A digital twin basin dynamic calibration method based on marmoset optimization algorithm

Abstract

The application discloses a digital twin basin dynamic calibration method based on a marmoset optimization algorithm, and comprises the following steps: acquiring basic geographic information and hydro-meteorological monitoring data of a target basin, and preprocessing the basic geographic information and the hydro-meteorological monitoring data to form a basin operation state data set; constructing a digital twin basin model according to the basin operation state data; constructing a runoff prediction error function and a flood peak time error function based on simulated runoff results and actual observed runoff results output by the digital twin basin model; establishing a multi-objective optimization model according to the runoff prediction error function and the flood peak time error function; jointly optimizing and solving specified parameters and state variables of the digital twin basin model by using the marmoset optimization algorithm to obtain an optimal parameter combination; and updating the optimal parameter combination to the digital twin basin model to realize dynamic calibration. Through the construction of the digital twin basin model and the dynamic optimization of the model parameters and the state variables by combining the marmoset optimization algorithm, high-precision simulation and real-time prediction of the runoff process under a complex basin environment are realized.

Description

Technical Field

[0001] This application belongs to the field of smart water conservancy and hydrological forecasting technology, and in particular relates to a digital twin watershed dynamic calibration method based on the mongoose optimization algorithm. Background Technology

[0002] With the increasing impact of global climate change and human activities on natural hydrological systems, watershed hydrological processes are exhibiting more significant nonlinear characteristics and uncertainties. Particularly in mountainous or arid / semi-arid watersheds, factors such as uneven spatial and temporal distribution of rainfall, dramatic changes in evapotranspiration, and dynamic variations in soil moisture content all exert complex influences on runoff formation. Traditional hydrological models typically rely on historical data for parameter calibration, using fixed parameters for long-term predictions after model establishment. However, in actual watershed environments, factors such as soil moisture content, groundwater storage, vegetation cover, and rainfall intensity are constantly changing. Fixed-parameter models often fail to consistently reflect the true watershed conditions, leading to a gradual increase in runoff prediction errors.

[0003] In recent years, with the development of IoT, remote sensing, and big data technologies, the ability to acquire watershed monitoring data has been continuously improved, enabling the continuous acquisition of large amounts of real-time hydrological and meteorological data. Against this backdrop, digital twin technology has gradually been introduced into the field of smart water conservancy. Digital twin watersheds, by constructing a digital model corresponding to the real watershed in virtual space, achieve real-time mapping and dynamic simulation of watershed hydrological processes, thus providing important technical support for hydrological forecasting and water resource allocation.

[0004] However, in practical applications, digital twin watershed models still face challenges such as difficulty in updating model parameters in real time and a gradual deviation between the model and the actual watershed state. When model parameters cannot be dynamically adjusted based on the latest monitoring data, the differences between the model simulation results and the actual watershed operation will accumulate, thus affecting the accuracy of runoff prediction. Summary of the Invention

[0005] The purpose of this application is to overcome the shortcomings of the prior art by providing a dynamic calibration method for digital twin watersheds based on the mongoose optimization algorithm. The optimization algorithm enables dynamic updating of model parameters and state variables, thereby improving the accuracy of runoff prediction.

[0006] The objective of this application is achieved through the following technical solution: A method for dynamic calibration of digital twin watersheds based on the dwarf meerkat optimization algorithm, the method comprising: Acquire basic geographic information and hydrological and meteorological monitoring data of the target watershed, and preprocess the basic geographic information and the hydrological and meteorological monitoring data to form a watershed operation status dataset; A digital twin watershed model is constructed based on the watershed's operational status data; Based on the simulated runoff results output by the digital twin watershed model and the actual observed runoff results, a runoff prediction error function and a flood peak time error function are constructed. A multi-objective optimization model is established based on the runoff prediction error function and the flood peak time error function. The mongoose optimization algorithm is used to jointly optimize and solve the specified parameters and state variables of the digital twin watershed model to obtain the optimal parameter combination; The optimal parameter combination is updated into the digital twin watershed model to achieve dynamic calibration.

