Hydrodynamic model-fused navigation and power hub dynamic flood digital twin deduction method and system

Abstract

This invention relates to the field of digital twin technology for water conservancy projects, and discloses a method and system for dynamic flood digital twin simulation of navigation and hydropower hubs that integrates hydrodynamic models. The method collects multi-source observation data to form a spatiotemporally coded raw dataset, constructs a set of hydrodynamic boundary parameters, builds a two-dimensional unsteady hydrodynamic calculation model based on the parameter set, and updates the model parameters online by assimilating the latest monitoring data. The hydrodynamic simulation state field is mapped onto a three-dimensional terrain mesh to generate a three-dimensional twin rendering scene. Based on this scene, inundation indicators and scheduling benefit indicators are calculated, a set of scheduling optimization schemes is generated, and execution control commands are issued. This realizes dynamic simulation of flood processes, multi-objective optimized scheduling, and closed-loop control of navigation and hydropower hubs, improving the accuracy and timeliness of flood control decisions.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology for water conservancy projects, and in particular to a method and system for dynamic flood digital twin simulation of navigation and hydropower hubs that integrates hydrodynamic models. Background Technology

[0002] Flood control and hydrological monitoring at navigation and hydropower hubs typically rely on hydrological models to calculate cross-sectional water level and flow data, and employ two-dimensional charts or three-dimensional displays based on static water levels to aid decision-making. These technologies primarily provide hydrological data support to help decision-makers understand hydrological dynamics. Their characteristics include data presented in discrete numerical form, and three-dimensional displays often depict water surfaces at fixed elevations, lacking a direct and intuitive reflection of the dynamic evolution of floods.

[0003] When engineers make dispatching decisions during sudden floods, they find that existing systems cannot accurately simulate the real-time impact of gate opening and closing operations on upstream and downstream flow patterns. For example, after opening floodgates, the downstream flow velocity changes displayed by the system differ significantly from actual observations, making it difficult for decision-makers to accurately judge the expansion trend of the inundation area based on system data. The root cause of this problem is that existing technology does not integrate hydrodynamic mechanisms into the real-time simulation process, relying only on static data and simple models. This makes it impossible to reproduce the nonlinear flow disturbances caused by gate operations. At the same time, the data update mechanism lags behind actual hydrological changes, causing a disconnect between dynamic process simulation and reality. This disconnect can lead to intensified dam erosion in downstream areas after dispatching orders are issued due to insufficient flow velocity prediction, increasing the risk of damage to the key structure and prolonging the inundation time in low-lying areas, thus creating additional difficulties for rescue and relief efforts. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, the present invention provides the following technical solution:

[0005] A digital twin modeling method for dynamic flood simulation of navigation and hydropower hubs integrating hydrodynamic models includes:

[0006] Collect multi-source observation data to form a spatiotemporally encoded raw dataset, and construct a hydrodynamic boundary parameter set based on the spatiotemporally encoded raw dataset;

[0007] A two-dimensional unsteady hydrodynamic calculation model is constructed based on the hydrodynamic boundary parameter set. The latest monitoring data is assimilated online to obtain the hydrodynamic simulation state field. The model parameters of the two-dimensional unsteady hydrodynamic calculation model are updated based on the hydrodynamic simulation state field.

[0008] The hydrodynamic simulation state field is mapped onto a 3D terrain mesh to generate a 3D twin rendering scene, and interactive visualization deduction and data rereading are driven based on the 3D twin rendering scene;

[0009] Based on the 3D twin rendering scene, the flooding index and scheduling efficiency index are calculated, and a set of scheduling optimization schemes is generated. Based on the set of scheduling optimization schemes, execution control instructions are generated for users to refer to and make decisions.

[0010] Furthermore, the specific method for collecting multi-source observation data to form a spatiotemporally coded raw dataset, and constructing a hydrodynamic boundary parameter set based on the spatiotemporally coded raw dataset is as follows:

[0011] Collect multi-source observation data, perform unified spatiotemporal benchmark transformation and quality labeling on the multi-source observation data, and form a spatiotemporally coded raw dataset;

[0012] A set of hydrodynamic boundary parameters was constructed based on the spatiotemporal encoded original dataset.

[0013] Furthermore, the process of forming the spatiotemporally encoded raw dataset includes the following steps:

[0014] Multi-source observation data are obtained by collecting upstream cross-sectional water level and flow sequences from hydrological telemetry stations, basin surface rainfall intensity sequences from radar rain gauges, gate opening sequences from gate control systems, generator output sequences and generator flow sequences from generator monitoring systems, and three-dimensional point cloud data of riverbed and bank slopes from survey vessel-borne multibeam echo sounders or UAV-borne lidar.

[0015] A unified spatiotemporal benchmark conversion was performed on the multi-source observation data. The timestamps of all time-series data in the multi-source observation data were converted to Coordinated Universal Time. The plane coordinates of all coordinate data in the multi-source observation data were uniformly converted to the 2000 National Geodetic Coordinate System. The elevation benchmarks of all elevation data in the multi-source observation data were uniformly converted to the 1985 National Elevation Benchmark.

[0016] Data quality checks and missing data completion processing are performed on the multi-source observation data after unified spatiotemporal benchmark conversion to generate spatiotemporally coded raw datasets.

[0017] Furthermore, the construction of the hydrodynamic boundary parameter set includes the following steps:

[0018] The water level and flow sequences corresponding to the upstream control sections are retrieved from the spatiotemporal encoded original dataset according to the spatial index. The upper boundary inflow process is constructed with the time index as the horizontal axis. At the same time, the water level sequences corresponding to the downstream reference sections are retrieved from the spatiotemporal encoded original dataset according to the spatial index. The lower boundary water level process is constructed with the time index as the horizontal axis.

[0019] From the spatiotemporal encoded raw dataset, retrieve the gate opening sequence corresponding to each floodgate according to the spatial index, retrieve the unit output sequence and unit flow sequence corresponding to each generator unit according to the spatial index, and construct the gate outflow relationship table and the unit outflow relationship table.

[0020] Three-dimensional point cloud data of riverbed and bank slope are retrieved from the spatiotemporal encoded original dataset according to the spatial index. Delaunay triangulation is performed on the three-dimensional point cloud data of riverbed and bank slope to generate an unstructured triangular terrain grid covering the reservoir area of ​​the navigation and hydropower hub, the hub building area and the downstream river section. At the same time, a roughness value is assigned to each triangular grid cell to form a roughness distribution map.

[0021] The upper boundary inflow process, lower boundary water level process, gate outflow relationship table, unit outflow relationship table, unstructured triangular terrain grid and roughness distribution map are assembled into a hydrodynamic boundary parameter set.

[0022] Furthermore, the specific method for constructing a two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic boundary parameter set, obtaining the hydrodynamic simulation state field by online assimilation of the latest monitoring data, and updating the model parameters of the two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic simulation state field is as follows:

[0023] A two-dimensional unsteady hydrodynamic calculation model is constructed based on the hydrodynamic boundary parameter set. The two-dimensional unsteady hydrodynamic calculation model is solved by time stepping to generate the hydrodynamic simulation state field.

[0024] The latest monitoring data is assimilated online, and the model parameters of the two-dimensional unsteady hydrodynamic calculation model are updated based on the hydrodynamic simulation state field.

[0025] Furthermore, the generation of the hydrodynamic simulation state field includes the following steps:

[0026] The unstructured triangular terrain grid and roughness distribution map in the hydrodynamic boundary parameter set are read. The unstructured triangular terrain grid is used as the basis for spatial discretization of the computational domain. The Manning roughness coefficient of each grid cell in the roughness distribution map is assigned to the corresponding computational cell to establish the spatial discretization framework of the two-dimensional unsteady hydrodynamic calculation model. The two-dimensional unsteady hydrodynamic calculation model is constructed based on the two-dimensional shallow water equation set.

[0027] Read the upper boundary inflow process, lower boundary water level process, gate outflow relationship table and unit outflow relationship table from the hydrodynamic boundary parameter set. Set the upper boundary inflow process as the flow boundary condition of the grid cell at the upstream inlet of the two-dimensional unsteady hydrodynamic calculation model, and set the lower boundary water level process as the water level boundary condition of the grid cell at the downstream outlet of the two-dimensional unsteady hydrodynamic calculation model. For the grid cells at the locations of each floodgate and generator unit of the navigation and power hub, set them as internal boundary conditions.

[0028] The two-dimensional unsteady hydrodynamic calculation model was spatially discretized using the finite volume method. The discretized equations were solved by time step using an explicit Runge-Kutta time integration scheme. The water depth, average lateral velocity, average longitudinal velocity, and water surface elevation of all grid cells obtained at each time step were written into the hydrodynamic simulation state field using time and spatial indices as double keys.

[0029] Furthermore, updating the model parameters of the two-dimensional unsteady hydrodynamic calculation model includes the following steps:

[0030] In the time-step solution of the two-dimensional unsteady hydrodynamic calculation model, the latest water level and flow rate observations are read from the spatiotemporally encoded original dataset at preset assimilation intervals. The latest water level and flow rate observations are compared with the water surface elevation and flow velocity values ​​at the corresponding spatial locations and time steps in the hydrodynamic simulation state field, and the observation residual vector is calculated.

[0031] Based on the observation residual vector, the model parameters of the two-dimensional unsteady hydrodynamic calculation model are updated using the ensemble Kalman filter method.

[0032] The updated model parameters are substituted into the two-dimensional unsteady hydrodynamic calculation model to continue the subsequent time step solution. The calculation results of the subsequent time steps are used to overwrite and update the water depth, lateral depth average velocity, longitudinal depth average velocity and water surface elevation values ​​at the corresponding time steps and spatial locations in the hydrodynamic simulation state field.

[0033] Furthermore, the specific method for mapping the hydrodynamic simulation state field onto a three-dimensional terrain mesh to generate a three-dimensional twin rendering scene, and driving interactive visualization deduction and data back-reading based on the three-dimensional twin rendering scene, is as follows:

[0034] The hydrodynamic simulation state field is mapped onto a 3D terrain mesh to generate a 3D twin rendering scene;

[0035] Interactive visualization simulation and data backreading driven by 3D twin rendering scene.

[0036] Furthermore, the specific method for generating execution control instructions based on the scheduling optimization scheme set for user reference in decision-making is as follows:

[0037] Calculate flooding and scheduling efficiency indicators based on 3D twin rendering scenes;

[0038] Generate a set of scheduling optimization schemes, and generate execution control instructions based on the set of scheduling optimization schemes for users to refer to and make decisions.

[0039] A digital twin simulation system for dynamic floods in navigation and hydropower hubs, integrating a hydrodynamic model, is used to implement the aforementioned method for dynamic flood simulation of navigation and hydropower hubs using an integrated hydrodynamic model. The system includes:

[0040] Multi-source data acquisition and hydrodynamic boundary parameter construction module: used to acquire multi-source observation data, form a spatiotemporally encoded raw dataset, and construct a hydrodynamic boundary parameter set based on the spatiotemporally encoded raw dataset;

[0041] Hydrodynamic Model Construction and Online Assimilation Module: This module is used to construct a two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic boundary parameter set, assimilate the latest monitoring data online to obtain the hydrodynamic simulation state field, and update the model parameters of the two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic simulation state field.

[0042] 3D Twin Rendering and Interactive Inference Module: Used to map the hydrodynamic simulation state field onto a 3D terrain mesh to generate a 3D twin rendering scene, and drive interactive visualization inference and data rereading based on the 3D twin rendering scene;

[0043] The scheduling optimization and instruction issuance module is used to calculate flooding indicators and scheduling efficiency indicators based on the 3D twin rendering scene, generate a set of scheduling optimization schemes, and generate execution control instructions based on the set of scheduling optimization schemes for users to refer to and make decisions.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0045] This invention effectively solves the problems of inconsistent formats and time-consuming manual conversion of multi-source heterogeneous data in traditional flood control scheduling of navigation and hydropower hubs by constructing a spatiotemporally coded raw dataset and a hydrodynamic boundary parameter set. The spatiotemporally coded raw dataset performs unified spatiotemporal benchmark conversion and quality labeling on multi-source observation data from hydrological telemetry stations, gate control systems, etc., reducing the time from data acquisition to usability for hydrodynamic models from days to minutes, avoiding the low efficiency of modeling preparation caused by manually writing format conversion scripts in traditional systems. The hydrodynamic boundary parameter set automatically aggregates boundary conditions, topographic information, and parameter information based on the spatiotemporally coded raw dataset, enabling adaptive updates of modeling boundary conditions as observation data is updated, solving the problem of simulation results being out of sync with actual operating conditions caused by the use of lagging data in traditional offline modeling.

[0046] The two-dimensional unsteady hydrodynamic calculation model solves the problem of traditional models having fixed parameters after offline calibration and being unable to adapt to changes in working conditions such as riverbed scouring and silting or gate wear by using Kalman filtering to assimilate the latest monitoring data online and dynamically correcting the roughness coefficient and flow coefficient. The model automatically activates a high-resolution calculation mode during flood season scenarios and switches to a low-resolution mode during normal operation, achieving a dynamic balance between calculation accuracy and efficiency and improving the simulation stability under complex working conditions.

[0047] The 3D twin rendering scene directly links the hydrodynamic simulation state field with the 3D rendering mesh based on an index mapping table, synchronizing water surface morphology, flow velocity texture, and calculation results frame by frame. This solves the problem of information isolation between traditional 3D visualization and hydrodynamic calculation results. It supports interactive cross-section selection and data readback, allowing decision-makers to directly obtain quantitative data from any location, upgrading 3D visualization from a display tool to an analysis tool and improving the intuitiveness of scheduling decisions.

