A Dynamic Estimation Method for Watershed Hydrological Model Parameters Based on Digital Twin Technology

Abstract

This invention discloses a dynamic estimation method for watershed hydrological model parameters based on digital twin technology. A digital twin of the watershed is constructed based on its physical characteristics data and historical hydrological and meteorological observation data. Sensors are deployed within the watershed to achieve real-time monitoring of multiple characteristic elements, enabling real-time tracking and information synchronization of the digital twin. Based on the historical and real-time synchronized digital twin, a combination of data assimilation methods and optimization algorithms is used to estimate the state variables and parameters in the simulation model, ultimately achieving dynamic estimation of the watershed hydrological model parameters. This method fully utilizes real-time monitoring data of the watershed, accurately reflecting the dynamic changes in watershed characteristics under changing environmental conditions. It achieves the goals of intelligent management, simulation, automatic optimization, real-time updates, and reasonable prediction, improving the accuracy of runoff simulation in watersheds under changing environments.

Description

Technical Field

[0001] This invention belongs to the field of watershed hydrological models, and in particular relates to a method for dynamic estimation of watershed hydrological model parameters based on digital twin technology. Background Technology

[0002] As an important tool for studying the natural laws of hydrology and solving practical hydrological problems, the accurate estimation of parameters of watershed hydrological models plays a crucial role in the simulation and forecasting of watershed runoff. Current watershed hydrological model parameter estimation methods generally use optimization algorithms for calibration based on historical hydrological data. The assumption is that the model parameters are constant within the watershed, meaning that the parameters calibrated based on historical hydrological data can be directly used for watershed runoff forecasting in the present and future. The main implementation steps of the current method are: (1) Selecting a complete historical hydrological data sequence, such as rainfall, potential evaporation, and runoff data; (2) Selecting the objective function for the optimization parameters, generally using the minimum sum of squared errors between simulated runoff and measured runoff as the objective function; (3) Using optimization algorithms to optimize the parameters to be estimated in the hydrological model, among which the more commonly used optimization algorithms include the SCE-UA algorithm and the genetic algorithm.

[0003] Therefore, existing methods for estimating hydrological model parameters have the following problems: (1) they only use historical data for parameter calibration, failing to fully utilize real-time monitoring data to further reduce the uncertainty of parameter estimation; (2) they cannot accurately reflect the dynamic changes in watershed characteristics under the influence of environmental changes; and (3) they cannot achieve the goals of intelligent management, perception analysis, simulation, automatic optimization, real-time updates, and reasonable prediction. Digital twin technology has the characteristics of real-time monitoring, intelligent judgment, and accurate prediction, tapping the hidden value of massive amounts of information and predicting future situations better, thus achieving true automation. Therefore, it is necessary to explore and invent a dynamic estimation method for watershed hydrological model parameters based on digital twin technology to solve the existing technical problems. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic estimation method for watershed hydrological model parameters based on digital twin technology, which can make full use of historical and real-time monitoring data, so that the parameters of the hydrological model can better reflect the changes in watershed characteristic conditions, thereby improving the accuracy of watershed hydrological model runoff simulation and forecasting under the influence of environmental changes.

[0005] The present invention adopts the following technical solution:

[0006] A method for dynamic estimation of parameters of a watershed hydrological model based on digital twin technology includes the following steps:

[0007] (1) Based on the physical characteristics data and hydro-meteorological data of the watershed, construct a digital twin of the watershed based on historical data. The physical characteristics data include digital elevation (DEM) data, watershed land use type (LULC) data, and watershed soil data. The hydro-meteorological data include precipitation, potential evaporation, and runoff.

[0008] (2) Real-time monitoring of multiple feature elements of the basin based on sensors deployed within the basin to obtain a real-time synchronized digital twin of the basin;

[0009] (3) Based on the real-time synchronized digital twin of the watershed, optimization algorithms and data assimilation methods were successively used to realize the dynamic estimation of watershed hydrological model parameters. The specific operation steps are as follows:

[0010] (3.1) The parameter sensitivity analysis method is used to conduct sensitivity analysis on the set of parameters of the hydrological model used in the target watershed, and the sensitive parameter set and the non-sensitive parameter set are selected.

[0011] (3.2) An optimization algorithm is used to calibrate the set of parameters of the hydrological model and determine the parameter values ​​of the set of insensitive parameters;

[0012] (3.3) The state transition equation and observation equation of the hydrological model sensitivity parameter set are constructed by using the data assimilation method, and the sensitivity parameter set is dynamically estimated based on the real-time tracked observation values ​​of multiple feature elements of the watershed.