[0007] Furthermore, the basic geographic information includes digital elevation models, river network structures, land use types, and soil type data, and the hydrological and meteorological monitoring data includes rainfall, temperature, evapotranspiration, soil moisture, and hydrological station flow data; The preprocessing of the basic geographic information and the hydrological and meteorological monitoring data to form a basin operation status dataset specifically includes: The basic geographic information and the hydrological and meteorological monitoring data are cleaned, time scale unified, and spatially interpolated to form a basin operation status dataset.

[0008] Furthermore, the construction of the runoff prediction error function and the flood peak time error function specifically includes: Constructing the runoff prediction error function: ; in, This indicates that the model simulates the flow. This represents the measured flow rate, and N represents the length of the time series. This is the runoff prediction error function; Construct the peak flood time error function: ; in, This indicates the model's simulation of the flood peak arrival time. Indicates the actual time of arrival of the flood peak. This is the peak flood time error function.

[0009] Furthermore, the establishment of the multi-objective optimization model specifically includes: Taking into account both runoff prediction error and peak flow time error, the comprehensive objective function includes: ; in, and Here, represents the weighting coefficients, and F represents the overall optimization objective function.

[0010] Furthermore, the step of using the mongoose optimization algorithm to jointly optimize and solve the specified parameters and state variables of the digital twin watershed model to obtain the optimal parameter combination specifically includes: Update the location of individual pygmy mongooses: ; in, Let be the position of the i-th pygmy mongoose individual at the t-th iteration. This represents the optimal solution in the current iteration, where r is a random coefficient and t represents the number of iterations. By iteratively updating the location of individual pygmy mongooses, the value of the objective function F is gradually reduced, thereby obtaining the optimal combination of model parameters.

[0011] Furthermore, the method also includes: When new hydrological and meteorological monitoring data is received, the model parameters and state variables are updated to achieve continuous calibration.

[0012] The beneficial effects of this application are as follows: This application enables dynamic updating of digital twin watershed models, allowing model parameters to be automatically adjusted based on real-time monitoring data. It also achieves efficient parameter optimization through the mongoose optimization algorithm, thereby obtaining more accurate runoff prediction results in complex hydrological environments and providing reliable technical support for watershed water resources management, flood warning, and smart water conservancy decision-making. Attached Figure Description

[0013] Figure 1 This is a flowchart of the digital twin watershed dynamic calibration method based on the dwarf mongoose optimization algorithm in this application; Figure 2 This is a schematic diagram of the structure of a digital twin watershed model; Figure 3 This is a schematic diagram of the parameter optimization process of the dwarf mongoose optimization algorithm; Figure 4 This is a schematic diagram of the dynamic calibration and rolling prediction process of a digital twin watershed; Figure 5 This is a schematic diagram of the multi-source hydrological and meteorological data fusion structure. Detailed Implementation

[0014] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0015] Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] In practical applications, digital twin watershed models still face challenges such as difficulty in updating model parameters in real time and a gradual deviation between the model and the actual watershed state. When model parameters cannot be dynamically adjusted based on the latest monitoring data, the differences between the model simulation results and the actual watershed operation will continue to accumulate, thus affecting the accuracy of runoff prediction.

[0017] To address the aforementioned technical problems, the following embodiments of a digital twin watershed dynamic calibration method based on the mongoose optimization algorithm are proposed in this application.

[0018] This embodiment provides a digital twin watershed dynamic calibration method based on the dwarf meerkat optimization algorithm, referring to... Figure 1 ,like Figure 1 The diagram shown is a flowchart of the digital twin watershed dynamic calibration method based on the dwarf meerkat optimization algorithm in this embodiment. The method includes the following steps: S100: Multi-source hydrological and meteorological data acquisition. Acquire topographic data, land use data, hydrological monitoring data, and meteorological monitoring data of the target watershed, and preprocess rainfall, temperature, evapotranspiration, and flow data to form a watershed operation status dataset.

[0019] As one implementation method, in step S100 of this embodiment, multi-source hydrological and meteorological data are fused. By constructing a data fusion weight model, the reliability of different data sources is evaluated, thereby obtaining unified watershed hydrological input data.