[0048] The scheduling optimization scheme set generates Pareto optimal schemes that meet safety constraints through parallel simulation of multiple candidate schemes, quantitative calculation of inundation and benefit indicators, and non-dominated ranking. This solves the problems of traditional scheduling relying on experience and lacking quantitative evaluation. After the closed-loop issuance of execution control commands, the actual operating status automatically flows back to the data acquisition layer, triggering a full-link update and forming a closed-loop control of data acquisition-simulation-decision-execution. This compresses the scheduling response cycle to the minute level, improving the timeliness and accuracy of flood control scheduling, and meeting the technical requirements of general control or regulation systems for dynamic response and precise control. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0050] Figure 1 This is a flowchart of the dynamic flood digital twin simulation method for navigation and hydropower hubs that integrates hydrodynamic models in this invention;

[0051] Figure 2 This is a schematic diagram illustrating the construction of the spatiotemporal coding original dataset in an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of an unstructured triangular terrain grid in an embodiment of the present invention;

[0053] Figure 4 This is a schematic diagram illustrating the setting of internal boundary conditions of the gate in an embodiment of the present invention;

[0054] Figure 5 This is a schematic diagram of online assimilation using ensemble Kalman filtering in an embodiment of the present invention;

[0055] Figure 6 This is a schematic diagram of interactive section selection and data readback in an embodiment of the present invention;

[0056] Figure 7 This is a schematic diagram of the overlay of vector graphics layers for the protected target in an embodiment of the present invention;

[0057] Figure 8 This is a schematic diagram illustrating the calculation of the flooding index in an embodiment of the present invention;

[0058] Figure 9 This is a schematic diagram of multi-objective sorting of scheduling optimization scheme set in an embodiment of the present invention;

[0059] Figure 10 This is a schematic diagram of the closed-loop execution of control commands in an embodiment of the present invention;

[0060] Figure 11 This is a functional module diagram of the dynamic flood digital twin simulation system for navigation and hydropower hubs that integrates hydrodynamic models, as described in this invention. Detailed Implementation

[0061] 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.

[0062] Example 1:

[0063] Please see Figure 1 As shown, this embodiment provides a method for dynamic flood digital twin simulation of navigation and hydropower hubs by integrating hydrodynamic models, including:

[0064] S1: Collect multi-source observation data to form a spatiotemporally encoded raw dataset, and construct a hydrodynamic boundary parameter set based on the spatiotemporally encoded raw dataset; further, S1 includes:

[0065] S11: Collect multi-source observation data, perform unified spatiotemporal benchmark transformation and quality labeling on the multi-source observation data, and form a spatiotemporally coded raw dataset; further, forming the spatiotemporally coded raw dataset includes the following steps:

[0066] S111: Upstream cross-section water level and flow sequences are collected from hydrological telemetry stations; basin-wide rainfall intensity sequences are collected from radar rain gauges; gate opening sequences are collected from the gate control system; generator output and flow sequences are collected from the generator monitoring system; and three-dimensional point cloud data of the riverbed and bank slopes are collected from a multibeam echo sounder on a survey vessel or a lidar system on an unmanned aerial vehicle (UAV). All of the above data are collectively recorded as multi-source observation data. The hydrological telemetry stations refer to automatic hydrological monitoring stations deployed at the upstream control cross-section and downstream reference cross-section of the navigation and hydropower hub, capable of outputting water level and flow values ​​at fixed sampling intervals. The gate control system refers to the PLC control units of each floodgate and ship lock gate of the navigation and hydropower hub, capable of outputting gate opening values ​​at fixed sampling intervals. The generator monitoring system refers to the distributed control system of the generator units of the navigation and hydropower hub, capable of outputting the active power output and corresponding flow through the generators at fixed sampling intervals. The three-dimensional point cloud data of the riverbed and bank slopes is a discrete set of three-dimensional coordinate points obtained by measuring the underwater riverbed elevation using a shipborne multibeam echo sounder or by using an UAV-borne lidar to obtain the elevation of the bank slopes and beaches. Each point contains both planar coordinates and elevation values. The multi-source observation data covers all the key physical quantities involved in the flood control and dispatching of the navigation and hydropower hub, ensuring that the boundary conditions and initial conditions of the subsequent hydrodynamic model can be directly obtained from the measured data, avoiding simulation biases caused by using empirical assumptions.

[0067] S112: Perform a unified spatiotemporal benchmark conversion on the multi-source observation data. Convert the timestamps of all time-series data in the multi-source observation data to Coordinated Universal Time (UTC). Convert the plane coordinates of all coordinate-based data in the multi-source observation data to the CGCS2000 National Geodetic Coordinate System. Convert the elevation datum of all elevation-based data in the multi-source observation data to the 1985 National Elevation Datum. The purpose of this unified spatiotemporal benchmark conversion is to eliminate data spatial location and time alignment deviations caused by time benchmark offsets, coordinate system definition differences, and different elevation reference surfaces between different sensors and measurement systems. In the scenario of a navigation and hydropower hub, hydrological telemetry stations typically use Beijing time, gate control systems may use local system time, and shipboard measurement equipment may use GPS time. If the timestamps are not unified to Coordinated Universal Time (UTC), the records from different data sources at the same physical moment will not be correctly aligned, directly affecting the temporal consistency of the subsequent hydrodynamic model boundary conditions. Similarly, riverbed point cloud data may use a local independent coordinate system or a WGS84 coordinate system, while the gate location coordinates may use an engineering construction coordinate system. The inconsistency between the coordinate system and the elevation datum will cause the gate structure to shift in position in the terrain grid, which in turn will cause errors in the spatial positioning of the gate's internal boundary in the hydrodynamic model.

[0068] S113: Perform data quality inspection and missing data completion processing on the multi-source observation data after unified spatiotemporal benchmark conversion to generate a spatiotemporally coded original dataset. Specifically, the data quality inspection method is as follows: For each time-series data record, calculate the rate of change between two adjacent sampling times. When the rate of change exceeds three times the standard deviation of the historical statistical rate of change for this type of data, the record is labeled as an outlier. For the three-dimensional point cloud data of riverbed and bank slope, calculate the average elevation difference between each point and its eight nearest spatial neighbors. When the average elevation difference exceeds three times the standard deviation of the statistical value of topographic relief in the area, the point is labeled as an outlier. The three-times-standard-deviation threshold is determined based on the assumption of normal distribution: under the assumption of normal distribution, three times the standard deviation covers 99.7% of the normal observation range. Records exceeding this range are statistically extreme deviations and can be identified as anomalies caused by sensor failure or signal interference. The missing data completion process is as follows: For time-series data records labeled as outliers and missing data periods due to communication interruptions, a linear interpolation method is used to fill in the missing data periods using the nearest valid record values ​​before and after the missing data periods; for outlier points in the 3D point cloud data of riverbeds and bank slopes labeled as outliers, the average elevation of the eight nearest normal points in their spatial neighborhood is used for replacement. The linear interpolation method refers to calculating the estimated value of the intermediate time based on the valid record values ​​at both ends of the missing data period, proportionally according to the time interval. After the missing data completion process, a data quality label and a scene label are added to each data record. The data quality label is a three-valued discrete label, with values ​​including original valid, interpolated complete, and outlier replacement, representing the original value directly output by the sensor, the estimated value filled by the interpolation method, and the corrected value after neighborhood replacement, respectively. The scenario labels are enumerated type labels, with values ​​including four scenario types: normal operation, flood season flood, gate regulation, and unit start-up and shutdown. The determination rules are as follows: when the water level value in the upstream section water level sequence exceeds 70% of the design flood level of the navigation and power hub, it is labeled as flood season flood; when the cumulative change in the gate opening sequence within three consecutive sampling intervals exceeds 5% of the gate's full opening stroke, it is labeled as gate regulation; when the unit output sequence shows a jump from zero to non-zero or from non-zero to zero, it is labeled as unit start-up and shutdown. Records that do not meet any of the above conditions are labeled as normal operation. The design flood level and gate full opening stroke are fixed engineering parameters explicitly given in the navigation and power hub design documents. All data records, after unified spatiotemporal benchmark conversion, data quality inspection, and missing data completion processing, are organized into a structured data set with spatial and temporal indices as double keys, denoted as the spatiotemporally encoded original dataset. The spatial index refers to the spatial location identifier corresponding to each data record. For time-series data, it is the CGCS2000 plane coordinates of the station or equipment to which it belongs; for point cloud data, it is the CGCS2000 plane coordinates of each point. The time index refers to the Coordinated Universal Time (UTC) timestamp corresponding to each data record.The spatiotemporal encoded original dataset uses spatial and temporal indices as double keys, enabling subsequent steps to efficiently retrieve required data within any spatial range and time interval, avoiding the computational overhead of full traversal. Data quality labels allow the subsequent hydrodynamic model to assign different confidence weights based on the reliability of the data source when assimilating observational data. Interpolated and supplemented data are given lower weights, while original, valid data is given higher weights, thus reducing the interference of low-quality data on model calibration. Scene labels allow subsequent steps to automatically select appropriate model parameter configuration strategies for different operating conditions. More dense time steps and higher spatial resolution are used in flood season scenarios, while sparser time steps are used in normal operating scenarios to conserve computational resources. See also. Figure 2 This is a schematic diagram illustrating the construction of the spatiotemporal coding raw dataset provided in an embodiment of this application. For example... Figure 2 As shown, the left side lists five types of heterogeneous data sources involved in the flood control and dispatching scenario of the navigation and hydropower hub, including hydrological telemetry stations, radar rain gauges, gate control systems, unit monitoring systems, and measuring equipment. Each data source is labeled with its original time reference and coordinate system, intuitively reflecting the differences in spatiotemporal reference standards between different sensor systems. The middle processing box shows three sequential data preprocessing stages: unified spatiotemporal reference conversion, data quality detection, and missing data completion. This indicates that all raw data achieves format and reference standardization after being processed through these three stages. The right side shows the data structure of the generated spatiotemporally coded raw dataset in a table format, organized with spatial and temporal indices as double keys. Each record ends with a data quality label and a scene label. In the actual operation of the navigation and hydropower hub, hydrological telemetry stations use Beijing time, gate control systems use local system time, and measuring shipboard equipment uses GPS time. If these time reference differences are not eliminated during the data access stage, the boundary conditions of the subsequent hydrodynamic model will not be correctly aligned on the time axis, causing a mismatch between gate opening changes and the corresponding upstream and downstream water level records. The introduction of data quality labels enables the subsequent online assimilation process of ensemble Kalman filtering to automatically identify low-reliability records with interpolation completion or abnormal replacement and reduce their assimilation weights, avoiding bias in model parameters due to sensor failure data. The introduction of scene labels enables the subsequent hydrodynamic calculation model to automatically switch calculation strategies according to the current operating conditions, shortening the time step to improve inference accuracy in flood scenarios during the flood season, and extending the time step to save computing resources in normal operating scenarios.

[0069] Specifically, the spatiotemporally encoded raw dataset is a structured data collection with spatial and temporal indices as double keys. Each data record contains five attribute fields: a unified Coordinated Universal Time (UTC) timestamp, CGCS2000 plane coordinates, elevation values ​​under the 1985 National Height Datum, data quality labels, and scene labels. The core function of the spatiotemporally encoded raw dataset is to unify the raw observation information from five heterogeneous data sources—hydrological telemetry stations, radar rain gauges, gate control systems, unit monitoring systems, and measurement equipment—into an intermediate data structure with consistent spatiotemporal references and quality metadata. In traditional navigation and hydropower hub flood control scheduling systems, these various types of data are stored in their own independent databases or file systems, with formats including fixed-length text format for hydrological telemetry stations, OPC data format for gate control systems, and LAS binary format for point cloud data. Data engineers need to manually write format conversion scripts and check the coordinate system and elevation datum item by item. A single modeling preparation usually takes several days, which is difficult to meet the timeliness requirements for rapid modeling under sudden flood conditions. The spatiotemporally encoded raw dataset undergoes format unification, benchmark conversion, and quality labeling during the data access phase, transforming modeling preparation from manual operations to an automated pipeline process. This reduces the time from data acquisition to usable data for hydrodynamic models from days to minutes. Data quality labels provide a confidence level for subsequent data assimilation in the hydrodynamic model, enabling the model to automatically reduce the weight of interpolated data in parameter correction, preventing spurious data caused by sensor malfunctions from skewing model parameters. Scene labels provide condition identification for subsequent adaptive parameter configuration, allowing the model to automatically increase the computational grid resolution and shorten the time step when it detects a switch from normal operation to flood season, ensuring sufficient spatiotemporal accuracy in flood projection during critical periods.