[0013] Furthermore, the watershed multi-feature elements mentioned in step (2) include watershed soil moisture content, watershed evaporation, and watershed hydrological control section runoff.

[0014] Furthermore, in the real-time monitoring of multiple feature elements of the watershed described in step (2), the deployment of watershed soil moisture content and watershed evaporation observation sensors adopts a multi-point deployment method based on the spatial topographic features of the watershed, while the deployment of watershed hydrological control section runoff sensors adopts a single-point deployment method at the total outlet section of the watershed.

[0015] Furthermore, when constructing the state transition equation for the set of hydrological model sensitivity parameters in step (3.3), the initial values ​​of the sensitivity parameters are the optimal values ​​estimated by the optimization algorithm in step (3.2).

[0016] Furthermore, the data assimilation estimation method in step (3.3) is multi-source data assimilation that integrates three types of observation data: watershed soil moisture content, watershed evaporation, and watershed hydrological control section runoff.

[0017] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0018] (1) Existing technologies can only estimate parameters and predict runoff through historical measured hydrological and meteorological data. However, the method of this invention constructs a real-time digital twin through real-time data feedback, which is more in line with the actual situation of the watershed under changing environment. Compared with existing technologies, it can make full use of the real-time data of the watershed.

[0019] (2) Existing technologies can generally only consider the case where the basin climate conditions and underlying surface conditions do not change, and assume that the parameters of the model are constant. However, the method of this invention considers the characteristics of hydrological model parameters changing over time, and can more accurately reflect the changes in the characteristic conditions of the basin under changing environments.

[0020] (3) Existing technologies cannot achieve the goals of intelligent management, automatic optimization, and real-time updates, while the method of the present invention has the characteristics of real-time monitoring, intelligent judgment and accurate prediction, which can better simulate future situations and achieve true automation. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0022] This invention utilizes digital twin technology to construct a runoff simulation model centered on a watershed digital twin and a hydrological model, based on watershed hydrological and meteorological data and related parameters. Sensors are used to track and update the watershed digital twin. Then, based on the real-time synchronized watershed digital twin, data assimilation technology is used to dynamically estimate the parameters of the runoff simulation model. Figure 1 As shown, the technical solution of the present invention specifically includes the following steps:

[0023] Step 1: Based on the physical characteristics data of the watershed and hydrological and meteorological data, construct a digital twin of the watershed based on historical data;

[0024] A three-dimensional physical model is established based on historical meteorological, hydrological, land use, soil, digital elevation, and water conservancy engineering data of the watershed. At the same time, a logical model is established through graphical methods, and the relationships between various data elements are fed back into the physical model. A suitable watershed hydrological model is selected, and a runoff simulation model of the watershed is established using meteorological and hydrological data. Finally, the watershed entity is transformed into a three-dimensional twin, forming a watershed digital twin based on the physical model, logical model, and runoff simulation model.

[0025] Step 2: Based on the sensors deployed within the watershed, real-time monitoring of multiple characteristic elements of the watershed is carried out to obtain a real-time synchronized digital twin of the watershed.

[0026] Sensors are deployed throughout the target watershed to collect relevant parameters, as well as meteorological, hydrological, soil, topographic, and hydraulic engineering data. These sensors specifically include those for monitoring meteorological parameters (rainfall, temperature, wind speed, humidity, evaporation, etc.), hydrological parameters (water level, flow rate, soil moisture, etc.), and geographic parameters. The sensors should be distributed as evenly as possible to effectively capture the dynamic changes in watershed characteristics under varying environmental conditions. The data collected by the sensors is stored in a digital twin database, enabling dynamic real-time updates.

[0027] Step 3: Based on the real-time synchronized digital twin of the watershed, optimization algorithms and data assimilation methods are successively used to dynamically estimate the watershed hydrological model parameter θ. The specific operation process includes:

[0028] (1) For the selected set of hydrological model parameters Θ, based on historical hydrological and meteorological data of the watershed, sensitivity analysis methods (such as Extended Fourier Amplitude Sensitivity Analysis (EFAST method), Morris screening method, and Sobol method) are used to analyze the sensitivity of hydrological model parameters and screen out the set of sensitive parameters Θ. s Insensitive parameter set Θ ns .

[0029] (2) Use optimization algorithms (such as genetic algorithm, SCE-UA algorithm) to calibrate the hydrological model parameter set Θ and determine the set of insensitive parameters Θ. ns The parameter value.