[0020] S200: Construction of a digital twin watershed model. A digital twin watershed hydrological model is constructed based on watershed operational status data to simulate rainfall-runoff generation, slope runoff, and river runoff processes. (Refer to...) Figure 2 ,like Figure 2 The diagram shown is a schematic of the structure of a digital twin watershed model.

[0021] As one implementation method, the digital twin watershed hydrological model in this embodiment is constructed based on digital elevation model, river network structure, soil type and land use information, and is used to map the real watershed hydrological status in real time.

[0022] S300: Runoff Simulation and Error Calculation. Runoff prediction error functions and peak flow time error functions are constructed based on the simulated runoff results output by the digital twin watershed model and the actual observed runoff results.

[0023] The runoff prediction error function is: ; in: The model represents the simulated flow rate; Indicates the measured flow rate; N represents the length of the time series.

[0024] The peak flood time error function is: ; in: To simulate the peak flood time; This refers to the actual peak flood time.

[0025] S400: Establish a multi-objective optimization model. A multi-objective optimization model is established based on the runoff prediction error function and the peak flood time error function.

[0026] The comprehensive objective function of the multi-objective optimization model is: ; in: and These are the weighting coefficients.

[0027] S500: Meerkat Optimization Algorithm for Model Parameter Optimization. The meerkat optimization algorithm is used to jointly optimize and solve the key parameters and state variables of a digital twin watershed model to obtain the optimal parameter combination.

[0028] In this embodiment, the pygmy meerkat optimization algorithm simulates the foraging behavior of pygmy meerkats to perform global search and local optimization of infiltration coefficient, evapotranspiration correction coefficient, river confluence parameters, and soil water state variables in the digital twin watershed model. (Refer to...) Figure 3 ,like Figure 3 The diagram shows the parameter optimization process of the dwarf meerkat optimization algorithm.

[0029] The mongoose optimization algorithm continuously updates candidate solutions in the search space by simulating the foraging behavior of a mongoose population. Let the position of the i-th mongoose individual at the t-th iteration be: ; in: d represents the parameter dimension; This represents the parameter value of the i-th individual in the j-th dimension.

[0030] During the algorithm iteration process, the individual position is updated using the following update formula: ; in: This is the optimal solution in the current iteration; r is a random coefficient; t represents the number of iterations.

[0031] By iteratively updating the location of individual pygmy mongooses, the value of the objective function F is gradually reduced, thereby obtaining the optimal combination of model parameters.

[0032] S600: Dynamic Model Calibration and Runoff Prediction Output. The optimal parameter combination is updated into the digital twin watershed model to achieve dynamic model calibration, and runoff prediction results for future periods are output based on the calibrated model.

[0033] As one implementation method, in step S600 of this embodiment, the state variables in the digital twin watershed model are dynamically updated through an error feedback mechanism. These state variables include soil moisture content, groundwater storage, surface water storage, and snowmelt storage. (Refer to...) Figure 4 ,like Figure 4 The diagram shows the process of dynamic calibration and rolling prediction of a digital twin watershed.

[0034] When new hydrological monitoring data arrives, the optimization process described above can be repeated to update the model parameters and state variables, enabling continuous calibration and rolling prediction of the digital twin watershed model. This method ensures the model remains consistent with the actual watershed conditions, thereby improving the reliability and stability of runoff prediction.

[0035] Through the above technical solutions, this embodiment can realize the dynamic updating of the digital twin watershed model, enabling the model parameters to be automatically adjusted according to real-time monitoring data, and achieving efficient parameter optimization through the mongoose optimization algorithm, thereby obtaining more accurate runoff prediction results in complex hydrological environments, and providing reliable technical support for watershed water resources management, flood warning and smart water conservancy decision-making. Example 1

[0036] This embodiment proposes a digital twin watershed dynamic calibration and runoff forecasting method based on the dwarf meerkat optimization algorithm. This method targets watersheds with complex terrain, complex climate characteristics, and multi-source monitoring conditions. By collecting basic geographical data, hydrological and meteorological monitoring data, and watershed status observation data, a digital twin watershed model corresponding to the real watershed is constructed. During model operation, the dwarf meerkat optimization algorithm is introduced to dynamically calibrate key parameters and state variables of the model, thereby improving the accuracy, stability, and real-time response capability of runoff forecasting.