[0070] S12: Construct a hydrodynamic boundary parameter set based on the spatiotemporal encoded original dataset; further, the construction of the hydrodynamic boundary parameter set includes the following steps:

[0071] S121: Retrieve the water level and flow sequences corresponding to the upstream control sections from the spatiotemporal coding raw dataset according to the spatial index, and construct the upper boundary inflow process with the time index as the horizontal axis. The upper boundary inflow process is a continuous time series curve with Coordinated Universal Time (UTC) as the independent variable and the flow rate of the upstream control section as the dependent variable. Its physical meaning is the amount of water entering the navigation and hydropower hub reservoir area through the upstream control section per unit time. When both water level and flow sequences exist for the upstream control section in the spatiotemporal coding raw dataset, the flow sequence is directly taken as the data source for the upper boundary inflow process; when only the water level sequence exists and the flow sequence is missing, the water level value is converted into a flow value using the water level-flow relationship curve of the upstream control section. The water level-flow relationship curve is a single-value relationship curve fitted after multiple measured flow calibrations of the section, and is stored in the engineering attribute table of the spatiotemporal coding raw dataset. At the same time, retrieve the water level sequence corresponding to the downstream reference section from the spatiotemporal coding raw dataset according to the spatial index, and construct the lower boundary water level process with the time index as the horizontal axis. The lower boundary water level process is a continuous time series curve with Coordinated Universal Time (UTC) as the independent variable and the water level at the downstream reference section as the dependent variable. Its physical meaning is the change in water surface elevation at the downstream outlet of the navigation and hydropower hub over time. The upper boundary inflow process and the lower boundary water level process together constitute the open boundary conditions of the hydrodynamic model. The upper boundary inflow process determines the amount of water entering the model's computational domain, while the lower boundary water level process determines the water surface elevation constraint at the model's computational domain outlet; both are indispensable.

[0072] S122: Retrieve the gate opening sequence corresponding to each floodgate from the spatiotemporally encoded raw dataset according to the spatial index, and retrieve the unit output sequence and unit flow sequence corresponding to each generator unit according to the spatial index, constructing a gate outflow relationship table and a generator outflow relationship table. The gate outflow relationship table is a structured data table with gate number, time index, gate opening value, upstream water level value, and downstream water level value as input columns, and gate flow rate as the output column. Specifically, the gate flow rate is calculated as follows: when the gate opening value is less than the difference between the upstream water level value and the gate bottom plate elevation, it is determined to be a downstream outflow state, and the gate flow rate is calculated using the orifice outflow formula, which is:

[0073] ;

[0074] in For the gate flow rate, The orifice outflow coefficient is... For the clear width of the gate, This represents the gate opening value. Take the acceleration due to gravity as 9.81 meters per second squared. This is the difference between the water level upstream of the gate and the elevation of the gate center. The flow coefficient... The gate's flow coefficient is determined by referring to empirical curves in hydraulic engineering design manuals based on its structural type and opening ratio. Different gate types correspond to different empirical curves. When the gate opening value is greater than or equal to the difference between the upstream water level and the gate bottom elevation, it is determined to be a weir crest overflow state. The gate flow rate is calculated using the broad-crested weir overflow formula, which is:

[0075] ;

[0076] in The overflow flow coefficient of the broad-crested weir. This represents the difference between the upstream water level of the gate and the weir crest elevation. The unit outflow relationship table is a structured data table with unit number, time index, and unit output value as input columns and unit flow rate as output column. When the original spatiotemporal encoded dataset directly contains unit flow rate sequences, the corresponding values ​​of the unit flow rate sequences are directly filled into the unit flow rate column of the unit outflow relationship table. When only the unit output value sequences are contained, the unit output value is converted into a unit flow rate value using the unit characteristic curve. The unit characteristic curve is the curve showing the correspondence between output and flow rate of the unit under different head conditions, provided by the unit manufacturer and stored in the engineering attribute table of the original spatiotemporal encoded dataset. The gate outflow relationship table and the unit outflow relationship table transform discrete gate opening data and unit output data into a continuous flow rate time process, enabling the hydrodynamic model to receive the outflow boundary conditions of the gate and unit in the form of flow rate, without recalculating the gate flow rate within the model, thus reducing the coupling complexity of the model solution.

[0077] S123: Retrieve 3D point cloud data of the riverbed and bank slope from the spatiotemporally encoded raw dataset according to the spatial index. Perform Delaunay triangulation on the 3D point cloud data of the riverbed and bank slope to generate an unstructured triangular terrain mesh covering the reservoir area, the hub structure area, and the downstream river section. Delaunay triangulation is a method of connecting discrete point sets into a triangular mesh. Its characteristic is that the circumcircle of all triangles does not contain other discrete points. This feature ensures that the generated triangles are as close as possible to equilateral triangles, avoiding extremely long and narrow degenerate triangles. In the hydropower hub area, the area near hydraulic structures such as the main channel of the riverbed, gate piers, stilling basin bottom slab, and guide walls is subjected to mesh densification. The spacing between adjacent mesh nodes within the densified area is set to one-tenth of the minimum feature size of the structure in that area. The minimum feature size refers to the minimum value among the gate pier width, stilling basin tooth spacing, and guide wall thickness. In the reservoir area and downstream river section far from the structures, the spacing between adjacent mesh nodes is set to one-twentieth of the river channel width in that area. Mesh refinement provides the hydrodynamic model with sufficient spatial resolution to capture flow characteristics such as surface drops and backflow eddies in areas of rapid flow caused by gate opening and closing. A coarser mesh is used in areas with gentler flow to reduce computational load. Simultaneously, a roughness value is assigned to each triangular mesh cell, forming a roughness distribution map. The roughness value is determined based on the type of substrate material in the area where the mesh cell is located: preferably, a Manning roughness coefficient of 0.013 to 0.015 is used for concrete-lined areas, 0.025 to 0.035 for natural riverbed gravel areas, and 0.040 to 0.060 for vegetated floodplain areas. The specific value of the Manning roughness coefficient is selected within the above range based on the recommended values ​​for the corresponding substrate material in the hydraulic engineering design manual. The roughness distribution map stores the Manning roughness coefficient value for each mesh cell and is a key input parameter for calculating the frictional resistance term in the hydrodynamic model. Spatial differences in roughness values ​​directly affect the model's simulation accuracy of velocity distribution in different areas. See also... Figure 3 This is a schematic diagram of an unstructured triangular terrain mesh provided in an embodiment of this application. For example... Figure 3As shown, the entire computational domain, following the river's direction from left to right, consists of the upstream reservoir area, the navigation and hydropower hub structure area, and the downstream river section. The outer contour lines represent the boundaries on both sides of the river. In the figure, the triangular meshes in the reservoir area on the left and the downstream river section on the right have larger side lengths and lower mesh density, corresponding to sparse mesh configurations in areas with gentle flow. Near the central gate and stilling basin, the triangular meshes have smaller side lengths and significantly increased mesh density, forming a locally denser area centered on the gate piers and stilling basin. The gate structures and stilling basin are marked with gray rectangles, and the three gate piers separate the gate openings into independent flow channels. The dashed boxes mark the three areas: the sparse mesh area in the reservoir area, the dense mesh area near the gate, and the sparse mesh area in the downstream river section. In the flood control operation of the navigation and hydropower hub, the opening and closing of the gates will cause rapid changes in flow patterns near the gate openings, such as violent water level drops, high-speed jets, and hydraulic jumps in the stilling basin. The spatial scale of these flow characteristics is on the same order of magnitude as the gate pier width and the spacing between the stilling basin sills. If the same coarse grid as that used in the reservoir area is applied to these regions, the grid scale will be much larger than the characteristic scale of the rapidly changing flow, leading to over-smoothing of the water surface drop process and incorrect prediction of the hydraulic jump location. By setting the grid node spacing in the intensified areas to one-tenth of the minimum characteristic size of the structure, it is ensured that each rapidly changing flow feature is resolved by at least several grid cells. This allows the interface flux calculation of the finite volume method to capture the local hydraulic gradient changes at the gate. Meanwhile, coarse grids are used in the reservoir area and downstream river section with gentle flow to control the overall computational load.

[0078] S124: Assemble the upper boundary inflow process, lower boundary water level process, gate outflow relationship table, unit outflow relationship table, unstructured triangular terrain grid, and roughness distribution map into a hydrodynamic boundary parameter set. This hydrodynamic boundary parameter set is a structured parameter package containing six components. Each component carries a corresponding time index range and spatial index range, as well as data quality labels and scene labels inherited from the spatiotemporal encoded original dataset. The data quality label inheritance rule in the hydrodynamic boundary parameter set is: when more than 30% of the data records in a component carry interpolated or abnormally replaced data quality labels, that component is labeled as low confidence; otherwise, it is labeled as high confidence. The 30% threshold is determined based on the requirement in hydrological industry standards that the observation data integrity rate should not be less than 70%. The scene label inheritance rule in the hydrodynamic boundary parameter set is: take the scene label that appears most frequently within the same time interval in the spatiotemporal encoded original dataset as the scene label for the hydrodynamic boundary parameter set within that time interval. The hydrodynamic boundary parameter set serves as the sole data input interface for subsequent hydrodynamic model construction and solution. It organizes the boundary conditions, terrain information, and parameter information that were originally scattered across different data sources into a structured parameter package. This allows the hydrodynamic model construction process to be completed automatically by reading the standardized interface of the hydrodynamic boundary parameter set, without the need for manual configuration item by item.

[0079] Specifically, the construction process of the hydrodynamic boundary parameter set transforms the traditional workflow of manually compiling modeling boundary conditions into an automated data aggregation process based on the spatiotemporally encoded raw dataset. In traditional hydrodynamic modeling practices for navigation and hydropower hubs, modelers need to obtain water level and flow data from hydrological stations, gate and unit operation records from the hub's operation and management department, and topographic survey results from the surveying department. These data are then manually edited into the input file format required for the hydrodynamic model. This process is not only time-consuming and labor-intensive but also prone to errors introduced through manual transcription. The hydrodynamic boundary parameter set automates this manual process by defining explicit data retrieval rules, physical formulas, and assembly logic in S121 to S124, enabling the modeling boundary conditions to be automatically updated with the spatiotemporally encoded raw dataset. When the hydrological telemetry station receives new water level and flow data and writes it into the spatiotemporally encoded raw dataset, the upper boundary inflow process and the lower boundary water level process can be automatically extended to the latest time. When the gate control system issues a new opening command and writes it into the spatiotemporally encoded raw dataset, the gate outflow relationship table can automatically calculate the corresponding gate flow rate. This adaptive updating capability ensures that the hydrodynamic boundary parameter set always reflects the latest operating conditions of the navigation and power hub, avoiding the problem of simulation results being out of sync with actual operating conditions caused by the use of lagging data in traditional modeling. The data quality labels inherited in the hydrodynamic boundary parameter set provide hierarchical information on data reliability for the subsequent data assimilation steps of the hydrodynamic model. This allows the model to automatically distinguish between high-confidence and low-confidence boundary data during parameter calibration, assigning larger assimilation weights to high-confidence data and smaller assimilation weights to low-confidence data, thereby improving the robustness of model calibration. The scene labels inherited in the hydrodynamic boundary parameter set provide triggering conditions for the adaptive switching of the model's computational strategy. This allows the model to automatically activate high-resolution computation mode when the scene label is flood season, and switch to low-resolution computation mode when the scene label is normal operation, achieving a dynamic balance between computational accuracy and computational efficiency.

[0080] S2: A two-dimensional unsteady hydrodynamic calculation model is constructed based on the hydrodynamic boundary parameter set. The latest monitoring data is assimilated online to obtain the hydrodynamic simulation state field. The model parameters of the two-dimensional unsteady hydrodynamic calculation model are updated based on the hydrodynamic simulation state field. The specific method for updating the model parameters of the two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic simulation state field is as follows:

[0081] S21: Construct a two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic boundary parameter set, perform time-step solution on the two-dimensional unsteady hydrodynamic calculation model, and generate a hydrodynamic simulation state field; further, generating the hydrodynamic simulation state field includes the following steps:

[0082] S211: Read the unstructured triangular terrain mesh and roughness distribution map from the hydrodynamic boundary parameter set. Use the unstructured triangular terrain mesh as the basis for spatial discretization of the computational domain. Assign the Manning roughness coefficient of each mesh cell in the roughness distribution map to the corresponding computational cell to establish the spatial discretization framework of the two-dimensional unsteady hydrodynamic computational model. The two-dimensional unsteady hydrodynamic computational model is constructed based on a two-dimensional shallow water equation set, which includes three partial differential equations: a continuity equation, a momentum equation in the x-direction, and a momentum equation in the y-direction. The continuity equation is:

[0083] ;

[0084] in Because of the water depth, The depth-average flow velocity is denoted by x. The depth-average flow velocity is y. For time, The source and sink terms represent rainfall or lateral inflow per unit area. The momentum equation in the x-direction is:

[0085] in The elevation of the riverbed bottom. The roughness coefficient is Manning's coefficient. Let be the eddy viscosity coefficient. The momentum equations in the y-direction and x-direction are structurally symmetric. and Replace with and The eddy viscosity coefficient can then be obtained. The Smagorinsky subgrid turbulence model was used for calculations, which expresses the eddy viscosity as...

[0086] ;

[0087] in The Smagorinsky constant takes values ​​between 0.1 and 0.2. The equivalent side length of the current mesh cell. Let be the modulus of the strain rate tensor. The two-dimensional shallow water equations describe the conservation of water mass and momentum under shallow water conditions where the horizontal scale is much larger than the water depth. This system is applicable to flow scenarios in reservoir areas of navigation and hydropower projects and downstream river sections where the water depth is relatively small compared to the channel width. The terms on the right-hand side of the momentum equation represent the driving force of the riverbed slope, the bottom frictional resistance, and the turbulent diffusion effect, respectively. The bottom frictional resistance term incorporates the riverbed sediment characteristics into the model using the Manning roughness coefficient, while the turbulent diffusion term parameterizes turbulent effects smaller than the grid scale using the Smagorinsky subgrid turbulence model.