[0030] Using optimization algorithms to estimate the set of insensitive parameters Θ ns In the process of assembling the optimal values, considering the consistency of all parameters in the hydrological model during runoff simulation, a parameter optimization objective function that minimizes the sum of squared errors between simulated and measured runoff values ​​is first constructed based on historical hydrological and meteorological data (including precipitation, potential evaporation, and runoff). This objective function is then used to calibrate the entire parameter set Θ. Subsequently, for the non-sensitive parameter set, the parameter values ​​remain unchanged using the current optimal values ​​throughout the subsequent process. For the sensitive parameter set Θ s The optimal values ​​of all parameters in this calibration This will serve as the initial value in the next step, as explained in step (3).

[0031] The superscript 'c' indicates the optimal value of the hydrological model parameters obtained from calibration; the superscripts 1 to n and 1 to m represent the parameter numbers; n represents the number of sensitive parameters and m represents the number of non-sensitive parameters.

[0032] (3) Construct a set of hydrological model sensitivity parameters Θ using data assimilation methods (such as ensemble Kalman filtering and particle filtering). s The state transition equations and observation equations are as follows:

[0033]

[0034] y t+1 =h(x t+1 ,θ t+1 )+ξ (2)

[0035] In the formula: θ∈Θ s For sensitivity parameters; θ t+1 θ t ... t+1 x t Let be the model state variables at times t+1 and t, respectively, and let η be the normal distribution error with a mean of 0; u t+1 The model input data at time t+1 typically includes rainfall P and potential evapotranspiration PET; y t+1 The simulated values ​​of the model observation variables at time t+1 include the actual evapotranspiration ET, soil moisture content W, and runoff Q; ξ is the normal distribution error with a mean of 0; f and h are also present.

[0036] Here, "watershed hydrological model" refers to the parameters. Specifically, when t = 1, i.e., at the initial time, the parameters take the values ​​from the previous time.

[0037] The optimal value obtained by calibration in one step

[0038] Once the observed values ​​of evapotranspiration, soil moisture content, and runoff at time t+1 are obtained, the process of dynamically estimating sensitive parameters using the observed data can be described by the following equation:

[0039]

[0040]

[0041] In the formula: Z represents the extended vector of the hydrological model's state variables and sensitivity parameters, i.e. For t+1

[0042] The k-th predicted set item in the filtered time step; Let be the k-th updated set item in the filter at time t+1; For time t+1, the observation error term is added. The calculated kth observation set term; K t+1The Kalman gain factor at time t+1 can be calculated using the following equation:

[0043] K t+1 =COV zy (COV yy +W t+1 ) -1 (5)

[0044]

[0045]

[0046]

[0047]

[0048] Where: COV zy COV represents the covariance between the expanded vector z and the observed vector y after adding the perturbation. yy This represents the covariance of the observed vector y after the perturbation; These represent the average values ​​of the perturbation-expanded vector and the observed vector, respectively. Let N and T represent the state and the Nth term in the observation vector, respectively; N represents the set number; and the superscript T denotes the transpose of the matrix. Specifically, the observation vector is a vector of observed variables consisting of watershed evapotranspiration, soil moisture content, and runoff, i.e. In the above formula, ET t+1 W represents the actual evaporation at time t+1. t+1 Q is the soil moisture content at time t+1. t+1 It is the runoff at time t+1.