[0037] In practical implementation, the first step is to acquire the basic geographic information of the target watershed. This basic geographic information includes a digital elevation model (DEM), watershed boundary data, river network structure data, land use data, soil type data, vegetation cover data, and necessary surface engineering intervention information. The DEM is used to characterize the topographic relief features of the watershed and to extract watershed slope, aspect, confluence paths, river network distribution, and sub-watershed division results. Watershed boundary data is used to define the spatial extent of the target simulation area. River network structure data is used to determine the positional relationships of the main channel, tributaries, and confluence nodes at various levels. Land use data is used to distinguish different underlying surface types such as cultivated land, forest land, grassland, bare land, residential areas, and water bodies. Soil type data is used to reflect the permeability, water storage capacity, and water conduction capacity of soils in different regions. Vegetation cover data is used to reflect the evapotranspiration capacity, surface roughness, and raindrop interception effect in different regions. If reservoirs, dams, irrigation canals, or artificial water storage facilities exist in the watershed, they can also be incorporated into the digital twin watershed model as watershed boundary conditions or local control factors.

[0038] While acquiring basic geographic information, further hydrological and meteorological monitoring data within the target watershed are obtained. This data includes rainfall, temperature, relative humidity, wind speed, sunshine duration, evapotranspiration, soil moisture, groundwater level, hydrological station flow rate, water level processes, and necessary snow and ice cover data. For high-altitude or snowmelt-prone watersheds, data on snow depth, snow water equivalent, freezing depth, and diurnal temperature range can be added to improve the model's ability to characterize snowmelt processes. This data can be obtained from various sources, including surface meteorological stations, hydrological stations, automatic monitoring terminals, remote sensing platforms, and reanalysis data. Considering the differences in sampling frequency, spatial resolution, and data accuracy among different data sources, unified processing of all types of data is necessary in practical applications.

[0039] Specifically, preprocessing hydrological and meteorological monitoring data can include missing value imputation, outlier identification, noise filtering, time scale unification, spatial scale mapping, unit standardization, and quality assessment. Missing value imputation can be based on interpolation between adjacent time periods, regression between adjacent stations, or historical averages for the same period. Outlier identification can be performed using thresholding, box plots, or statistical bias methods. Noise filtering can improve data stability using moving averages, weighted smoothing, etc. Time scale unification refers to standardizing data from different sampling frequencies to hourly, daily, or other preset scales. Spatial scale mapping involves converting point observation data or raster data to sub-basin scales, basin unit scales, or the average scale of the entire basin. Unit standardization is used to avoid biases caused by data with different dimensions in subsequent analysis. The processed data forms a basin operational status dataset, serving as the input basis for a digital twin basin model.

[0040] In this embodiment, the digital twin watershed model can be understood as a watershed virtual mapping system that simultaneously possesses "static geographic mapping capability" and "dynamic state update capability." The static part of the model consists of the watershed topographic structure, river network structure, land use type, soil type, and initial parameter values, while the dynamic part consists of real-time rainfall input, evapotranspiration changes, soil moisture content changes, groundwater recharge status, and river flow status. The model as a whole includes a runoff generation module, a runoff confluence module, a groundwater response module, and a state feedback module. The runoff generation module is used to calculate the runoff generation process based on rainfall input and underlying surface conditions; the runoff confluence module is used to calculate slope confluence, gully confluence, and river propagation processes; the groundwater response module is used to characterize the groundwater recharge relationship to baseflow and the recession characteristics; and the state feedback module is used to compare the model simulation results with real-time observations and trigger subsequent optimization updates.