[0088] S212: Read the upper boundary inflow process, lower boundary water level process, gate outflow relationship table, and unit outflow relationship table from the hydrodynamic boundary parameter set. Set the upper boundary inflow process as the flow boundary condition of the grid cell at the upstream inlet of the two-dimensional unsteady hydrodynamic calculation model, and set the lower boundary water level process as the water level boundary condition of the grid cell at the downstream outlet of the two-dimensional unsteady hydrodynamic calculation model. For the grid cells at the locations of each floodgate and generator unit of the navigation and power hub, set them as internal boundary conditions. The internal boundary conditions refer to the flow constraints applied to specific grid cells within the computational domain. Their physical meaning is that the gate or unit discharges water downstream at that location according to the flow rate specified in the gate outflow relationship table or the unit outflow relationship table. Specifically, the internal boundary conditions are implemented as follows: a virtual interface is inserted between adjacent grid cells upstream and downstream of the gate or unit location. On this virtual interface, the gate flow rate or unit flow rate at the corresponding moment in the gate outflow table or unit outflow table is used as the normal flow rate value passing through the virtual interface, replacing the flux calculation of the two-dimensional shallow water equations on that interface. The setting of internal boundary conditions enables the two-dimensional unsteady hydrodynamic calculation model to accurately reflect the control effect of gate opening and closing and unit start-up and shutdown on the upstream and downstream flow regimes within the computational domain, rather than simplifying the gate and unit to inlet and outlet conditions at the boundary of the computational domain. This allows it to capture nonlinear flow responses such as upstream water level rise and downstream velocity surge when the gate is partially opened. See also... Figure 4 This is a schematic diagram of the gate's internal boundary condition settings provided in an embodiment of this application. For example... Figure 4As shown, a gate structure is drawn at the center of the river's longitudinal profile. Upstream of the gate is the reservoir water, and downstream is the downstream river section water, with the upstream water level higher than the downstream water level, creating a water level difference. Two triangular cells are drawn on each side of the gate, representing adjacent computational grids upstream and downstream, respectively, separated by a virtual interface marked with dashed lines. A thick arrow points from the upstream grid through the virtual interface to the downstream grid, with the gate flow rate indicated next to the arrow. This means the normal flow rate value on the virtual interface is directly taken from the calculation results at the corresponding time in the gate outflow relationship table, replacing the conventional flux solution of the two-dimensional shallow water equations on this interface. During the actual operation of the navigation and hydropower hub, the partial opening or closing of the gate will generate a drastic nonlinear flow response upstream and downstream of the gate. The magnitude and propagation speed of the upstream water level rise and the downstream flow velocity surge depend on the coupling relationship between the gate opening, the upstream and downstream water level difference, and the gate structural parameters. Traditional hydrodynamic models typically simplify gates to inlet and outlet conditions at the boundary of the computational domain. This simplification neglects the gate's impact on the backwater level of the upstream reservoir and its disturbance effect on the downstream near-field flow. By setting up a virtual interface within the computational domain and replacing conventional flux calculations with physical flow rates provided by the gate outflow relationship table, the upstream backwater and downstream rapid flow caused by gate opening and closing can be fully captured within the same continuous computational domain. This provides decision-makers with a physically consistent simulation basis for accurately predicting the impact of gate operations on upstream and downstream water levels and velocities before issuing dispatch instructions.

[0089] S213: The two-dimensional unsteady hydrodynamic calculation model is spatially discretized using the finite volume method, and the discretized equations are solved in time steps using an explicit Runge-Kutta time integration scheme. The finite volume method treats each triangular mesh element as a control volume, integrating the two-dimensional shallow water equations over each control volume, transforming the partial differential equations into an algebraic equation system centered on the interface flux of the control volume. The interface flux is calculated using the Roe approximation Riemann solver. This solver estimates the interface flux by solving a linearized Riemann problem at the interface of adjacent mesh elements, accurately capturing discontinuous solutions such as hydraulic jumps and dam-break waves. The explicit Runge-Kutta time integration scheme is a third-order Runge-Kutta scheme with no increase in total variation. This scheme performs three substep calculations within each time step, with the result of each substep being a weighted combination of the results of the previous substep, ensuring stability while maintaining time accuracy and the stability condition of no increase in total variation. Time step size. The Courant-Friedrichs-Lewy condition is dynamically determined based on the Courant-Friedrichs-Lewy condition, which requires the following:

[0090] ;

[0091] in The Courant number takes values ​​between 0.3 and 0.5. For the first The equivalent side length of each grid cell. For the first Water depth per grid cell For the first The velocity modulus of each grid cell, min, represents the minimum ratio of the spatial step size to the corresponding velocity across all computational cells, serving as the upper limit constraint for the time step. The Courant-Friedrichs-Lewy condition ensures that the information propagation speed does not exceed the speed that the grid resolution can distinguish, thus ensuring the numerical stability of the explicit time integration scheme. The water depth, x-direction depth-average velocity, y-direction depth-average velocity, and water surface elevation values ​​of all grid cells obtained at each time step are written into the hydrodynamic simulation state field using time and spatial indices as double keys. The hydrodynamic simulation state field is a four-dimensional data structure using time and spatial indices as double keys. The spatial dimension corresponds to the grid cell number in the unstructured triangular terrain grid, and the time dimension corresponds to the Coordinated Universal Time (UTC) timestamp for each computational time step. Each spatiotemporal grid point stores four physical quantities: water depth, x-direction depth-average velocity, y-direction depth-average velocity, and water surface elevation. The water surface elevation value in the hydrodynamic simulation state field is equal to the riverbed elevation of the corresponding grid cell plus the water depth value. The hydrodynamic simulation state field serves as a unified storage container for all calculation results of the two-dimensional unsteady hydrodynamic calculation model, providing a spatiotemporally continuous physical quantity data source for subsequent visualization rendering and scheduling index calculation.

[0092] Specifically, the construction and solution process of the two-dimensional unsteady hydrodynamic calculation model transforms the boundary conditions, topographic information, and parameter information in the hydrodynamic boundary parameter set into a spatiotemporally continuous hydrodynamic field with physical consistency. In traditional flood control scheduling of navigation and hydropower hubs, hydrodynamic models are usually run offline. Modelers manually configure boundary conditions and parameters before the flood arrives and start the calculation. After the calculation is completed, the results are exported as a static report. This offline operation mode means that when the operating conditions change during the flood, the existing calculation results cannot reflect the latest water situation changes, and decision-makers can only rely on experience to make scheduling decisions. The two-dimensional unsteady hydrodynamic calculation model sets the upper boundary inflow process and the lower boundary water level process as time-varying boundary conditions, and sets the gate outflow relationship table and the unit outflow relationship table as time-varying internal boundary conditions, so that the model solution process can reflect the latest changes in boundary conditions step by step. The hydrodynamic simulation state field stores the full-field calculation results of each time step with a unified spatiotemporal index structure, so that subsequent steps can retrieve the required water level, flow velocity, and water depth information according to any time interval and spatial range without rerunning the model. The Roe approximate Riemann solver can accurately capture the wavefront propagation process of sudden drops or rises in water level when the gate suddenly opens or closes, avoiding spurious water level jumps caused by numerical oscillations. The Smagorinsky subgrid turbulence model enables the model to reasonably parameterize turbulence effects that cannot be directly resolved at the grid scale, avoiding excessive velocity gradients and water surface instability caused by neglecting turbulence diffusion.

[0093] S22: Assimilate the latest monitoring data online and update the model parameters of the two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic simulation state field; further, updating the model parameters of the two-dimensional unsteady hydrodynamic calculation model includes the following steps:

[0094] S221: In the time-step solution process of the two-dimensional unsteady hydrodynamic calculation model, with a preset assimilation period At intervals, the latest water level and flow rate observations are read from the spatiotemporally encoded original dataset. These observations are then compared with the corresponding water surface elevation and flow velocity values ​​at the corresponding spatial locations and time steps in the hydrodynamic simulation state field, and the observation residual vector is calculated. The assimilation period is... The time interval between two consecutive data assimilation operations is determined based on the scene label in the hydrodynamic boundary parameter set: when the scene label is flood season, the assimilation period is... Set to twice the sampling interval of the hydrological telemetry station; when the scene label is set to normal operation, the assimilation period is... The sampling interval is set to 10 times that of the hydrological telemetry stations. The observation residual vector is a one-dimensional vector with a length equal to the number of available observation stations at the current time, and its i-th... The element is the first The observation residual vector is the difference between the measured water level at each observation station and the water surface elevation of the corresponding grid cell in the hydrodynamic simulation state field at the current time step. The observation residual vector reflects the degree of deviation between the simulation results of the two-dimensional unsteady hydrodynamic calculation model and the actual observation. The larger the absolute value of each element in the observation residual vector, the lower the simulation accuracy of the model in that region, and the more necessary it is to reduce the deviation through parameter correction.

[0095] S222: Based on the observation residual vector, the ensemble Kalman filter method is used to update the model parameters of the two-dimensional unsteady hydrodynamic calculation model. The ensemble Kalman filter method is a sequential data assimilation method based on Monte Carlo sampling. Its core idea is to maintain a set of ensemble samples of the state and parameters of the two-dimensional unsteady hydrodynamic calculation model, approximate the covariance structure of the model error through the statistical characteristics of the ensemble samples, and when new observation data arrives, calculate the Kalman gain matrix using the ensemble covariance and the observation residual vector, thereby correcting the state and parameter values ​​of each sample in the ensemble. Specifically, the execution steps of the ensemble Kalman filter method are as follows: First, generate a set of model parameters. Extract the Manning roughness coefficient of each region in the roughness distribution map and the flow coefficient in the gate outflow relationship table from the hydrodynamic boundary parameter set as parameters to be assimilated. Randomly perturb each parameter to be assimilated within its physically reasonable range according to a Gaussian distribution, generating parameter samples, which is the number of ensemble samples. The value of these samples is determined according to the dimension of the parameter to be assimilated, and is usually set to an integer between 2 and 5 times the dimension of the parameter to be assimilated. Secondly, a time-step solution of one assimilation cycle is run on each parameter sample using a two-dimensional unsteady hydrodynamic calculation model to obtain simulated state samples. Next, the ensemble mean and ensemble covariance matrix of the water surface elevation values ​​at each observation station location are calculated for each simulated state sample, as well as the cross-covariance matrix between the simulated state samples and the parameters to be assimilated. Then, the Kalman gain matrix is ​​calculated using the ensemble covariance matrix, the cross-covariance matrix, and the observation residual vector. The formula for calculating the Kalman gain matrix is:

[0096] ;

[0097] in The cross-covariance matrix is ​​the sum of the simulated state samples and the parameters to be assimilated. for The set covariance matrix of simulated state samples at the locations of observation stations. This is the observation error covariance matrix. This is a diagonal matrix, where the diagonal elements represent the measurement error variance of the water level observations at each observation station. The measurement error variance is determined based on the instrument accuracy specifications of the hydrological telemetry stations. The covariance matrix of the observation errors is then used to calculate this. For observation records in the original spatiotemporal encoded dataset whose data quality labels are interpolation completion or anomalous replacement, the corresponding measurement error variance is amplified to four times the measurement error variance of the original valid records, thereby reducing the weight of low-quality data on parameter correction. Finally, the Kalman gain matrix is ​​used... and observation residual vector pair Each parameter sample is corrected individually, and the correction formula is as follows:

[0098] ;

[0099] in For the first The prior values ​​of each parameter sample. For the first The posterior values ​​of each parameter sample. This is the vector of measured observations corresponding to the observed residual vector. This is the linear mapping matrix that maps the model state to the observation space for the observation operator. Let... The mean of each posterior parameter sample is used as the updated model parameter value. The updated Manning roughness coefficient is written back to the roughness distribution map, and the updated flow coefficient is written back to the gate outflow relationship table. This completes the online update of relevant parameters in the hydrodynamic boundary parameter set, enabling the two-dimensional unsteady hydrodynamic calculation model to use the updated model parameters in subsequent time-step solutions. See also Figure 5 This is a schematic diagram of online assimilation using ensemble Kalman filtering provided in an embodiment of this application. For example... Figure 5As shown, two assimilation moments are marked on the horizontal time axis, dividing the time axis into two assimilation periods of different lengths. The first segment on the left represents the assimilation period under normal operating conditions, with a longer period, indicating that data assimilation is performed at a lower frequency when the water situation is stable to save computational resources. The second segment on the right represents the assimilation period under flood conditions, with a shorter period, indicating that data assimilation is performed at a higher frequency during floods to track rapidly changing water conditions. Several dashed trajectories starting from the left-hand starting point represent the simulated state of the ensemble samples. The trajectories gradually diverge over time, reflecting the accumulation of simulation errors caused by model parameter uncertainties. At the first assimilation moment, the solid circles represent the latest measured water level observations read from the spatiotemporally encoded original dataset. The deviation between the observed values ​​and the simulated trajectories of the ensemble samples is the observation residual vector. After parameter correction using the Kalman gain matrix, the dispersion range of the subsequent ensemble sample trajectories shrinks, indicating that the model parameters have been corrected by the observed data to a state closer to the actual operating conditions. In flood control and dispatching of hydropower projects, factors such as riverbed scouring and silting, seasonal growth of aquatic plants, and wear of gate seals can cause the roughness coefficient and flow coefficient to drift over time. If the model parameters remain fixed after initial calibration, the deviation between the simulation results and actual observations will accumulate continuously with the flood process. The online assimilation mechanism automatically corrects the model parameters using the latest observation data in each assimilation cycle, enabling the model to sense and adapt to these gradually changing operating conditions and maintain simulation accuracy within an acceptable range.