[0049] The embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A method for dynamic estimation of parameters of a watershed hydrological model based on digital twin technology, characterized in that, Includes the following steps: (1) Based on the physical characteristics data and hydro-meteorological data of the watershed, construct a digital twin of the watershed based on historical data. The physical characteristics data include digital elevation (DEM) data, watershed land use type (LULC) data, and watershed soil data. The hydro-meteorological data include precipitation, potential evaporation, and runoff. A three-dimensional physical model is established based on meteorological, hydrological, land use, soil, digital elevation, and water conservancy engineering data from historical periods of the watershed. Simultaneously, a logical model is established graphically, and the relationships between various data elements are fed back into the physical model. A suitable watershed hydrological model is selected, and a runoff simulation model of the watershed is established using the watershed's meteorological and hydrological data. Finally, the watershed entity is transformed into a three-dimensional twin, forming a digital twin of the watershed based on the physical model, logical model, and runoff simulation model. (2) Real-time monitoring of multiple feature elements of the basin based on sensors deployed within the basin to obtain a real-time synchronized digital twin of the basin; (3) Based on the real-time synchronized digital twin of the watershed, optimization algorithms and data assimilation methods were successively used to realize the dynamic estimation of watershed hydrological model parameters. The specific operation steps are as follows: (3.1) The parameter sensitivity analysis method is used to conduct sensitivity analysis on the set of parameters of the hydrological model used in the target watershed, and the sensitive parameter set and the non-sensitive parameter set are selected. (3.2) An optimization algorithm is used to calibrate the set of parameters of the hydrological model and determine the parameter values ​​of the set of insensitive parameters. Specifically: First, based on historical hydrological and meteorological data, a parameter optimization objective function is constructed to minimize the sum of squared errors between simulated and measured runoff values, and this function is calibrated for all parameter sets Θ. Subsequently, for the non-sensitive parameter set, the parameter values ​​remain unchanged using the current optimal values ​​throughout the subsequent process. For the sensitive parameter set Θ s The optimal values ​​of all parameters in this calibration Then it will be used as the initial value in the next step; The superscript 'c' indicates the optimal value of the hydrological model parameters obtained from calibration; the superscripts 1 to n and 1 to m represent the parameter number; n represents the number of sensitive parameters and m represents the number of non-sensitive parameters. (3.3) The state transition equation and observation equation of the hydrological model sensitivity parameter set are constructed using the data assimilation method. The sensitivity parameter set is dynamically estimated based on the real-time tracked observation values ​​of multiple watershed features. Specifically: A set of sensitive parameters Θ for the hydrological model is constructed using data assimilation methods. s The state transition equations and observation equations are as follows: y t+1 =h(x t+1 ,i t+1 )+ξ (2) In the formula: θ∈Θ s For sensitive parameters; θ t+1 θ t ... t+1 x t Let be the model state variables at times t+1 and t, respectively, and let η be the normal distribution error with a mean of 0; u t+1 The model input data at time t+1 includes rainfall P and potential evapotranspiration PET; t+1 Let ξ be the simulated value of the model observation variable at time t+1; ξ be the normal distribution error with a mean of 0; f and h both represent the watershed hydrological model here; in particular, when t=1, that is, at the initial time, the parameter values ​​are the optimal values ​​obtained from the calibration in the previous step. After obtaining the observed values ​​of evapotranspiration, soil moisture content, and runoff at time t+1, the process of dynamically estimating sensitive parameters using the observed data is described by the following equation: In the formula: Z represents the extended vector of the hydrological model's state variables and sensitivity parameters, i.e. Let be the k-th predicted set item in the filter at time t+1; Let be the k-th updated set item in the filter at time t+1; For time t+1, the observation error term is added. The calculated kth observation set term; K t+1 The Kalman gain factor at time t+1 is calculated using the following equation: K t+1 =COV zy (COV) yy +W t+1 ) -1 (5) Where: COV zy COV represents the covariance between the expanded vector z and the observed vector y after adding the perturbation. yy This represents the covariance of the observed vector y after the perturbation; These represent the average values ​​of the perturbation-expanded vector and the observed vector, respectively. Let N and T represent the state and the Nth term in the observation vector, respectively; N represents the set number; the superscript T denotes the transpose of the matrix; specifically, the observation vector is a vector of observed variables consisting of watershed evapotranspiration, soil moisture content, and runoff, i.e. In the formula ET t+1 W represents the actual evaporation at time t+1. t+1 Q is the soil moisture content at time t+1. t+1 It is the runoff at time t+1.

2. The method for dynamic estimation of watershed hydrological model parameters based on digital twin technology according to claim 1, characterized in that, The watershed multi-feature elements mentioned in step (2) include watershed soil moisture content, watershed evaporation, and watershed hydrological control section runoff.

3. The method for dynamic estimation of watershed hydrological model parameters based on digital twin technology according to claim 2, characterized in that, In step (2), the real-time monitoring of multiple features of the watershed is carried out by deploying multiple monitoring sensors for watershed soil moisture content and watershed evaporation according to the spatial topographic features of the watershed, and by deploying runoff sensors at the watershed hydrological control section at a single point at the total outlet section of the watershed.

4. The method for dynamic estimation of watershed hydrological model parameters based on digital twin technology according to claim 1, characterized in that, When constructing the state transition equation for the set of hydrological model sensitivity parameters in step (3.3), the initial values ​​of the sensitivity parameters are the optimal values ​​estimated by the optimization algorithm in step (3.2).

5. The method for dynamic estimation of watershed hydrological model parameters based on digital twin technology according to claim 1, characterized in that, The data assimilation estimation method in step (3.3) is multi-source data assimilation that integrates three types of observation data: watershed soil moisture content, watershed evaporation, and watershed hydrological control section runoff.

Images