[0041] In the specific modeling process, the entire target watershed can be divided into multiple spatial computational units according to topography and river network structure. Each spatial computational unit has independent land use attributes, soil parameters, and state variables. Rainfall input first triggers the runoff generation process within each unit. The model determines, based on the current soil moisture content, evapotranspiration conditions, and infiltration capacity, how much rainfall is converted into surface runoff, how much rainfall enters the soil reservoir, and how much water recharges groundwater. Surface runoff further enters the river network system through slope confluence and propagates downstream within the river network. At each time step, the model outputs the simulated flow results for the corresponding node, cross section, or target station.

[0042] To evaluate the results of the digital twin watershed model, an error evaluation mechanism needs to be constructed.

[0043] In this embodiment, a flow error evaluation relationship is established by comparing the simulated flow rate output by the model with the measured flow rate at the hydrological station. Let the simulated flow rate at the t-th time step be... The actual flow rate was Then the runoff prediction error function can be expressed as: ;

[0044] Where N is the total number of time steps within the evaluation period. This represents the simulated flow rate at time step t. This represents the measured flow rate at time step t. This is the runoff prediction error function. This evaluation function is used to measure the overall deviation between the model-simulated flow rate and the measured flow rate.

[0045] In this embodiment, the parameters to be optimized in the model may include infiltration coefficient, soil water storage capacity parameter, evapotranspiration correction coefficient, slope runoff coefficient, river channel runoff propagation parameter, groundwater recession coefficient, river channel roughness parameter, and necessary snowmelt factor parameters. Furthermore, to further improve dynamic calibration capabilities, the current soil moisture content, surface water storage state, or baseflow state can be incorporated as auxiliary optimization parameters into the parameter vector. Thus, model optimization involves not only adjusting static parameters but also correcting the instantaneous state of the watershed, thereby enhancing the model's adaptability to sudden rainfall events and state transitions.

[0046] After obtaining the evaluation function and the set of parameters to be optimized, the mongoose optimization algorithm is introduced for dynamic optimization. The mongoose optimization algorithm is a swarm intelligence optimization algorithm that simulates the cooperative foraging behavior of mongoose groups. In this algorithm, each individual represents a set of candidate parameter combinations, and the entire population continuously searches in the parameter space, gradually retaining and strengthening parameter combinations with better fitness. In this embodiment, during population initialization, an initial set of parameter combinations can be generated based on empirical parameter ranges, historical calibration results, and physical feasibility constraints. Then, each set of parameters is input into the digital twin watershed model to calculate the corresponding evaluation function value, thereby measuring the current individual's performance.

[0047] In subsequent iterations, the mongoose optimization algorithm updates the positions of each candidate solution based on the current best individual and population distribution. This update process is not a simple random perturbation, but rather balances approaching the current best solution with exploring new regions in the search space. This improves search efficiency and helps avoid getting trapped in local optima. After each iteration, the system recalculates the error evaluation values ​​for all individuals and updates the current optimal parameter combination. If the preset iteration limit is reached, or the error decrease is less than a preset threshold in multiple consecutive iterations, the optimization process is considered converged, and the current optimal parameter combination is written back to the digital twin watershed model as the dynamic calibration result of this round.

[0048] After the optimal parameter combination is written back to the model, the system predicts runoff for several future time periods based on the updated digital twin watershed model. At this point, the digital twin watershed model is much closer to the actual watershed conditions than the initial model, thus the output runoff prediction results have higher reliability. For short-term prediction scenarios, the system can output the flow change process for the next few hours to one day; for medium-term prediction scenarios, it can combine forecasted rainfall input to output the runoff process curve for the next few days. The prediction results can be directly used for flood warnings, reservoir scheduling, irrigation water allocation, and watershed management decisions.

[0049] To achieve continuous dynamic updates, the digital twin watershed model in this embodiment is not "optimized once and used permanently." Instead, it continuously executes a closed-loop process of "observation-comparison-optimization-update-prediction" as new rounds of observation data arrive. In other words, within each update cycle, the system reacquires the latest monitoring data, re-evaluates model errors, and decides whether to activate the Jurassic optimization algorithm for recalibration based on the error level. In this way, the model can continuously track the changes in the real watershed under different weather conditions, seasonal stages, and underlying surface states, thereby significantly improving the model's adaptability to complex watershed systems.