[0100] S223: Substitute the updated model parameters into the two-dimensional unsteady hydrodynamic calculation model and continue to perform subsequent time-step solutions. Use the calculation results of subsequent time steps to overwrite and update the water depth, x-direction depth-average velocity, y-direction depth-average velocity, and water surface elevation values ​​at the corresponding time steps and spatial locations in the hydrodynamic simulation state field. In the next assimilation cycle... Upon arrival, operations S221 and S222 are repeated to form a continuous online data assimilation process.

[0101] Specifically, the online data assimilation mechanism is a core technical feature that distinguishes two-dimensional unsteady hydrodynamic computational models from traditional offline hydrodynamic models. In traditional hydrodynamic modeling practices, model parameter calibration is typically performed offline using one or more historical flood events at the initial stage of modeling. After calibration, the parameters remain fixed. Even if subsequent changes in riverbed scouring and deposition, aquatic plant growth, and gate wear during actual flood events cause shifts in the roughness coefficient and flow coefficient, the model cannot automatically perceive and adapt to these changes. This leads to an accumulation of discrepancies between the model simulation results and actual observations over time. The ensemble Kalman filter method addresses this by implementing an online data assimilation mechanism in each assimilation cycle. The model parameters are corrected using the latest measured water level and flow rates, enabling them to track changes in actual operating conditions. For example, when the riverbed downstream of the hydropower hub experiences a localized decrease in roughness due to flooding, the water surface elevation in the hydrodynamic simulation state field for that region will be higher than the measured water level. The corresponding element in the observation residual vector will be positive. The ensemble Kalman filter method uses the Kalman gain matrix to transform this positive residual into a negative correction for the Manning roughness coefficient in that region, reducing the roughness coefficient and thus lowering the water surface elevation in subsequent time steps, making the simulation results closer to actual observations. This adaptive parameter correction capability allows the two-dimensional unsteady hydrodynamic calculation model to no longer rely on fixed parameters calibrated offline in a single step, but instead continuously self-corrects with the continuous input of observation data, improving the simulation accuracy under long-term and complex operating conditions. (Observation error covariance matrix) The method used to handle the amplified measurement error variance of low-quality data effectively suppresses the impact of outliers on parameter correction when sensor malfunctions cause abnormal observations to enter the assimilation process. This avoids the risk of false observations leading the model parameters in the wrong direction. Assimilation period. The adaptive adjustment strategy based on scene labels enables the model to assimilate observation data at a higher frequency during flood season to track rapidly changing water conditions, and at a lower frequency during normal operation to save computing resources, thus achieving a dynamic balance between computational accuracy and computational efficiency.

[0102] S3: Map the hydrodynamic simulation state field onto a 3D terrain mesh to generate a 3D twin rendering scene, and drive interactive visualization deduction and data rereading based on the 3D twin rendering scene; the specific method for mapping the hydrodynamic simulation state field onto a 3D terrain mesh to generate a 3D twin rendering scene, and driving interactive visualization deduction and data rereading based on the 3D twin rendering scene is as follows:

[0103] S31: Map the hydrodynamic simulation state field onto a three-dimensional terrain mesh to generate a three-dimensional twin rendering scene; further, generating the three-dimensional twin rendering scene includes the following steps:

[0104] S311: Load all node coordinates and unit topology relationships of the unstructured triangular terrain mesh into the 3D rendering engine to generate a 3D terrain rendering mesh. The 3D rendering engine refers to a software framework that supports real-time 3D graphics rendering, capable of receiving mesh vertex coordinates and topology relationships as input and outputting a frame-by-frame updated 3D image. The 3D terrain rendering mesh is a set of triangular patches formed by converting the CGCS2000 plane coordinates and elevation values ​​under the 1985 National Height Datum of each node in the unstructured triangular terrain mesh into 3D vertex coordinates in the 3D rendering engine coordinate system. The three vertex coordinates of each triangular patch correspond one-to-one with the three nodes of the corresponding triangular unit in the unstructured triangular terrain mesh. Simultaneously, load the 3D structural model of the navigation and power hub buildings, including the 3D geometry of structures such as gate leaves, gate piers, stilling basins, powerhouses, lock chambers, and guide walls. Place the 3D structural model onto the 3D terrain rendering mesh according to the actual positions of each structure in the CGCS2000 coordinate system, forming a complete 3D scene base containing terrain and buildings.

[0105] S312: Generate a 3D water surface rendering mesh above the water body coverage area of ​​the 3D terrain rendering mesh. The 3D water surface rendering mesh is a set of triangular facets covering the water body area. Its planar topology is consistent with the topology of the unstructured triangular terrain mesh in the water body area; that is, each triangular facet in the 3D water surface rendering mesh has the same node number and adjacency relationship as the corresponding triangular cell in the unstructured triangular terrain mesh in the water body area. The planar coordinates of each vertex in the 3D water surface rendering mesh are the same as the planar coordinates of the corresponding node in the unstructured triangular terrain mesh. The elevation coordinates are initially set to the riverbed bottom elevation value of the corresponding node, and are dynamically updated to the actual water surface elevation value after receiving data from the hydrodynamic simulation state field. An index mapping table is established, which is a two-column integer lookup table. The first column is the mesh cell number of the water body area in the unstructured triangular terrain mesh in the 2D unsteady hydrodynamic calculation model, and the second column is the vertex number group of the corresponding triangular facet in the 3D water surface rendering mesh. The index mapping table enables the physical quantities of each computational unit in the hydrodynamic simulation state field to be directly located to the corresponding rendering vertex in the 3D water surface rendering mesh through a table lookup operation, eliminating the need for spatial search or coordinate matching calculations. This reduces the time complexity of a single mapping operation from that based on spatial search to... Reduced to table lookup-based ,in This represents the total number of grid cells.

[0106] S313: Real-time reading of water surface elevation and flow velocity values ​​at the current time step and adjacent time steps in the hydrodynamic simulation state field. The water surface elevation values ​​are assigned to the elevation coordinates of the corresponding vertices in the 3D water surface rendering mesh via an index mapping table. When the timestamp of the rendering frame falls between two adjacent computation time steps in the hydrodynamic simulation state field, the water surface elevation values ​​of the two adjacent computation time steps are linearly interpolated according to the linear ratio between the rendering frame timestamp and the timestamps of the two computation time steps. The resulting water surface elevation interpolation is then assigned to the elevation coordinates of the corresponding vertices in the 3D water surface rendering mesh. When the difference in water surface elevation between adjacent computational units in the hydrodynamic simulation state field exceeds one-tenth of the equivalent side length of the adjacent units, an auxiliary rendering vertex is added at the midpoint of the common edge of the adjacent units. The elevation coordinates of the auxiliary rendering vertex are taken as the arithmetic mean of the water surface elevation values ​​of the two adjacent units. This process is used to avoid visual gaps between triangular patches in areas of rapid water surface change. The average depth velocity values ​​in the x and y directions of the hydrodynamic simulation state field are combined into a two-dimensional flow velocity vector. The direction angle of the two-dimensional flow velocity vector is mapped to the scrolling direction of the surface texture of the 3D water surface rendering mesh, and the magnitude of the two-dimensional flow velocity vector is mapped to the texture scrolling speed. This results in a dynamic flow texture effect on the surface of the 3D water surface rendering mesh that matches the actual flow direction and velocity. The texture scrolling direction refers to the offset direction of the UV coordinates of the 3D water surface rendering mesh surface map in each rendering frame, and the texture scrolling speed refers to the offset of the UV coordinates in each rendering frame. The offset is proportional to the magnitude of the two-dimensional flow velocity vector, and the scaling factor is determined according to the conversion relationship between the spatial scale of the 3D rendering engine and the physical spatial scale. The vertex elevation and texture offset of the 3D water surface rendering mesh are updated frame by frame to form a 3D twin rendering scene. The 3D twin rendering scene is a real-time updated 3D image presented in the 3D rendering engine after the 3D terrain rendering mesh, the 3D structural model, and the 3D water surface rendering mesh are superimposed.

[0107] Specifically, the generation process of the 3D twin rendering scene directly associates the physical calculation results in the hydrodynamic simulation state field with the vertex attributes of the 3D rendering mesh through an index mapping table. This ensures that the shape and flow texture of the water surface in each frame of the 3D image are driven by the latest hydrodynamic calculation results, rather than a static stretching of the water surface based on a fixed elevation. In traditional 3D visualization systems for navigation and hydropower hubs, the water surface is usually simplified to a plane, with its elevation set according to the upstream water level telemetry value. The water surface elevation is the same throughout the reservoir area and the downstream river section, failing to reflect the upstream and downstream water level differences caused by the opening and closing of the gates and the water level drop in the stilling basin. The 3D twin rendering scene independently assigns the water surface elevation value of each calculation unit in the hydrodynamic simulation state field to the corresponding rendering vertex, making the water surface elevation in the 3D image present a continuously changing curved surface shape in space. This can intuitively display the backwater area upstream of the gate, the water level drop downstream of the gate, and the hydraulic jump position in the stilling basin. The mapping of flow velocity vectors to texture scrolling direction and speed ensures that the direction and speed of water flow in the 3D rendering are consistent with the physical calculation results. Decision-makers can intuitively determine the mainstream direction of water flow and the location of high-velocity areas after the gate is opened by observing the flow direction and speed of the water surface texture, obtaining spatial distribution information without needing to consult numerical reports. Temporal linear interpolation processing ensures that the frame rate of the 3D rendering is not limited by the hydrodynamic calculation time step. Even if the hydrodynamic calculation time step is several seconds, the 3D rendering can still update smoothly at a rate of 30 or 60 frames per second, avoiding visual discontinuities caused by abrupt changes in water surface morphology. Auxiliary rendering vertex insertion processing for rapidly changing water surface areas eliminates visual gaps caused by excessive elevation differences between adjacent triangular faces, ensuring that rapidly changing flow states such as hydrojumps and water surface drops appear as continuous curved transitions rather than broken steps in the 3D rendering.

[0108] S32: Based on a 3D twin rendering scene, drive interactive visualization deduction and data rereading; further, the driving interactive visualization deduction and data rereading includes the following steps:

[0109] S321: Overlay an engineering operation status layer onto the 3D twin rendering scene. The engineering operation status layer includes gate opening status labels, unit operation status labels, and a water level alarm information panel. The gate opening status labels are text labels placed above the 3D structural model of each gate leaf in the 3D twin rendering scene. The text label content includes the gate number and the current opening value. When the gate opening value is within the most recent assimilation cycle... When changes occur within the system, the background color of the text label switches from the default gray to yellow to indicate changes in opening. The unit operation status label refers to an icon placed at the location of each unit's power plant in the 3D twin-rendered scene. The icon is circular; when the output value of the corresponding unit in the unit output sequence is greater than zero, the icon is green indicating operation; when the output value is equal to zero, the icon is red indicating shutdown. The water level alarm information panel is a semi-transparent rectangular panel fixed in the upper right corner of the 3D twin-rendered scene. The panel content includes the current upstream section water level, downstream section water level, water level change trend, and the determination result of whether the warning water level has been exceeded. The warning water level is a fixed water level value clearly defined in the flood control plan for the navigation and power hub. When the upstream section water level exceeds the warning water level, the border color of the water level alarm information panel switches from the default blue to red, and the water surface elevation corresponding to the warning water level is displayed as a red semi-transparent plane in the 3D twin-rendered scene. The engineering operation status layer overlays gate opening, unit status, and water condition early warning information scattered across different monitoring systems onto a 3D twin rendering scene. This allows decision-makers to obtain complete information on spatial water condition distribution and engineering operation status within the same window without having to frequently switch between 3D screens and table monitoring interfaces.

[0110] S322: Supports interactive cross-section selection and data readback operations in the 3D twin rendering scene. Specifically, when a user draws a straight line segment in any direction in the 3D twin rendering scene, the 3D rendering engine converts the coordinates of the start and end points of the straight line segment from the rendering coordinate system to the CGCS2000 coordinate system. It retrieves the numbers of all grid cells intersecting the straight line segment in the unstructured triangular terrain mesh, extracts the water surface elevation, water depth, and flow velocity values ​​of these grid cells from the hydrodynamic simulation state field at the current time step through an index mapping table, arranges them along the direction of the straight line segment according to their distance from the starting point, and generates a cross-sectional water surface line map and a cross-sectional flow velocity distribution map, which are displayed in a floating window in the 3D twin rendering scene. The cross-sectional water surface line map is a two-dimensional polyline graph with the distance from the starting point as the horizontal axis and the water surface elevation and riverbed bottom elevation as the vertical axes. The area between the water surface elevation polyline and the riverbed bottom elevation polyline is filled in blue to represent water, and the area below the riverbed bottom elevation polyline is filled in brown to represent the riverbed. The cross-sectional velocity distribution map is a two-dimensional line graph with the distance from the starting point as the horizontal axis and the velocity modulus as the vertical axis. When a user clicks on the 3D structural model of any gate or unit in the 3D twin rendering scene, the 3D rendering engine extracts the water surface elevation and velocity values ​​of the corresponding internal boundary conditions at all calculated time steps from the hydrodynamic simulation state field based on the clicked object's number, generating water level time process curves and velocity time process curves, which are displayed in a floating window in the 3D twin rendering scene. The water level time process curve is a two-dimensional line graph with Coordinated Universal Time (UTC) as the horizontal axis and water surface elevation as the vertical axis. The velocity time process curve is a two-dimensional line graph with Coordinated Universal Time (UTC) as the horizontal axis and velocity modulus as the vertical axis. When the user drags the time cursor in the floating window, the water surface morphology in the 3D twin rendering scene synchronously jumps to the screen driven by the hydrodynamic simulation state field data of the time step corresponding to the time cursor, realizing a playback in the time dimension. Data readback operations enable decision-makers to directly obtain quantitative physical data at any location and time within a 3D twin-rendered scene, elevating 3D visualization from a mere visual display tool to a quantitative analysis tool. See also Figure 6 This is a schematic diagram of interactive section selection and data readback provided in an embodiment of this application. For example... Figure 6As shown, the left side is a top-down view of the river channel, filled with an unstructured triangular mesh. A thick solid line segment running through the river channel represents the cross-section selection line manually drawn by the user in the 3D twin rendering scene. Within the area where this line segment intersects with the river channel mesh, the triangular mesh cells that are crossed are highlighted with a bold border, indicating that the 3D rendering engine has automatically retrieved all mesh cell numbers intersecting with the cross-section line. The right side is a cross-section water surface line map generated after extracting data from the hydrodynamic simulation state field based on these mesh cells. The horizontal axis is the distance from the starting point, the vertical axis is the elevation, the upper broken line is the water surface elevation line, and the lower broken line is the riverbed bottom elevation line. The water distribution on the cross-section is filled with light blue between the two broken lines. During the flood control scheduling decision-making process of the navigation and hydropower hub, after observing an abnormal water surface morphology in a certain area of ​​the 3D image, the decision-maker needs to immediately know the specific water depth and flow velocity values ​​of that area to determine whether it exceeds the safety threshold. Traditional systems require decision-makers to exit the 3D view, switch to post-processing software, and manually locate the corresponding position and view the values. This workflow is difficult to meet the timeliness requirements during the rapid evolution of floods. By directly linking the vertices of the 3D rendered mesh with the calculation units of the hydrodynamic simulation state field through an index mapping table, the user's drawing operations in the 3D view can be instantly converted into spatial queries of the hydrodynamic simulation state field. The filled area between the water surface elevation line and the riverbed bottom elevation line in the cross-section water surface line map intuitively shows the water flow area and water depth distribution at each location of the cross-section, enabling decision-makers to simultaneously obtain spatial morphological perception and quantitative data analysis capabilities on the same interface.

