Hydrological model parameter calibration method based on normalized Nash efficiency coefficient

By normalizing the Nash efficiency coefficient, normalized Nash efficiency coefficient is formed as the objective function of parameter rate determination of hydrological model, which solves the problem that a single objective function is difficult to comprehensively evaluate the model performance, and achieves higher quality parameter rate determination and simulation result accuracy.

CN119004253AInactive Publication Date: 2024-11-22CHINA INST OF WATER RESOURCES & HYDROPOWER RES
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410996468.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the existing hydrological model parameter rate determination method, it is difficult for a single objective function to comprehensively evaluate the model performance, and the value range and optimization direction of different objective functions are different, which affects the optimization effect.

Method used

The hyperbolic tangent function and mathematical transformation are used to normalize the Nash efficiency coefficients to form the normalized Nash efficiency coefficients as the multi-objective optimization objective function, and realize the dimensionless and order-of-magnitude similarity between different dimension objective functions.

Benefits of technology

Through normalization processing, the value range of the objective function is limited, allowing the multi-objective optimization algorithm to find higher quality solutions, and improve the accuracy and consistency of the parameter rate determination of the hydrological model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004253A_ABST
    Figure CN119004253A_ABST
Patent Text Reader

Abstract

The invention provides a normalized Nash efficiency coefficient-based hydrological model parameter calibration method, which comprises data collection and further comprises the following steps of: generating a parameter group total sample; calculating Nash efficiency coefficients NSE corresponding to the sample points; performing normalization processing on the Nash efficiency coefficient NSE, and recording the Nash efficiency coefficient NSE as a new Nash efficiency coefficient NNSE; calibration of multi-target hydrological model parameters is carried out; and outputting optimal parameters of the hydrological model. According to the hydrological model parameter calibration method based on the normalized Nash efficiency coefficient provided by the invention, the Nash efficiency coefficient NSE is normalized through the hyperbolic tangent function and mathematical transformation, and the normalized Nash efficiency coefficient NNSE is used as a hydrological model parameter calibration target function, so that the defects of the existing parameter calibration target function are effectively overcome.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of watershed hydrological models, in particular to a hydrological model parameter calibration method based on normalized Nash efficiency coefficients. Background Art

[0002] Hydrological models are generalized models of real and complex hydrological phenomena in real environments. They are important tools for carrying out hydrological forecasting, flood control decision-making and other work. At present, the physical meaning of some parameters in hydrological models is relatively weak. Therefore, the values ​​of these parameters cannot be determined by measurement. They can only be estimated through rainfall-runoff data. The process of estimating parameters is called parameter calibration. In the process of using hydrological models, the results of model parameter calibration will directly affect the accuracy of model simulation results. Since the 1930s, with the development of hydrological model technology, many scholars have studied parameter calibration methods. This problem has always been a hot and difficult issue of concern to the water conservancy engineering community and academia.

[0003] After years of development, the hydrological model parameter calibration method has become relatively mature. At present, the hydrological model parameter calibration method is mainly divided into two categories: manual trial and error method and automatic optimization method. Among them, the automatic optimization method sets rules and allows the computer to automatically optimize the parameters within the value range. This method not only saves time and effort, but also avoids the deviation caused by human subjective factors. Therefore, the automatic optimization method has become the mainstream method for parameter calibration.

[0004] In August 2023, the magazine "Jiangxi Water Conservancy Science and Technology" published an article entitled "Simulation Analysis of Multiple Hydrological Forecasting Models in Wuyuan Area" by Zhao Lanlan, Zhao Bing, Hong Bo and Shi Peng. The article takes Wuyuan area in Jiangxi as the research area, selects Xin'anjiang model, BP neural network model, flood forecasting model based on soil hydrodynamics and GR3 model, and simulates 18 floods since 2017 respectively, and conducts a comprehensive comparative analysis from the aspects of certainty coefficient and flood pass rate. The results show that the four models have good effects in Wuyuan area, and the average certainty coefficient exceeds 0.8, among which the BP neural network model has the highest pass rate of 0.977; the flood pass rate exceeds 60%, and the flood process simulation effect of Xin'anjiang model and GR3 model is equivalent. In actual forecasting, a combination of multiple models can be used for comprehensive analysis to improve the accuracy of flood forecasting. The disadvantage of this article is that it only uses the certainty coefficient as a statistical indicator of the degree of matching between the simulation results of the hydrological model and the actual observed flow process line, which makes it difficult to comprehensively evaluate the performance of the hydrological model.