[0050] Furthermore, in this embodiment, parameter boundary constraints and physical consistency constraints can also be set. For example, the infiltration coefficient must not exceed the reasonable range of the actual soil water conductivity, the evapotranspiration correction coefficient should be kept within the empirically acceptable range, and the soil moisture content must not exceed the upper limit of saturation or fall below the residual moisture content threshold. Through the above constraints, it can be ensured that the parameter solutions obtained by the dwarf mongoose optimization algorithm during the optimization process not only have better error performance mathematically, but are also physically reasonable and interpretable, thereby enhancing the credibility and feasibility of the present invention in engineering applications.

[0051] This embodiment establishes a dynamic calibration mechanism for digital twin watersheds to meet real-time operational needs. This mechanism organically combines multi-source monitoring data, a digital twin watershed model, and the Jurassic optimization algorithm, enabling the model to have adaptive update capabilities and real-time rolling forecasting capabilities. Compared with traditional methods that use fixed parameters for long-term forecasting, this embodiment can adjust the model promptly when the watershed state changes, reducing the accumulation of errors caused by parameter mismatch. Therefore, it has higher forecast accuracy and stronger application value under complex geographical conditions, strong nonlinear response conditions, and sudden rainfall scenarios. Example 2

[0052] In this embodiment, based on the digital twin watershed dynamic calibration and runoff forecasting method based on the mongoose optimization algorithm described in Embodiment 1, a multi-source hydrological and meteorological data fusion mechanism is introduced to address the impact of insufficient spatial distribution or large measurement errors of single observation data on watershed runoff prediction results. By fusing rainfall and hydrological information from different data sources, the spatiotemporal distribution characteristics of rainfall in the real watershed can be reflected to a greater extent, thereby improving the accuracy of the input data of the digital twin watershed model and further enhancing the reliability of the model simulation results. (Refer to...) Figure 5 ,like Figure 5 The diagram shown is a schematic diagram of the multi-source hydrological and meteorological data fusion structure.

[0053] In this embodiment, the rainfall input data sources include rainfall data from ground meteorological observation stations, remote sensing-retrieved rainfall data, and meteorological reanalysis rainfall data. Ground meteorological station rainfall data is typically collected by automatic rain gauges permanently deployed within or around the watershed. While offering high observational accuracy, the limited number of stations results in a relatively small spatial coverage area, potentially failing to fully reflect the spatial distribution of rainfall in complex terrain conditions. Remote sensing rainfall data, acquired through satellite or radar observations, offers good spatial continuity and covers a wide area. However, it may be affected by topographical obstruction, cloud structure, and algorithmic retrieval errors in mountainous or topographically complex regions. Meteorological reanalysis data, obtained by fusing meteorological models with observational data, offers good temporal continuity and can provide supplementary information in areas with missing observational data. By comprehensively utilizing these multiple data sources, the limitations of a single data source can be overcome.

[0054] Before data fusion, different data sources need to be standardized. Since different data sources differ in temporal resolution, spatial resolution, and data format, data standardization is the first step. For example, data with different temporal resolutions are converted to the same time interval, such as 1 hour or 3 hours, and discrete observation point data are converted to watershed unit-scale data using spatial interpolation methods. Simultaneously, data quality assessment and anomaly detection are necessary to remove obviously anomalous observations and appropriately supplement missing data. Through these processes, a unified data input format can be established, providing a foundation for subsequent data fusion.

[0055] After data preprocessing, a weighted fusion method is used to fuse multi-source rainfall data. By assigning different weights to different data sources, rainfall information can be comprehensively evaluated based on data quality and reliability, resulting in a fusion result that more closely approximates the actual rainfall distribution. The fused rainfall data can be represented as follows: ; in This represents rainfall data from the i-th data source. This represents the weighting coefficient of the corresponding data source, where P represents the fused rainfall data. The weighting coefficient can be set based on the reliability of the data source, historical error patterns, and data integrity. For example, if a data source shows relatively stable performance in historical comparative analysis, its weight can be appropriately increased; conversely, if a data source exhibits significant errors in certain regions or time periods, its weight can be decreased.