[0111] Specifically, the interactive visualization simulation and data readback functions solve the information isolation problem between traditional 3D visualization systems and hydrodynamic calculation results. In traditional systems, 3D images and hydrodynamic calculation results are stored in separate software environments. After viewing the water surface morphology on the 3D image, if decision-makers need to know the specific water level or flow velocity value at a certain location, they need to switch to the hydrodynamic calculation post-processing software, manually select the corresponding grid cell, and view the numerical results. This process is not only time-consuming but also prone to location selection errors due to differences in the coordinate systems of the two software programs. The 3D twin rendering scene directly associates the vertices of the 3D rendering grid with the calculation cells of the hydrodynamic simulation state field through an index mapping table. This allows users' clicks or drawing operations in the 3D image to be immediately converted into spatial queries of the hydrodynamic simulation state field, presenting both 3D spatial morphology and 2D quantitative curves on the same interface. The cross-sectional water surface line map allows decision-makers to intuitively judge the distribution of water flow area and water depth at various locations on the cross-section, thereby estimating whether the flow capacity of the cross-section meets the flood discharge requirements. The flow velocity distribution map allows decision-makers to identify high-velocity areas and assess the scour risk of these areas. Water level time-course curves enable decision-makers to track historical trends in water levels at specific locations and predict future directions of change. Time-cursor-driven 3D playback allows decision-makers to repeatedly observe key periods of flood evolution, compare changes in inundation extent at different times, and provide intuitive spatiotemporal references for the formulation of dispatch plans.

[0112] S4: Calculate flooding and scheduling efficiency indicators based on the 3D twin rendering scene, generate a set of scheduling optimization schemes, and generate execution control instructions based on the set of scheduling optimization schemes for user reference and decision-making; the specific method for calculating flooding and scheduling efficiency indicators based on the 3D twin rendering scene, generating a set of scheduling optimization schemes, and issuing execution control instructions based on the set of scheduling optimization schemes is as follows:

[0113] S41: Calculate flooding indicators and scheduling efficiency indicators based on the 3D twin-rendered scene; further, the calculation of flooding indicators and scheduling efficiency indicators includes the following steps:

[0114] S411: Overlay a protective target vector layer in a 3D twin-rendered scene. The protective target vector layer contains three types of vector elements: town building outline polygons, transportation infrastructure line segments, and key protective target locations. Each type of vector element carries CGCS2000 planar coordinates and attribute fields. The attribute fields for the town building outline polygons include building name, building purpose, and ground elevation. The attribute fields for the transportation infrastructure line segments include road name, road grade, and road surface elevation. The attribute fields for the key protective target locations include target name, target category, and target ground elevation. The target category includes five types: school, hospital, pumping station, substation, and communication base station. Project all vector elements in the protective target vector layer onto the planar coordinates of an unstructured triangular terrain grid. For each vector element, determine the set of all grid cell numbers it covers. Combine each vector element and its covered grid cell number set into a flooding analysis unit. The flooding analysis unit is the basic unit of analysis and evaluation, using a protective target vector element as its spatial scope and the hydrodynamic simulation results of all grid cells covered by that vector element as its data source. See also... Figure 7 This is a schematic diagram of the overlay of vector graphics layers for the protected target provided in an embodiment of this application. For example... Figure 7 As shown, the bottom layer is an unstructured triangular terrain grid covering the river channel and its banks. Three types of vector elements from the protected target vector layer are overlaid on top of this grid. A town building outline is marked with a solid polygon in the upper left corner, covering four triangular grid cells below it. The covered grid cells are highlighted with gray fill. A dashed broken line traverses the riverbanks, representing transportation infrastructure segments. A key protected target location is marked with a pentagram in the upper right corner, indicating its category as a school. A dashed box encloses the building outline polygon and its covered grid cells as a whole, with a label indicating that it constitutes a flood analysis unit. In downstream flood control management of the hydropower hub, different types of protected targets have varying flood tolerance. Densely populated areas such as schools and hospitals have much higher safety requirements for water depth and flow velocity than general buildings, and transportation roads have independent standard requirements for safe wading flow velocity. By binding each vector element to its covered grid cell as an independent inundation analysis unit, the calculation of subsequent inundation indicators can be accurate to each specific protection target. This allows decision-makers to directly see the differences in the maximum inundation depth and flow velocity exceeding the limit for a certain school or road when comparing scheduling schemes. This realizes the transformation of flood impact assessment from a general regional evaluation to a refined quantification of each target.

[0115] S412: Traverse all calculated time steps of data in the hydrodynamic simulation state field and calculate the inundation index for each inundation analysis unit. The inundation index includes four components: maximum inundation depth, inundation start time, inundation end time, and cumulative duration of flow velocity exceeding the limit. The maximum inundation depth is calculated as follows: Traverse the depth values ​​of all time steps in the hydrodynamic simulation state field across all grid cell numbers covered by the inundation analysis unit, and take the maximum depth value of all grid cells across all time steps as the maximum inundation depth of the inundation analysis unit. The inundation start time is calculated as follows: When the water surface elevation value of any grid cell first exceeds the ground elevation of the corresponding vector feature of the inundation analysis unit, record the Coordinated Universal Time (UTC) timestamp of that time step as the inundation start time. The ground elevation is determined by the corresponding attribute field value based on the vector feature type: ground elevation field value for urban building outline polygons, road surface elevation field value for traffic infrastructure segments, and target ground elevation field value for key protection target locations. The inundation termination time is calculated as follows: after the inundation start time, when the water surface elevation values ​​of all grid cells covered by the inundation analysis unit fall below the ground elevation, the Coordinated Universal Time (UTC) timestamp of this time step is recorded as the inundation termination time. The cumulative duration of flow velocity exceeding the limit is calculated as follows: a flow velocity safety threshold is set, determined based on the type of vector element corresponding to the inundation analysis unit. For urban building outline polygons, the flow velocity safety threshold is the critical flow velocity for human standing stability, i.e., the flow velocity when the product of water depth and flow velocity equals the safety threshold. For traffic infrastructure segments, the flow velocity safety threshold is the critical flow velocity for vehicle wading. For key protection target locations, the flow velocity safety threshold is the maximum allowable flow velocity specified in the flood control design standard for that type of facility. The critical flow velocity for human standing stability is derived by inversely from the condition that the product of water depth and flow velocity does not exceed the safety threshold. The safety threshold is the upper limit of the product of water depth and flow velocity for an adult to maintain stable standing on flat ground; this upper limit is taken from experimental research data in the field of flood emergency management. The number of time steps in the grid cells covered by the inundation analysis unit that exceeded the corresponding velocity safety threshold was counted, and multiplied by the hydrodynamic calculation time step to obtain the cumulative duration of velocity exceeding the limit. The four components of the inundation index quantify the impact of flooding on each protected target from different dimensions: the maximum inundation depth reflects the risk of physical damage; the inundation start and end times reflect the warning time window and inundation duration; and the cumulative duration of velocity exceeding the limit reflects the risks of scour damage and personnel safety. See also... Figure 8 This is a schematic diagram illustrating the calculation of flooding indicators provided in an embodiment of this application. For example... Figure 8As shown, the left side is the water level time-matter curve, with time on the horizontal axis and water surface elevation on the vertical axis. A horizontal dashed line marks the ground elevation of the protected target. The water level curve gradually rises from the initial low water level and crosses the ground elevation line; the moment of crossing is the start of inundation, marked by a vertical dashed line on the time axis. The water level curve continues to rise to its peak and then gradually falls back; the moment it crosses the ground elevation line again is the end of inundation. The period between the start and end of inundation is marked by a double-headed arrow as the inundation duration. During this period, the area between the water level curve and the ground elevation line is filled with light blue to represent the inundated water body. The vertical double-headed arrow between the peak water level curve and the ground elevation line indicates the maximum inundation depth. The right side is a simplified cross-sectional view of the protected target. The rectangular building sits above the ground elevation, and the light blue filled area represents the flood inundation range. The water surface line is marked by a dashed line, and the vertical double-headed arrow between the water surface line and the ground also indicates the maximum inundation depth, corresponding to the peak water depth in the water level curve on the left. In flood control scheduling of navigation and hydropower hubs, the maximum inundation depth directly determines the degree of structural damage to buildings and the magnitude of indoor property loss; the duration of inundation determines the soaking time of building foundations and the post-disaster recovery cost; and the inundation start time provides an available response window for emergency management departments to evacuate personnel. By extracting and quantifying these four components from the hydrodynamic simulation state field target by target, decision-makers can accurately assess the differences in the impact of each scheme on each protection target during the scheduling scheme comparison phase.

[0116] S413: Calculate scheduling efficiency indicators based on the hydrodynamic simulation state field and hydrodynamic boundary parameter set. The scheduling efficiency indicators include three components: power generation, navigation water level guarantee rate, and ecological flow satisfaction rate. The power generation indicator is calculated as follows: The unit outflow time series of all calculated time intervals for each unit is read from the unit outflow relationship table; the upstream and downstream water surface elevations of each unit's location are read from the hydrodynamic simulation state field; and the gross head of each unit at each time step, i.e., the difference between the upstream and downstream water surface elevations (in the power generation calculation formula, this is expressed as the difference between the unit's overall efficiency). (Consideration of head loss), power generation formula:

[0117] ;

[0118] Calculate the power generation at each time step, where The overall efficiency of the unit is obtained from the unit characteristic curve. Take the density of water as 1000 kg per cubic meter. Take the acceleration due to gravity as 9.81 meters per second squared. For the unit's overflow rate at this time step, The gross head for this time step. The power generation of all time steps is multiplied by the corresponding calculation time step length and then summed to obtain the power generation index. The navigation water level guarantee rate index is calculated as follows: The time series of water surface elevation values ​​at the upstream and downstream approach channels of the lock are read from the hydrodynamic simulation state field. The percentage of time steps where the water surface elevation is greater than or equal to the minimum navigation water level is counted out of the total number of time steps, and this percentage is used as the navigation water level guarantee rate index. The minimum navigation water level is the minimum water surface elevation value required for safe navigation of ships as specified in the channel design documents. The ecological flow satisfaction rate index is calculated as follows: The time series of flow velocity values ​​at the downstream ecological section of the navigation and power hub is read from the hydrodynamic simulation state field. The flow rate across the section is calculated for each time step based on the flow velocity value and the cross-sectional area. The percentage of time steps where the flow rate across the section is greater than or equal to the minimum ecological flow is counted out of the total number of time steps, and this percentage is used as the ecological flow satisfaction rate index. The minimum ecological flow is the minimum flow value determined by the ecological environment protection management department based on the habitat requirements of downstream aquatic organisms. The three components of the scheduling efficiency index quantify the overall performance of the current scheduling plan from three dimensions: economic benefits, transportation benefits, and ecological benefits, enabling decision-makers to make a quantitative trade-off between flood control safety and overall benefits.