[0005] On August 30, 2023, the first online paper of "Hydrology" published by Li Yue, Liu Ji, Yang Shaokang, Zhao Xinrui and Shu Zhangkang was titled "Analysis of the Sensitivity and Uncertainty Impact of Different Objective Functions on the Parameters of the Xin'anjiang Model". The article proposed that the choice of objective function affects the parameter calibration results of the basin hydrological model. In order to further explore the influence of different objective functions on the sensitivity and uncertainty of hydrological model parameters and runoff simulation results, this paper takes the daily runoff simulation of Chitan Basin in Jinxi, Fujian Province as an example, and uses the SCEM-UA optimization algorithm to calibrate the Xin'anjiang model with three water sources. The Nash efficiency coefficient f1, root mean square error f2 and Kling-Gupta efficiency coefficient f3 are selected as the objective functions for parameter calibration. At the same time, a total of 6 calibration targets, namely F1, F2 and F3, which are combined multi-objective functions that prefer f1, f2 and f3 respectively, are constructed. The differences in runoff simulation accuracy under different calibration targets are compared, and the Sobol global sensitivity analysis method is used to quantitatively compare and analyze the sensitivity of Xin'anjiang model parameters under different objective functions. Finally, the uncertainty of runoff simulation under wet and dry seasons and different flow levels is analyzed based on the GLUE method. The results show that: (1) the identification results of the main sensitive parameters of the Xin'anjiang model are consistent under the six objective functions; (2) F1, F2, and F3 show higher runoff simulation accuracy than the single objective function; (3) In different hydrological periods and different flow levels, f1 shows a higher coverage rate and a smaller uncertainty interval than other objective functions. Under different objective functions, the uncertainty of the parameters in the dry season is the smallest, and the uncertainty of runoff simulation increases with the increase of flow level. The disadvantage of this article is that the three objective functions have different value ranges and optimization directions. Forcibly combining them into a new statistical indicator by addition will have an adverse effect on the optimization work. Summary of the invention

[0006] In order to solve the above-mentioned technical problems, the present invention proposes a hydrological model parameter calibration method based on normalized Nash efficiency coefficient, which normalizes the Nash efficiency coefficient (NSE) through hyperbolic tangent function and mathematical transformation, and uses the normalized Nash efficiency coefficient (NNSE) as the objective function for hydrological model parameter calibration, which effectively makes up for the shortcomings of the existing parameter calibration objective function.

[0007] The object of the present invention is to provide a method for calibrating hydrological model parameters based on normalized Nash efficiency coefficient, comprising data collection and the following steps:

[0008] Step 1: Generate total sample of parameter groups;

[0009] Step 2: Calculate the Nash efficiency coefficient NSE corresponding to the sample point;

[0010] Step 3: normalize the Nash efficiency coefficient NSE and record it as a new Nash efficiency coefficient NNSE;

[0011] Step 4: Calibrate the parameters of the multi-objective hydrological model;

[0012] Step 5: Output the optimal parameters of the hydrological model.

[0013] Preferably, the data collection includes collecting rainfall and evaporation sequence data applicable to the watershed for which the hydrological model is to be calibrated, and measured runoff sequence data or ideal runoff sequence data at the corresponding time.

[0014] In any of the above solutions, preferably, step 1 includes generating random values ​​within the range of each parameter value to form a parameter group sample point, and repeating this step to generate a parameter sample group.

[0015] In any of the above solutions, preferably, the step 2 includes bringing the parameter sample group into the hydrological model, obtaining the simulated flow process and calculating the Nash efficiency coefficient NSE.

[0016] In any of the above solutions, preferably, step 3 includes normalizing the Nash efficiency coefficient NSE using a hyperbolic tangent function and mathematical transformation.

[0017] In any of the above solutions, preferably, the calculation formula of the Nash efficiency coefficient NSE is:

[0018]

[0019] Among them, y i is the measured flow rate at the ith moment, is the predicted flow of the model at the i-th moment, is the mean of the measured flow, i is the time serial number of the flow data, and n is the length of the flow data sequence.

[0020] In any of the above solutions, preferably, the normalized new Nash efficiency coefficient NNSE formula is:

[0021] NNSE = 1-tanh(1-NSE).

