Hydrological model parameter time-varying optimization method and system

Abstract

The invention provides a hydrological model parameter time-varying optimization method and system, and relates to the technical field of hydrological models and hydrological prediction.The method comprises the steps that hydrological meteorological data and runoff observation data are divided into time windows according to a sliding window, parameter optimization calibration is conducted through an optimization algorithm, and an optimal parameter set is obtained; splicing according to a time sequence, and generating a continuous curve of parameters changing along with time; performing score comparison on the continuous curves, and determining the window with the highest score as a steady-state window; the optimal parameters of the steady-state window, the climate statistics of the steady-state window and the underlying surface features are stored in a parameter database; and finally, determining the optimal parameter of the target time period based on the continuous curve or the parameter database. According to the method, dynamic adjustment of hydrological model parameters is realized through the sliding window, the optimization algorithm and the parameter database, the simulation precision of the model is improved, the stability of parameter optimization is enhanced, and the risk of local optimum is reduced.

Description

Technical Field

[0001] This invention relates to the field of hydrological modeling and hydrological prediction technology, and in particular to a method and system for time-varying optimization of hydrological model parameters. Background Technology

[0002] Hydrological models are crucial tools for simulating watershed hydrological processes and managing water resources, widely used in runoff forecasting, water resource allocation, flood warning, and hydraulic engineering design. Traditional hydrological models typically assume fixed parameters during parameter calibration, neglecting the dynamic characteristics of watershed hydrological processes that vary seasonally or sub-seasonally. In complex watershed conditions, such as high-altitude mountainous areas or watersheds with drastic precipitation variations, runoff generation and confluence processes are influenced by multiple factors including precipitation, evapotranspiration, snow cover, and permafrost changes, exhibiting significant temporal variability and nonlinear characteristics.

[0003] Existing methods for calibrating hydrological model parameters have certain limitations: Firstly, most methods perform only one-time parameter optimization during the observation period, failing to reflect the changing patterns of parameters over time, thus leading to decreased simulation accuracy under different seasons or extreme events. Secondly, there is a lack of unified scientific basis for selecting the sliding window length, resulting in significant differences in optimization results across different time scales, making it difficult to determine a steady-state window to ensure the reliability of parameter optimization. Furthermore, existing methods lack sufficient storage and management of historical calibration results, lacking effective parameter databases and retrieval mechanisms, making it difficult to reuse historical data and provide rapid references for hydrological forecasting or resimulation. Simultaneously, traditional optimization methods often rely on single algorithms, easily getting trapped in local optima and having limited computational efficiency, failing to meet the needs of large-scale or high-resolution hydrological simulations. Therefore, designing a time-varying optimization method and system for hydrological model parameters is essential. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for time-varying optimization of hydrological model parameters, which enables dynamic adjustment, effective management and rapid retrieval of hydrological model parameters through sliding windows, optimization algorithms and parameter databases, thereby improving the simulation accuracy of the model at different time scales, enhancing the stability and global adaptability of parameter optimization, and reducing the risk of local optima.

[0005] To achieve the above objectives, the present invention provides the following solution: A time-varying optimization method for hydrological model parameters includes the following steps: The hydrological and meteorological data and runoff observation data are divided into sliding segments according to a preset sliding window to obtain the time window; The optimal set of parameters is obtained by optimizing and calibrating the time window using an optimization algorithm; the parameters include: runoff generation coefficient, water storage coefficient, and evapotranspiration coefficient. The optimal parameters are concatenated in chronological order, and a continuous curve showing the parameter changes over time is generated based on the concatenation result. The continuous curves of different time windows are scored and compared, and the time window with the highest score is determined as the steady-state window; The optimal parameters of the steady-state window, the climate statistics of the steady-state window, and the underlying surface characteristics are stored in the parameter database. During the forecasting or resimulation phase, the optimal parameters for the target time period are determined based on continuous curves or parameter databases.

[0006] Optionally, the length of the sliding window is not less than the number of parameters; hydrometeorological data include: precipitation data, temperature data, land use type, and vegetation index.

[0007] Optionally, the optimization algorithm includes: SCE-UA algorithm, genetic algorithm, particle swarm optimization algorithm, Bayesian optimization algorithm and gradient method.