[0119] S42: Generate a set of scheduling optimization schemes, and issue execution control instructions based on the set of scheduling optimization schemes; further, generating the set of scheduling optimization schemes and issuing execution control instructions includes the following steps:

[0120] S421: Generate a set of candidate scheduling schemes. Each candidate scheduling scheme in the set is expressed as a gate opening time series and a unit output time series. The gate opening time series is a piecewise constant time series with Coordinated Universal Time (UTC) as the independent variable and the opening value of each gate as the dependent variable. The duration of each segment is equal to the minimum execution interval of the scheduling command, which is determined based on the mechanical response time of the gate opening and closing mechanism. The unit output time series is a piecewise constant time series with Coordinated Universal Time (UTC) as the independent variable and the output value of each unit as the dependent variable. The candidate scheduling scheme is generated as follows: Using the latest state of the current hydrodynamic boundary parameter set, including the gate outflow relationship table and the unit outflow relationship table, as the baseline scheme, discrete candidate values ​​are generated for each gate opening value within a preset step range above and below the current opening value. Similarly, discrete candidate values ​​are generated for each unit output value within a preset step range above and below the current output value. The preset step range refers to the discrete adjustment increment of each gate opening value or unit output value based on the current value when generating the candidate scheduling scheme. The determination rules are as follows:

[0121] The preset step size for the gate opening is taken as the minimum adjustment amount of the gate opening and closing mechanism. The minimum adjustment amount refers to the minimum opening adjustment increment that the gate opening and closing mechanism can perform in a single operation. This is determined by the mechanical transmission accuracy of the gate opening and closing mechanism and is an inherent technical parameter of the gate equipment. For example, if the minimum adjustment amount of the gate opening and closing mechanism is 0.1 meters, then the preset step size for the gate opening is 0.1 meters.

[0122] The preset step size for unit output is taken as the minimum output adjustment amount of the unit speed governor. The minimum output adjustment amount refers to the minimum active power output adjustment increment that the unit speed governor can perform in a single operation. It is determined by the control accuracy of the speed governor and is an inherent technical parameter of the unit equipment. For example, if the minimum output adjustment amount of the unit speed governor is 1MW, then the preset step size for unit output is 1MW.

[0123] The discrete candidate opening values ​​of all gates and the discrete candidate output values ​​of all generating units are combined using a Cartesian product. One value is selected from the candidate opening value set of each gate, and one value is selected from the candidate output value set of each generating unit. All possible combinations of these values ​​form a candidate scheduling scheme set. When the number of schemes in the candidate scheduling scheme set exceeds a preset maximum number of schemes, a representative subset of schemes not exceeding the maximum number of schemes is extracted from all combinations using a Latin hypercube sampling method. The maximum number of schemes is determined based on the parallel computing capability of the high-performance computing environment and the computation time of a single scheme, ensuring that all candidate schemes can be calculated within a preset time limit. The Latin hypercube sampling method is a hierarchical random sampling method. Its characteristic is that the value range of each parameter to be sampled is equally divided into intervals equal to the number of samples. A value is randomly selected within each interval, ensuring that the sampling points uniformly cover the entire value range in each parameter dimension, avoiding local clustering that may occur with random sampling.

[0124] S422: For each candidate scheduling scheme in the candidate scheduling scheme set, replace the gate outflow relationship table and unit outflow relationship table in the hydrodynamic boundary parameter set with its gate opening time series and unit output time series. Use a two-dimensional unsteady hydrodynamic calculation model for rapid derivation calculation to generate the candidate hydrodynamic simulation state field corresponding to that candidate scheduling scheme. The rapid derivation calculation is the same as the time-step solution process in S213, using the same finite volume method and explicit Runge-Kutta time integration format. However, the online data assimilation operations in S221 and S222 are not performed in the rapid derivation calculation to reduce the computation time of a single scheme. The rapid derivation calculations of multiple candidate scheduling schemes are executed simultaneously in parallel in a high-performance computing environment, with each candidate scheme assigned an independent computation process. For each candidate hydrodynamic simulation state field, perform the inundation index and scheduling benefit index calculations in S411 to S413 to obtain the inundation index value and scheduling benefit index value corresponding to each candidate scheduling scheme.

[0125] S423: Based on the inundation index values ​​and scheduling benefit index values ​​of all candidate scheduling schemes, construct a set of optimized scheduling schemes. Specifically, first, perform a safety constraint check: for each candidate scheduling scheme, verify whether the maximum inundation depth of the inundation analysis unit for the key protection target type in all its inundation analysis units is less than the maximum allowable inundation depth specified in the flood control design standard for that type of protection target, and verify whether the cumulative duration of flow velocity exceeding the limit is less than the maximum allowable flow velocity exceeding the limit duration specified in the flood control design standard for that type of protection target. The maximum allowable inundation depth and the maximum allowable flow velocity exceeding the limit duration are fixed values ​​explicitly specified in the flood control scheme for each key protection target category of the navigation and hydropower hub. Candidate scheduling schemes that do not meet any safety constraint conditions are removed from the candidate scheduling scheme set. Then, a multi-objective ranking is performed on the remaining candidate scheduling schemes: minimizing the maximum value of the maximum inundation depth in all inundation analysis units (i.e., using the minimization-maximization criterion to minimize the most unfavorable maximum inundation depth in all inundation analysis units), maximizing the power generation index, maximizing the navigation water level guarantee rate index, and maximizing the ecological flow satisfaction rate index are the four optimization objectives. A non-dominated ranking method is used to rank the remaining candidate scheduling schemes. This non-dominated ranking method is a Pareto ranking method in multi-objective optimization. Its rule is: when candidate scheme A is not inferior to candidate scheme B in all four optimization objectives, and is strictly superior to candidate scheme B in at least one optimization objective, candidate scheme A is said to dominate candidate scheme B. All candidate schemes not dominated by any other candidate scheme are assigned to the first Pareto front layer; all candidate schemes not dominated by other candidate schemes after removing the first Pareto front layer are assigned to the second Pareto front layer, and so on. All candidate scheduling schemes in the first Pareto front layer are taken as the scheduling optimization scheme set. The dispatch optimization scheme set is an ordered set containing one or more candidate dispatch schemes. Each candidate dispatch scheme carries its gate opening time series, unit output time series, inundation index value, and dispatch benefit index value. The candidate dispatch schemes in the set are arranged from largest to smallest power generation index. All candidate dispatch schemes in the dispatch optimization scheme set satisfy safety constraints and there is no Pareto dominance relationship among the four optimization objectives. Decision-makers can select the final execution scheme based on the current flood control situation's priorities. When two candidate schemes have advantages and disadvantages in the four optimization objectives but do not dominate each other, both are included in the same Pareto front layer and are both retained in the dispatch optimization scheme set for decision-makers to choose from. The Pareto ranking does not forcibly rank the superiority or inferiority among mutually non-dominant schemes; instead, it presents all mutually non-dominant schemes as equivalent optimal compromise schemes with different trade-offs among multiple objectives, allowing decision-makers to select the final execution scheme based on the specific needs of the current flood control situation. See also Figure 9 This is a schematic diagram of multi-objective sorting of the scheduling optimization scheme set provided in the embodiments of this application. For example... Figure 9As shown, the horizontal axis represents the maximum inundation depth, with smaller values ​​being better, and the vertical axis represents the power generation index, with larger values ​​being better. Each marker within the coordinate area represents a candidate scheduling scheme. Four solid black dots are located in the upper left of the coordinate area, connected by a dashed broken line to form the first Pareto front layer. A side note indicates that this is the set of optimal scheduling schemes; these schemes do not exhibit a Pareto dominance relationship between the maximum inundation depth and the power generation index. The hollow gray dots distributed in the lower right of the front layer represent non-front-edge qualified schemes that pass the safety constraint check but are not located in the first Pareto front layer. The cross markers on the right side of the coordinate area represent schemes that were eliminated by the safety constraint check because the cumulative duration of the maximum inundation depth or flow velocity exceeding the limit for key protection targets exceeded the allowable value of the flood control design standard. In the actual decision-making of flood control scheduling in navigation and power hubs, there is an inherent contradiction between flood control safety and power generation benefits: increasing the discharge volume helps reduce downstream inundation risk but reduces the power generation head and turbine flow rate; decreasing the discharge volume helps maintain power generation benefits but may increase downstream inundation risk. By stratifying all candidate schemes through non-dominated sorting and using safety constraints as a hard threshold to eliminate schemes that do not meet the flood control safety baseline, the decision-makers obtain a set of optimized scheduling schemes that meet safety requirements. Furthermore, each scheme represents the optimal compromise between different trade-offs between flood control safety and economic benefits, allowing decision-makers to choose the focus based on the urgency of the current flood situation.

[0126] S424: Convert the gate opening time series and unit output time series of the final execution plan selected by the decision-maker from the scheduling optimization scheme set into execution control instructions for user reference and decision-making. The execution control instructions are structured instruction messages conforming to the communication protocol of the avionics hub scheduling automation system. Each execution control instruction contains four fields: target device number, target parameter name, target parameter value, and instruction execution timestamp. Specifically, for each constant segment of the gate opening time series in the final execution plan, an execution control instruction is generated. Its target device number is the device number of the corresponding gate, the target parameter name is the gate opening, the target parameter value is the gate opening value for that segment, and the instruction execution timestamp is the start Coordinated Universal Time (UTC) timestamp for that segment. For each constant segment of the unit output time series in the final execution plan, an execution control instruction is generated. Its target device number is the device number of the corresponding unit, the target parameter name is the unit output, the target parameter value is the unit output value for that segment, and the instruction execution timestamp is the start Coordinated Universal Time (UTC) timestamp for that segment. After user confirmation, all execution control commands are sent to the navigation and power hub dispatch control department in ascending order of execution timestamp. Upon confirmation from the dispatch control department, the dispatch automation system executes the commands sequentially when their corresponding execution timestamps arrive. After the execution control commands are sent, the actual execution status of the gate control system and the unit monitoring system is written into the spatiotemporally encoded original dataset as new gate opening sequences and unit output sequences. This triggers automatic updates to the hydrodynamic boundary parameter set and continuous online solving of the two-dimensional unsteady hydrodynamic calculation model, forming a complete closed loop from data acquisition, hydrodynamic simulation, three-dimensional visualization, dispatch optimization to control execution and back to data acquisition. (See also...) Figure 10 This is a schematic diagram of the closed-loop execution control instructions provided in the embodiments of this application. For example... Figure 10As shown, five rectangles are arranged from top to bottom: the scheduling optimization scheme set, the generation and execution of control commands, the gate and unit execution regulation, the actual gate opening sequence and the actual unit output sequence, and the spatiotemporal encoded raw dataset. Adjacent rectangles are connected by downward arrows to indicate unidirectional data and command transmission. Two loops extend from the bottom spatiotemporal encoded raw dataset rectangle, passing through the left and right sides respectively and returning to the top scheduling optimization scheme set rectangle, forming a complete closed loop. The label next to the right loop indicates that the automatic update of the hydrodynamic boundary parameter set is triggered, meaning that after the actual execution data is written into the spatiotemporal encoded raw dataset, it will automatically drive the hydrodynamic boundary parameter set to update its gate outflow relationship table and unit outflow relationship table. The label next to the left loop indicates that the model is re-solved and the scheme is regenerated, meaning that the updated hydrodynamic boundary parameter set will drive the two-dimensional unsteady hydrodynamic calculation model to continue solving online and updating the hydrodynamic simulation state field, thereby triggering a new round of scheduling optimization scheme generation. In the event of a sudden flood at a navigation and hydropower hub, the effectiveness of the dispatching plan may deviate from expectations due to the actual inflow exceeding the forecast. In traditional dispatching processes, the correction cycle from detecting the deviation to re-formulating and issuing the plan typically lasts from tens of minutes to several hours. This closed-loop mechanism allows the execution results of the dispatching plan to automatically flow back to the data acquisition layer and propagate downstream along the entire link. After each assimilation cycle, the system can automatically generate a set of corrected and optimized dispatching plans based on the latest operating conditions, compressing the correction response cycle from hours to minutes, thus gaining a time window for emergency dispatching.

[0127] Specifically, the process of constructing the optimized scheduling scheme set transforms the qualitative decision-making model of traditional flood control scheduling in hydropower hubs, which relies on the experience and judgment of dispatchers, into a quantitative multi-objective optimization decision-making model based on hydrodynamic physical calculations. In traditional flood control scheduling practices of hydropower hubs, after receiving upstream inflow forecasts, dispatchers typically select the scheduling scheme closest to the current inflow level from a pre-set fixed scheduling scheme table in the scheduling procedures. The number of schemes in the scheduling scheme table is limited and based on design conditions, failing to cover various upstream and downstream water level combinations and gate unit joint operation conditions that may occur during actual floods. When the actual flood process deviates from the design conditions, dispatchers need to make temporary adjustments to the scheduling scheme based on experience within a limited time. Such adjustments lack quantitative assessment of the consequences, and decision-makers cannot accurately know the differences in the inundation impact of different adjustment schemes on downstream towns, transportation facilities, and key protection targets before issuing scheduling instructions. They also cannot accurately make quantitative trade-offs between flood control safety and power generation benefits, navigation assurance, and ecological flow.

[0128] The scheduling optimization scheme set automatically generates a set of candidate scheduling schemes covering gate opening and unit output adjustment ranges based on the current actual operating conditions in S421. In S422, a two-dimensional unsteady hydrodynamic calculation model is used to physically simulate each candidate scheduling scheme and calculate inundation and scheduling benefit indices. In S423, safety constraint checks and non-dominated sorting are used to screen out the optimal schemes that meet safety requirements and do not have Pareto dominance relationships among multiple benefit objectives. This means that decision-makers no longer receive a single fixed scheme, but a set of comparable schemes that have been physically verified and evaluated for multiple objectives. Decision-makers can intuitively see the maximum inundation depth, inundation duration, and flow velocity exceedance time for each inundation analysis unit corresponding to each scheme in the scheduling optimization scheme set, as well as the power generation, navigation water level guarantee rate, and ecological flow satisfaction rate that the scheme can achieve, thereby making informed choices based on the current flood control situation. For example, when the flood peak is about to arrive and downstream towns face the risk of inundation, decision-makers can select the scheme with the smallest maximum inundation depth from the set of optimal scheduling schemes, even if the power generation index of the scheme is low; when the downstream inundation risk has been eliminated during the flood receding stage, decision-makers can select the scheme with the largest power generation index from the set of optimal scheduling schemes to improve economic efficiency.