[0022] In any of the above schemes, preferably, step 4 includes taking the new Nash efficiency coefficient NNSE as an objective function in the multi-objective optimization algorithm to carry out multi-objective hydrological model parameter calibration.

[0023] The present invention proposes a hydrological model parameter calibration method based on the normalized Nash efficiency coefficient. When performing multi-objective optimization, this method can achieve dimensionless and magnitude-similar processing between objective functions of different dimensions, and limit their value range to a bounded interval. This processing helps the multi-objective optimization algorithm find a higher quality solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The present invention is a flowchart of a preferred embodiment of a method for calibrating hydrological model parameters based on normalized Nash efficiency coefficients.

[0025] Figure 2 The flowchart is another preferred embodiment of the hydrological model parameter calibration method based on the normalized Nash efficiency coefficient according to the present invention.

[0026] Figure 3 It is a comparison diagram of flow simulation results according to an embodiment of the hydrological model parameter calibration method based on the normalized Nash efficiency coefficient according to the present invention. DETAILED DESCRIPTION

[0027] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0028] Embodiment 1

[0029] like Figure 1 As shown, step 100 is executed to collect data, which includes collecting rainfall and evaporation sequence data suitable for the watershed where the hydrological model is to be calibrated, as well as measured runoff sequence data or ideal runoff sequence data at the corresponding time.

[0030] Execute step 110 to generate a total sample of the parameter group, including generating random values ​​within the value range of each parameter to form a parameter group sample point, and repeat this step to generate a parameter sample group.

[0031] Execute step 120 to calculate the Nash efficiency coefficient NSE corresponding to the sample point, including bringing the parameter sample group into the hydrological model, obtaining the simulated flow process and calculating the Nash efficiency coefficient NSE.

[0032] The calculation formula of the Nash efficiency coefficient NSE is:

[0033]

[0034] Among them, y i is the measured flow rate at the ith moment, is the predicted flow of the model at the i-th moment, is the mean of the measured flow, i is the time serial number of the flow data, and n is the length of the flow data sequence.

[0035] Execute step 130 to normalize the Nash efficiency coefficient NSE and record it as a new Nash efficiency coefficient NNSE, including using a hyperbolic tangent function and mathematical transformation to normalize the Nash efficiency coefficient NSE. The formula of the normalized new Nash efficiency coefficient NNSE is:

[0036] NNSE = 1-tanh(1-NSE).

[0037] Execute step 140 to calibrate the parameters of the multi-objective hydrological model;

[0038] Execute step 150 to output the optimal parameters of the hydrological model.

[0039] Embodiment 2

[0040] The object of the present invention is to provide an improved objective function calculation method, namely the normalized Nash efficiency coefficient, NNSE. After normalization, the value range of the objective function is transformed from an infinite interval to a bounded interval. When performing multi-objective optimization, this method can achieve dimensionless and similar magnitude processing between objective functions of different dimensions, and limit their value ranges to a bounded interval. This processing helps the multi-objective optimization algorithm to find a higher quality solution.

[0041] like Figure 2 As shown, the present invention is achieved through the following technical solutions:

[0042] Step 1: Data collection: Collect rainfall and evaporation series data applicable to the watershed where the hydrological model is to be calibrated, as well as measured runoff series data or ideal runoff series data at the corresponding time.

[0043] Step 2: Generate the total sample of the parameter group: Generate random values ​​within the value range of each parameter to form a parameter group sample point, and repeat this step to generate a parameter sample group.

[0044] Step 3: Calculation of Nash efficiency coefficient (NSE) corresponding to the sample point: Bring the parameter sample group into the hydrological model, obtain the simulated flow process and calculate the Nash efficiency coefficient (NSE);

[0045] The calculation method of Nash efficiency coefficient (NSE) is shown in formula (1):

[0046]

[0047] Where: y i is the measured flow rate; is the predicted flow of the model; is the mean of the measured flow rate; n is the length of the data sequence. Obviously, the value range of the objective function is (-∞,1].

[0048] Step 4: Normalize the NSE, and the normalized NSE is recorded as NNSE.

[0049] The NSE normalization method uses the hyperbolic tangent function (tanh) and mathematical transformation to process the NSE. Under the condition that the domain is (-∞,1], the value range of the objective function (NNSE) after mathematical transformation is (0,1], where 1 means that the simulated flow rate is completely consistent with the measured flow rate, and 0 means that the simulation result is extremely poor. The normalization method is shown in formula (2):

[0050]

[0051] Step 5: Multi-objective hydrological model parameter calibration: Use NNSE as one of the objective functions in the multi-objective optimization algorithm to carry out multi-objective hydrological model parameter calibration.

