, and global files for runoff generation and confluence of hydrological models.

3. A hydrological forecasting method for correcting surface runoff simulation accuracy according to claim 1, characterized in that: The machine learning model is a random forest model, and the dynamic contribution weight of the grid surface runoff simulation value to the sub-basin outlet flow simulation value is expressed as: Among them, ΔMSE(X j ,t) indicates that node t is considering feature X j The change in mean square error when X j represents the daily surface runoff simulation value of the jth grid, MSE(t) represents the mean square error of node t, N1 and N2 represent the number of samples of child nodes t1 and t2 after splitting, respectively, and N t represents the number of samples of node t, MSE(t1) and MSE(t2) represent the mean square error of the two child nodes t1 and t2 after splitting, respectively, ω i represents the dynamic contribution weight of the surface runoff simulation value of the i-th grid to the outlet flow simulation value of the sub-basin, and T represents the number of nodes; The contribution of each grid surface runoff simulation value to the sub-basin outlet flow simulation value is expressed as: R Qi =ω i ×R i Among them, R Qi R represents the contribution of the simulated surface runoff value of the i-th grid to the simulated flow value of the sub-basin outlet; i represents the simulated value of surface runoff of the ith grid, ω i Represents the dynamic contribution weight of the surface runoff simulation value of the i-th grid to the sub-basin outlet flow simulation value.

4. A hydrological forecasting method for correcting surface runoff simulation accuracy according to claim 1, characterized in that: The calculation formula of the proportional correction coefficient K is: Among them, Q0 represents the observed flow at the outlet of the sub-basin, Q sim Represents the simulated value of the sub-basin outlet flow.

5. A hydrological forecasting method for correcting surface runoff simulation accuracy according to claim 1, characterized in that: The inverse correction coefficient C is expressed as: Among them, R Qi R represents the contribution of the simulated surface runoff value of the i-th grid to the simulated flow value of the sub-basin outlet; i Represents the simulated value of surface runoff of the i-th grid.

6. A hydrological forecasting method for correcting surface runoff simulation accuracy according to claim 1, characterized in that: The modified grid surface runoff simulation value expression is as follows: R ′ oi =K×R i ×(1+α×C) Among them, R ′ oi represents the modified grid surface runoff simulation value, K is the proportional correction coefficient, R i represents the simulated value of surface runoff of the ith grid, α is the error correction factor, and C is the reverse correction coefficient.

7. A hydrological forecasting method for correcting surface runoff simulation accuracy according to claim 1, characterized in that: The error correction factor α is selected based on the principle that the simulated value of the sub-basin outlet flow simulated by the modified hydrological model is closest to the observed value of the sub-basin outlet flow.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program comprises computer-readable instructions, and the processor is configured to call the computer-readable instructions to execute the method according to any one of claims 1 to 7.

Images