[0008] Optionally, the objective function of the optimization algorithm includes: the mean squared error function and the NSE function; the expression for the mean squared error function is: ;in, To observe runoff, For The simulated runoff obtained by running the hydrological model with parameters is t, where t is the interval. For the time window; the expression for the NSE function is: ;in, for The average of observations within the range.

[0009] Optionally, methods for generating continuous curves include: overlapping window weighted average method, moving average method, linear interpolation method, and spline interpolation method; the expression for the overlapping window weighted average method is: ;in, Let be the smoothing parameter value at time t. The optimal parameter values ​​obtained during calibration in the k-th time window are... These are the weighting coefficients; the expression for the moving average method is: ;in, Let be the smoothing parameter value after moving average at time t. Let be the parameter value at time t+j. The width is half the width of the sliding window.

[0010] Optionally, during the forecasting or resimulation phase, the optimal parameters for the target time period are determined based on continuous curves or a parameter database. Specific methods include: directly determining the optimal parameters from the continuous curves based on the target time period.

[0011] Optionally, the optimal parameters for the target period can be determined based on continuous curves or parameter databases. Specific methods also include: using a trained mapping function to predict the parameters of climate statistics and underlying surface characteristics to obtain the optimal parameters.

[0012] A time-varying optimization system for hydrological model parameters, comprising: The window segmentation module is used to segment hydrological and meteorological data and runoff observation data into time windows based on preset sliding windows; The parameter optimization module is used to optimize and calibrate the parameters of the time window through optimization algorithms to obtain the optimal set of parameters, including: runoff coefficient, water storage coefficient, and evapotranspiration coefficient. The curve generation module is used to stitch the optimal parameters in chronological order and generate a continuous curve of parameter changes over time based on the stitching results. The window filtering module is used to score and compare continuous curves at different time windows, and to determine the time window with the highest score as the steady-state window. The data storage module is used to store the optimal parameters of the steady-state window, the climate statistics of the steady-state window, and the underlying surface characteristics into the parameter database; The results output module is used to determine the optimal parameters for the target time period based on continuous curves or parameter databases during the forecasting or resimulation phase.

[0013] This invention discloses the following technical effects: The time-varying optimization method for hydrological model parameters provided by this invention includes: dividing hydrological and meteorological data and runoff observation data into time windows based on a preset sliding window; optimizing and calibrating the parameters within the time windows using an optimization algorithm to obtain a set of optimal parameters; stitching the optimal parameters together in chronological order and generating a continuous curve showing the parameters changing over time based on the stitching result; comparing and scoring the continuous curves of different time windows and determining the time window with the highest score as the steady-state window; storing the optimal parameters of the steady-state window, the climatological statistics of the steady-state window, and the underlying surface characteristics in a parameter database; and determining the optimal parameters for the target time period based on the continuous curve or the parameter database during the forecast or re-simulation stage. This method, through sliding windows, optimization algorithms, and parameter databases, achieves dynamic adjustment, effective management, and rapid retrieval of hydrological model parameters, improves the simulation accuracy of the model at different time scales, enhances the stability and global adaptability of parameter optimization, and reduces the risk of local optima. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of the time-varying optimization method for hydrological model parameters according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the hydrological model parameter time-varying optimization system according to an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figure 1 As shown in the figure, this embodiment of the invention provides a time-varying optimization method for hydrological model parameters, including the following steps: Step 100: Divide the hydrological and meteorological data and runoff observation data into time windows by sliding segments according to the preset sliding window; Step 200: Optimize and calibrate the time window parameters using an optimization algorithm to obtain the set of optimal parameters; Step 300: Concatenate the optimal parameters in chronological order and generate a continuous curve of parameter variation over time based on the concatenation result; Step 400: Compare and score the continuous curves of different time windows, and determine the time window with the highest score as the steady-state window; Step 500: Store the optimal parameters of the steady-state window, the climate statistics of the steady-state window, and the underlying surface characteristics into the parameter database; Step 600: In the forecast or resimulation phase, determine the optimal parameters for the target time period based on continuous curves or parameter databases.

[0019] Specifically, there is a corresponding relationship between the length of the sliding window and the number of parameters in the hydrological model. The window length must not be less than the number of parameters to ensure that the calibration process has sufficient data constraints. The hydrological and meteorological data include precipitation data, temperature data, land use type, and vegetation index. Both the hydrological and meteorological data and the runoff observation data are time-varying data.