[0052] Step 6: Output the optimal parameters of the hydrological model.

[0053] The present invention proposes a Nash efficiency coefficient normalization method, which can be used for multi-objective optimization of hydrological model parameters. The Nash efficiency coefficient (NSE) is normalized by hyperbolic tangent function and mathematical transformation, and the normalized Nash efficiency coefficient (NNSE) is used as the objective function for hydrological model parameter calibration, which effectively makes up for the shortcomings of the existing parameter calibration objective function.

[0054] Embodiment 3

[0055] like Figure 3 The comparison results of the flow simulation results based on the multi-objective optimization algorithm (the algorithm used in this example is NSGA-II) are shown. The rainfall-runoff data of the study area, the two objective functions and the Xin'anjiang hydrological model are combined to carry out parameter calibration. 2 ) and the Nash efficiency coefficient as the objective function of the flow simulation result is recorded as QTS1 (the orange-red line in the figure), and the determination coefficient (R 2 ) and the normalized Nash efficiency coefficient as the objective function is recorded as QTS2 (the blue line in the figure). Figure 2 The calculation results show that compared with QTS1, the simulation results of QTS2 are generally in better agreement with the observed values ​​(QTO, represented by black dots in the figure). The peak value of QTO is 426m 3 / s, the peak value of QTS1 is 347.23m 3 / s, the peak value of QTS2 is 412.47m 3 / s, which shows that the parameter calibration result of this method is better and this method can improve the accuracy of the simulation results.

[0056] In order to better understand the present invention, the above is described in detail in conjunction with the specific embodiments of the present invention, but it is not intended to limit the present invention. Any simple modifications made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention. Each embodiment in this specification focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referenced to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

Claims

1. A method for calibrating hydrological model parameters based on normalized Nash efficiency coefficients, comprising data collection, characterized in that: The following steps are also included: Step 1: Generate total sample of parameter groups; Step 2: Calculate the Nash efficiency coefficient NSE corresponding to the sample point. The calculation formula of the Nash efficiency coefficient NSE is: Among them, y i is the measured flow rate at the ith moment, is the predicted flow of the model at the i-th moment, is the mean of the measured flow, i is the time number of the flow data, and n is the length of the flow data sequence; Step 3: normalize the Nash efficiency coefficient NSE and record it as a new Nash efficiency coefficient NNSE; Step 4: Calibrate the parameters of the multi-objective hydrological model; Step 5: Output the optimal parameters of the hydrological model.

2. The method for calibrating hydrological model parameters based on normalized Nash efficiency coefficient according to claim 1, characterized in that: The data collection includes collecting rainfall and evaporation sequence data applicable to the watershed where the hydrological model is to be calibrated, as well as measured runoff sequence data or ideal runoff sequence data at the corresponding time.

3. The method for calibrating hydrological model parameters based on normalized Nash efficiency coefficient according to claim 2, characterized in that: The step 1 includes generating random values ​​within the range of each parameter value to form a parameter group sample point, and repeating this step to generate a parameter sample group.

4. The method for calibrating hydrological model parameters based on normalized Nash efficiency coefficient according to claim 3, characterized in that: The step 2 includes bringing the parameter sample group into the hydrological model, obtaining the simulated flow process and calculating the Nash efficiency coefficient NSE.

5. The method for calibrating hydrological model parameters based on normalized Nash efficiency coefficient according to claim 4, characterized in that: The step 3 includes normalizing the Nash efficiency coefficient NSE by using a hyperbolic tangent function and mathematical transformation.

6. The method for calibrating hydrological model parameters based on normalized Nash efficiency coefficient according to claim 5, characterized in that: The normalized new Nash efficiency coefficient NNSE formula is: NNSE = 1-tanh(1-NSE).

7. The method for calibrating hydrological model parameters based on normalized Nash efficiency coefficient according to claim 6, characterized in that: The step 4 includes taking the new Nash efficiency coefficient NNSE as an objective function in the multi-objective optimization algorithm and carrying out multi-objective hydrological model parameter calibration.

Citation Information

Patent Citations

  • Multi-basin parallel calibration method for hydrological model in geographic space

    CN111259522A

  • Simulation method for watershed daily scale water balance

    CN116561988A