[0056] After obtaining the fused rainfall data, it is input into a digital twin watershed model for hydrological process simulation. Because the fused rainfall data better reflects the spatial distribution characteristics of rainfall within the watershed, the accuracy of the model simulation results is significantly improved. During model operation, the system calculates the runoff of each sub-watershed based on the current rainfall input and the initial state of the watershed, and gradually collects the water at the watershed outlet through slope runoff and river runoff processes, thus obtaining the runoff simulation results for each time step.

[0057] To ensure that the model output is consistent with the actual watershed conditions, this embodiment also introduces the mongoose optimization algorithm to dynamically update the model parameters. When there is a significant deviation between the model-simulated flow and the measured flow at the hydrological station, the system will automatically activate the optimization algorithm to readjust the model parameters. The mongoose optimization algorithm searches for the parameter combination that minimizes the prediction error in the parameter space by continuously updating the positions of individuals in the population. During the optimization process, each mongoose individual corresponds to a set of model parameter combinations. By running the digital twin watershed model and calculating the error function value, the quality of the parameter combination can be determined. As the algorithm iterates, the parameter combinations in the population gradually converge towards the optimal solution.

[0058] Furthermore, in this embodiment, soil moisture observation data can be incorporated to correct the model's internal state variables. Soil moisture is a crucial factor influencing rainfall-runoff processes. When there is a significant difference between the soil moisture content simulated by the model and actual observations, relevant state variables can be adjusted using optimization algorithms. This approach further enhances the model's ability to simulate watershed hydrological processes.

[0059] By introducing a multi-source data fusion mechanism, this embodiment can effectively improve the reliability of watershed rainfall input data, and dynamically adjust the model parameters through the mongoose optimization algorithm, so that the digital twin watershed model can more accurately reflect the real watershed state, thereby significantly improving the accuracy of runoff prediction. Example 3

[0060] In this embodiment, based on Embodiments 1 and 2, a real-time rolling prediction system for a digital twin watershed is further constructed to achieve continuous monitoring and dynamic prediction of watershed runoff processes. This system drives the digital twin watershed model through real-time monitoring data and continuously updates model parameters through an error feedback mechanism, enabling the model to continuously adapt to changes in the watershed environment, thereby improving the stability and real-time performance of runoff prediction.

[0061] During system operation, real-time hydrological and meteorological data are first acquired through a watershed monitoring network. This network includes meteorological observation stations, hydrological monitoring stations, automatic rain gauges, and soil moisture monitoring equipment. Meteorological observation stations collect meteorological information such as rainfall, temperature, wind speed, and evapotranspiration; hydrological monitoring stations acquire data on river level and flow changes; automatic rain gauges increase rainfall monitoring density; and soil moisture monitoring equipment monitors the water content of the watershed soil. Through these monitoring devices, the watershed's hydrological environment can be monitored in real time, and information on the watershed's operational status can be obtained promptly.

[0062] After acquiring real-time monitoring data, the system first preprocesses the data, including noise reduction, outlier detection, and time scale standardization. Since different monitoring devices may have different sampling frequencies, all monitoring data needs to be converted to the same time interval. Simultaneously, outlier detection methods can identify and remove obviously erroneous data records, thereby improving data quality.

[0063] After data preprocessing, the processed data is input into a digital twin watershed model for simulation calculations. The digital twin watershed model calculates the watershed runoff variation process based on current rainfall input and watershed state variables, and outputs runoff predictions for future periods. The model outputs predicted flow at each time step and compares it with real-time observed flow at hydrological stations to evaluate the accuracy of the model's predictions.

[0064] When the prediction error exceeds a preset threshold, the system automatically triggers a model parameter update mechanism. During the parameter update process, the mongoose optimization algorithm is introduced to re-optimize the model parameters. This algorithm simulates the foraging behavior of a mongoose group, continuously searching for better solutions in the parameter space to find the parameter combination that minimizes the prediction error. By continuously updating the model parameters, the digital twin watershed model can more closely approximate the real watershed conditions.