[0020] In this embodiment, the parameter optimization process in step 200 is implemented using parallelization or hybrid optimization algorithms, including: Shuffled Complex Evolution (SCE-UA) algorithm, genetic algorithm, particle swarm optimization algorithm, Bayesian optimization algorithm, and gradient method. The target parameters for optimization include: runoff generation coefficient, water storage coefficient, and evapotranspiration coefficient, all of which are hydrological model parameters.

[0021] Specifically, the objective function used in the parameter calibration using the optimization algorithm within the window is any of the following forms or a combination thereof: 1) Mean Squared Error (SSE) function, the expression is: ; in, To observe runoff, For The simulated runoff obtained by running the hydrological model with parameters t is time. For time windows.

[0022] 2) The Nash–Sutcliffe efficiency (NSE) function is expressed as: ; in, for The average of observations within the range.

[0023] In this embodiment, step 300, which involves concatenating and smoothing the optimal parameters within the window into a continuous parameter timing sequence, employs one or more of the following methods: 1) Overlapping window weighted average method, the expression is: ; in, Let be the smoothing parameter value at time t. In the first k The optimal parameter values ​​obtained by time window calibration These are the weighting coefficients.

[0024] 2) Moving average method, the expression is: ; in, Let be the smoothing parameter value after moving average at time t. The original parameter sequence in t + j Parameter values ​​at time, The width of the sliding window is a positive integer. The number of sampling points for the moving average, including time points. arrive common One point.

[0025] 3) Linear interpolation or spline interpolation.

[0026] In this embodiment, step 400 selects time windows of different lengths and sequentially performs steps 100 to 300 to obtain a set of candidate window lengths. The comprehensive score of each candidate window is calculated and the one with the highest score is selected as the steady-state window.

[0027] Specifically, each record in the parameter database includes: window start and end times, optimal parameter vector, climate statistics within the window (including temperature and precipitation), underlying surface characteristics (including land use type and vegetation index), and window calibration evaluation indicators (selected in this embodiment). NSE or J ).

[0028] In this embodiment, the parameter acquisition method in step 600 can be any of the following methods: 1) Determine the optimal parameters directly from the continuous curve based on the target time period.

[0029] 2) Input climate statistics and underlying surface features into the trained mapping function to obtain prediction parameters, and use the obtained parameters to drive the hydrological model for simulation; wherein the mapping method of the mapping function supports nearest neighbor retrieval based on similarity measure (in this embodiment, Euclidean distance or cosine similarity) or index, which is used to retrieve the most similar window records in history.

[0030] like Figure 2 As shown, this embodiment of the invention also provides a time-varying optimization system for hydrological model parameters, comprising: The window segmentation module is used to segment hydrological and meteorological data and runoff observation data into time windows based on preset sliding windows; The parameter optimization module is used to optimize and calibrate the parameters of the time window through optimization algorithms to obtain the optimal set of parameters, including: runoff coefficient, water storage coefficient, and evapotranspiration coefficient. The curve generation module is used to stitch the optimal parameters in chronological order and generate a continuous curve of parameter changes over time based on the stitching results. The window filtering module is used to score and compare continuous curves at different time windows, and to determine the time window with the highest score as the steady-state window. The data storage module is used to store the optimal parameters of the steady-state window, the climate statistics of the steady-state window, and the underlying surface characteristics into the parameter database; The results output module is used to determine the optimal parameters for the target time period based on continuous curves or parameter databases during the forecasting or resimulation phase.

[0031] The beneficial effects of this invention are as follows: 1) By dynamically adjusting the parameters of the hydrological model, the seasonal and sub-seasonal changes of the watershed hydrological process can be captured more accurately, significantly improving the simulation accuracy of the model at different time scales. 2) By comprehensively evaluating the calibration results of different window lengths, the steady-state window can be determined more scientifically, thereby enhancing the stability and global adaptability of parameter optimization and reducing the risk of local optima; 3) By establishing a parameter database, the optimal parameters, climate statistics and underlying surface information for each time window are recorded, which realizes the effective management and rapid retrieval of historical parameter information. In the forecast stage, the parameter time series can be directly called or the parameters can be predicted through the mapping function, so that the hydrological model can quickly obtain the optimal parameters in re-simulation and short-term / sub-seasonal prediction, which improves the efficiency and reliability of hydrological simulation and forecast. 4) This invention is compatible with a variety of optimization algorithms and supports parallelization and hybrid optimization implementation. It can adapt to different watershed types and climatic conditions, has strong versatility and engineering application value, significantly improves the simulation accuracy and stability of hydrological models under complex watershed conditions, provides a scientific basis for water resource scheduling, flood control management and water conservancy engineering design, and has broad prospects for promotion and application.