[0065] After updating the parameters, the system uses the updated model to continue predicting runoff for future periods. When new monitoring data arrives, the system repeats the above process, thus forming a continuous, cyclical prediction mechanism. This real-time rolling update mechanism ensures that the model maintains high prediction accuracy throughout its operation.

[0066] Furthermore, this embodiment can analyze historical prediction results and evaluate the model's operational status by comparing the error trends between the prediction results and actual observation data. When the system detects a continuous increase in prediction error, it can automatically increase the optimization frequency of model parameters, thereby enhancing the model's adaptability. Simultaneously, the system can also generate flood warning information based on the prediction results, providing decision support for basin flood control scheduling and water resource management.

[0067] Through the aforementioned real-time rolling forecasting system, the digital twin watershed model can continuously track changes in the actual watershed status and update model parameters based on the latest monitoring data, thereby significantly improving the accuracy of runoff forecasting. This system can be widely applied in areas such as watershed flood early warning, water resource allocation, and smart water management, and has promising application prospects in complex watershed environments.

[0068] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A digital twin watershed dynamic calibration method based on the dwarf mongoose optimization algorithm, characterized in that, The method includes: Acquire basic geographic information and hydrological and meteorological monitoring data of the target watershed, and preprocess the basic geographic information and the hydrological and meteorological monitoring data to form a watershed operation status dataset; A digital twin watershed model is constructed based on the watershed's operational status data; Based on the simulated runoff results output by the digital twin watershed model and the actual observed runoff results, a runoff prediction error function and a flood peak time error function are constructed. A multi-objective optimization model is established based on the runoff prediction error function and the flood peak time error function. The mongoose optimization algorithm is used to jointly optimize and solve the specified parameters and state variables of the digital twin watershed model to obtain the optimal parameter combination; The optimal parameter combination is updated into the digital twin watershed model to achieve dynamic calibration.

2. The digital twin watershed dynamic calibration method based on the mongoose optimization algorithm as described in claim 1, characterized in that, The basic geographic information includes digital elevation models, river network structure, land use types, and soil type data; the hydrological and meteorological monitoring data includes rainfall, temperature, evapotranspiration, soil moisture, and hydrological station flow data. The preprocessing of the basic geographic information and the hydrological and meteorological monitoring data to form a basin operation status dataset specifically includes: The basic geographic information and the hydrological and meteorological monitoring data are cleaned, time scale unified, and spatially interpolated to form a basin operation status dataset.

3. The digital twin watershed dynamic calibration method based on the mongoose optimization algorithm as described in claim 1, characterized in that, The construction of the runoff prediction error function and the flood peak time error function specifically includes: Construct the runoff prediction error function: ； in, This indicates that the model simulates the flow. This represents the measured flow rate, and N represents the length of the time series. This is the runoff prediction error function; Construct the peak flood time error function: ； in, This indicates the model's simulation of the flood peak arrival time. Indicates the actual time of arrival of the flood peak. This is the peak flood time error function.

4. The digital twin watershed dynamic calibration method based on the mongoose optimization algorithm as described in claim 3, characterized in that, The establishment of the multi-objective optimization model specifically includes: Taking into account both runoff prediction error and peak flow time error, the comprehensive objective function includes: ； in, and Here, represents the weighting coefficients, and F represents the overall optimization objective function.

5. The digital twin watershed dynamic calibration method based on the mongoose optimization algorithm as described in claim 4, characterized in that, The step of using the mongoose optimization algorithm to jointly optimize and solve the specified parameters and state variables of the digital twin watershed model to obtain the optimal parameter combination specifically includes: Update the location of individual pygmy mongooses: ； in, Let be the position of the i-th pygmy mongoose individual at the t-th iteration. This represents the optimal solution in the current iteration, where r is a random coefficient and t represents the number of iterations. By iteratively updating the location of individual pygmy mongooses, the value of the objective function F is gradually reduced, thereby obtaining the optimal combination of model parameters.

6. The digital twin watershed dynamic calibration method based on the mongoose optimization algorithm as described in claim 1, characterized in that, The method further includes: When new hydrological and meteorological monitoring data is received, the model parameters and state variables are updated to achieve continuous calibration.

Images