[0032] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0033] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A time-varying optimization method for hydrological model parameters, characterized in that, Includes the following steps: The hydrological and meteorological data and runoff observation data are divided into sliding segments according to a preset sliding window to obtain the time window; The time window is optimized and calibrated using an optimization algorithm to obtain a set of optimal parameters, including the runoff generation coefficient, water storage coefficient, and evapotranspiration coefficient. The optimal parameters are concatenated in chronological order, and a continuous curve showing the parameter variation over time is generated based on the concatenation result. The continuous curves of different time windows are scored and compared, and the time window with the highest score is determined as the steady-state window; The optimal parameters of the steady-state window, the climate statistics of the steady-state window, and the underlying surface characteristics are stored in the parameter database. During the forecasting or resimulation phase, the optimal parameters for the target time period are determined based on the continuous curve or the parameter database.

2. The time-varying optimization method for hydrological model parameters according to claim 1, characterized in that, The length of the sliding window is not less than the number of parameters; the hydrological and meteorological data includes precipitation data, temperature data, land use type, and vegetation index.

3. The time-varying optimization method for hydrological model parameters according to claim 1, characterized in that, The optimization algorithms include: SCE-UA algorithm, genetic algorithm, particle swarm optimization algorithm, Bayesian optimization algorithm, and gradient method.

4. The time-varying optimization method for hydrological model parameters according to claim 1, characterized in that, The objective function of the optimization algorithm includes: a mean squared error function and an NSE function; the expression for the mean squared error function is: ;in, To observe runoff, For The simulated runoff obtained by running the hydrological model with parameters t is time. The time window is defined as follows; the expression for the NSE function is: ;in, for The average of observations within the range.

5. The time-varying optimization method for hydrological model parameters according to claim 1, characterized in that, The methods for generating the continuous curve include: overlapping window weighted average method, moving average method, linear interpolation method, and spline interpolation method; the expression for the overlapping window weighted average method is: ;in, Let be the smoothing parameter value at time t. The optimal parameter value obtained by calibration in the k-th time window is... The weighting coefficients are used; the expression for the moving average method is: ;in, Let be the smoothing parameter value after moving average at time t. Let be the parameter value at time t+j. The width is half the width of the sliding window.

6. The time-varying optimization method for hydrological model parameters according to claim 1, characterized in that, In the forecasting or resimulation phase, the optimal parameters for the target time period are determined based on the continuous curve or the parameter database. Specifically, the optimal parameters are determined directly from the continuous curve according to the target time period.

7. The time-varying optimization method for hydrological model parameters according to claim 6, characterized in that, The optimal parameters for the target time period are determined based on the continuous curve or the parameter database. The specific method further includes: using a trained mapping function to predict the parameters of the climate statistics and the underlying surface features to obtain the optimal parameters.

8. A time-varying optimization system for hydrological model parameters, characterized in that, include: The window segmentation module is used to segment hydrological and meteorological data and runoff observation data into time windows based on preset sliding windows; The parameter optimization module is used to optimize and calibrate the parameters of the time window through an optimization algorithm to obtain the optimal set of parameters; the parameters include: runoff coefficient, water storage coefficient and evapotranspiration coefficient; The curve generation module is used to stitch the optimal parameters in chronological order and generate a continuous curve of parameter changes over time based on the stitching result. The window filtering module is used to score and compare the continuous curves of different time windows, and to determine the time window with the highest score as the steady-state window; The data storage module is used to store the optimal parameters of the steady-state window, the climate statistics of the steady-state window, and the underlying surface characteristics into the parameter database; The results output module is used to determine the optimal parameters for the target time period based on the continuous curve or the parameter database during the forecast or resimulation phase.

Images