[0129] The candidate scheduling scheme set is generated by discretizing and combining the baseline schemes according to the minimum adjustment amount of the equipment. This ensures that each candidate scheme is physically feasible and avoids infeasible schemes that exceed the gate travel range or the rated output range of the unit. When the number of combinations is too large, the Latin hypercube sampling method is used for dimensionality reduction sampling. Compared with pure random sampling, this method can more evenly cover the parameter space with the same number of samples, reducing the risk of missing optimal schemes due to sampling bias. The design choice of not performing online data assimilation in the rapid extrapolation calculation is based on the following considerations: the purpose of candidate scheme extrapolation is to compare the relative differences between different scheduling strategies, rather than to obtain the absolutely accurate prediction value of a certain scheme. Therefore, omitting the data assimilation step while maintaining the consistency of the model's physical framework can reduce the computation time of a single scheme, allowing more candidate schemes to be evaluated within the limited extrapolation time limit, thus expanding the coverage of the optimization search space. The process of generating execution control commands converts the gate opening time series and unit output time series from the scheduling optimization scheme set into structured command messages that the scheduling automation system can directly recognize and execute. This eliminates the intermediate step in the traditional scheduling process where the dispatcher verbally issues commands followed by manual operation of the gate opening and closing mechanisms and unit speed governors by the operators. The four fields in the execution control command—target equipment number, target parameter name, target parameter value, and command execution timestamp—cover all the information required for the scheduling automation system to execute a scheduling command. This eliminates the need for secondary manual interpretation and parameter conversion during command transmission, reducing execution deviations caused by delays in manual operation or misreading of parameters. The setting of the command execution timestamp allows the scheduling automation system to execute gate opening adjustments and unit output adjustments sequentially according to the time sequence, ensuring that the adjustment sequence between multiple gates and units is consistent with the sequence designed in the candidate scheduling scheme, and avoiding drastic fluctuations in upstream and downstream water levels caused by simultaneous adjustments of multiple devices.

[0130] In the traditional flood control scheduling process of navigation and hydropower hubs, there is often an information gap between scheduling decisions and execution. The hydrological information used by dispatchers when formulating scheduling plans comes from static data at the time of plan formulation. After the plan is issued and executed, dispatchers need to passively observe the execution effect through the monitoring system. When the execution effect deviates from expectations, it is necessary to collect hydrological data again, recalculate, re-formulate the plan, and reissue instructions. The entire correction cycle usually takes tens of minutes to several hours. This step, by automatically feeding back the execution results of the scheduling plan to the data acquisition layer and triggering subsequent automatic updates across the entire chain, compresses the correction cycle to one assimilation cycle of hydrodynamic model calculation. Including the total time consumed by parallel simulation of candidate solutions, a closed-loop correction response can be achieved within minutes with the support of a high-performance computing environment. This minute-level closed-loop correction capability is particularly crucial for emergency dispatching under sudden flood conditions exceeding the forecast level. When a sudden increase in water inflow occurs during a flood, exceeding the forecast level, the system can complete the entire process from sensing the change in water inflow to outputting a corrected dispatching plan and issuing execution instructions within minutes, gaining a time window for dispatching decisions and reducing the risk of downstream flooding caused by response delays.

[0131] By integrating the technical steps S1 to S4, this invention constructs a complete technical chain from multi-source observation data acquisition, spatiotemporal coding and quality labeling, automatic construction of hydrodynamic boundary parameter sets, online solution and data assimilation of two-dimensional unsteady hydrodynamic models, generation of three-dimensional twin rendering scenes and interactive data rereading, quantitative calculation of inundation indicators and scheduling benefit indicators, generation of multi-objective scheduling optimization schemes, to the closed-loop issuance of execution control commands. The spatiotemporal coding raw dataset serves as the data foundation for the entire chain, achieving integrated organization of multi-source heterogeneous data through unified spatiotemporal benchmarks and quality labeling. The hydrodynamic boundary parameter set, acting as a bridge between the data layer and the model layer, automatically aggregates scattered observation data into structured inputs that can be directly consumed by the hydrodynamic model. The two-dimensional unsteady hydrodynamic calculation model, through solving shallow water equations and an online assimilation mechanism using ensemble Kalman filtering, elevates static offline hydrodynamic simulation to dynamic real-time hydrodynamic extrapolation. The 3D twin rendering scene directly links hydrodynamic calculation results with the 3D rendering mesh through an index mapping table, achieving frame-by-frame synchronous driving from physical quantities to visual representations and interactive data back-reading from visual representations to physical quantities in a two-way linkage. The scheduling optimization scheme set, through parallel simulation of multiple schemes, quantitative evaluation of inundation indicators and scheduling benefit indicators, and multi-objective optimization of non-dominated sorting, transforms the 3D twin rendering scene from a passive display tool into a decision engine that actively outputs scheduling strategies. The closed-loop issuance of execution control commands and the automatic feedback of execution status form the system's continuous self-correction capability, enabling the entire digital twin system to evolve synchronously with actual operating conditions throughout the entire lifecycle of the flood process.

[0132] Example 2:

[0133] This embodiment, based on Embodiment 1, provides a dynamic flood digital twin simulation system for navigation and hydropower hubs that integrates hydrodynamic models, such as... Figure 11 As shown, it includes:

[0134] Multi-source data acquisition and hydrodynamic boundary parameter construction module: used to acquire multi-source observation data, form a spatiotemporally encoded raw dataset, and construct a hydrodynamic boundary parameter set based on the spatiotemporally encoded raw dataset;

[0135] Hydrodynamic Model Construction and Online Assimilation Module: This module is used to construct a two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic boundary parameter set, assimilate the latest monitoring data online to obtain the hydrodynamic simulation state field, and update the model parameters of the two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic simulation state field.

[0136] 3D Twin Rendering and Interactive Inference Module: Used to map the hydrodynamic simulation state field onto a 3D terrain mesh to generate a 3D twin rendering scene, and drive interactive visualization inference and data rereading based on the 3D twin rendering scene;

[0137] The scheduling optimization and instruction issuance module is used to calculate flooding indicators and scheduling efficiency indicators based on the 3D twin rendering scene, generate a set of scheduling optimization schemes, and generate execution control instructions based on the set of scheduling optimization schemes for users to refer to and make decisions.

Claims

1. A method for dynamic flood digital twin simulation of navigation and hydropower hubs integrating hydrodynamic models, characterized in that, The method includes: S1: Collect multi-source observation data, perform unified spatiotemporal benchmark conversion and quality labeling on the multi-source observation data to form a spatiotemporally coded raw dataset, and construct a hydrodynamic boundary parameter set based on the spatiotemporally coded raw dataset; the multi-source observation data includes upstream cross-section water level and flow sequences collected from hydrological telemetry stations, basin surface rainfall intensity sequences collected from radar rain gauges, gate opening sequences collected from gate control systems, unit output and flow sequences collected from unit monitoring systems, and three-dimensional point cloud data of riverbed and bank slope collected from survey vessel-borne multibeam echo sounders or UAV-borne lidar; The specific method for constructing the hydrodynamic boundary parameter set based on the spatiotemporally encoded raw dataset is as follows: The water level and flow sequences corresponding to the upstream control sections are retrieved from the spatiotemporally encoded raw dataset according to the spatial index, and the upper boundary inflow process is constructed with the time index as the horizontal axis. Simultaneously, the water level sequences corresponding to the downstream reference sections are retrieved from the spatiotemporally encoded raw dataset according to the spatial index, and the lower boundary water level process is constructed with the time index as the horizontal axis. The gate opening sequences corresponding to each flood discharge gate are retrieved from the spatiotemporally encoded raw dataset according to the spatial index, and the generator output sequences and generator sets corresponding to each generator unit are retrieved according to the spatial index. Flow sequences were used to construct gate outflow relationship tables and unit outflow relationship tables. Three-dimensional point cloud data of the riverbed and bank slopes were retrieved from the spatiotemporally encoded raw dataset according to spatial indexes. Delaunay triangulation was performed on the riverbed and bank slope three-dimensional point cloud data to generate an unstructured triangular terrain grid covering the reservoir area, the hub structure area, and the downstream river section. Simultaneously, a roughness value was assigned to each triangular grid cell to form a roughness distribution map. The upper boundary inflow process, lower boundary water level process, gate outflow relationship table, unit outflow relationship table, unstructured triangular terrain grid, and roughness distribution map were assembled into a hydrodynamic boundary parameter set. S2: Construct a two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic boundary parameter set, obtain the hydrodynamic simulation state field by online assimilation of the latest monitoring data, and update the model parameters of the two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic simulation state field; The specific method for updating the model parameters of the two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic simulation state field is as follows: A two-dimensional unsteady hydrodynamic calculation model is constructed based on the hydrodynamic boundary parameter set. The two-dimensional unsteady hydrodynamic calculation model is then solved using time-stepping to generate a hydrodynamic simulation state field. The two-dimensional unsteady hydrodynamic calculation model is constructed based on a two-dimensional shallow water equation set, using an unstructured triangular terrain grid as the spatial basis and assigning a Manning roughness coefficient. Open and internal boundary conditions are set. After discretization using the finite volume method and explicit Runge-Kutta integration, the hydraulic parameters of the grid cells are written into the state field according to the spatiotemporal index. The latest monitoring data is assimilated online. Based on the hydrodynamic simulation state field, the observed residual vector is calculated using measured data taken according to the assimilation period. The roughness and flow coefficient are updated using ensemble Kalman filtering. These are substituted into the model to continue solving and updating the state field. This process is repeated to update the model parameters of the two-dimensional unsteady hydrodynamic calculation model. S3: Map the hydrodynamic simulation state field to a 3D terrain mesh to generate a 3D twin rendering scene, and drive interactive visualization deduction and data rereading based on the 3D twin rendering scene; S4: Calculate flooding index and scheduling efficiency index based on the 3D twin rendering scene, generate a set of scheduling optimization schemes, and generate execution control instructions based on the set of scheduling optimization schemes for user reference and decision-making.

2. The method for dynamic flood digital twin simulation of navigation and hydropower hubs based on the fusion of hydrodynamic models according to claim 1, characterized in that, The hydrodynamic boundary parameter set is constructed based on the spatiotemporal encoded original dataset. The hydrodynamic boundary parameter set is a structured parameter package containing six components. Each component carries a corresponding time index range and spatial index range, as well as data quality labels and scene labels inherited from the spatiotemporal encoded original dataset.

3. The method for dynamic flood digital twin simulation of navigation and hydropower hubs based on the fusion of hydrodynamic models according to claim 2, characterized in that, The process of forming the spatiotemporally encoded raw dataset includes the following steps: A unified spatiotemporal benchmark conversion was performed on the multi-source observation data. The timestamps of all time-series data in the multi-source observation data were converted to Coordinated Universal Time. The plane coordinates of all coordinate data in the multi-source observation data were uniformly converted to the 2000 National Geodetic Coordinate System. The elevation benchmarks of all elevation data in the multi-source observation data were uniformly converted to the 1985 National Elevation Benchmark. Data quality checks and missing data completion processing are performed on the multi-source observation data after unified spatiotemporal benchmark conversion to generate spatiotemporally coded raw datasets.

4. The method for dynamic flood digital twin simulation of navigation and hydropower hubs based on the fusion of hydrodynamic models according to claim 3, characterized in that, The process of generating the hydrodynamic simulation state field includes the following steps: The unstructured triangular terrain grid and roughness distribution map in the hydrodynamic boundary parameter set are read. The unstructured triangular terrain grid is used as the basis for spatial discretization of the computational domain. The Manning roughness coefficient of each grid cell in the roughness distribution map is assigned to the corresponding computational cell to establish the spatial discretization framework of the two-dimensional unsteady hydrodynamic calculation model. The two-dimensional unsteady hydrodynamic calculation model is constructed based on the two-dimensional shallow water equation set. Read the upper boundary inflow process, lower boundary water level process, gate outflow relationship table and unit outflow relationship table from the hydrodynamic boundary parameter set. Set the upper boundary inflow process as the flow boundary condition of the grid cell at the upstream inlet of the two-dimensional unsteady hydrodynamic calculation model, and set the lower boundary water level process as the water level boundary condition of the grid cell at the downstream outlet of the two-dimensional unsteady hydrodynamic calculation model. For the grid cells at the locations of each floodgate and generator unit of the navigation and power hub, set them as internal boundary conditions. The two-dimensional unsteady hydrodynamic calculation model was spatially discretized using the finite volume method. The discretized equations were solved by time step using an explicit Runge-Kutta time integration scheme. The water depth, average lateral velocity, average longitudinal velocity, and water surface elevation of all grid cells obtained at each time step were written into the hydrodynamic simulation state field using time and spatial indices as double keys.

5. The method for dynamic flood digital twin simulation of navigation and hydropower hubs based on the fusion of hydrodynamic models according to claim 4, characterized in that, The steps for updating the model parameters of the two-dimensional unsteady hydrodynamic calculation model are as follows: In the time-step solution of the two-dimensional unsteady hydrodynamic calculation model, the latest water level and flow rate observations are read from the spatiotemporally encoded original dataset at preset assimilation intervals. The latest water level and flow rate observations are compared with the water surface elevation and flow velocity values ​​at the corresponding spatial locations and time steps in the hydrodynamic simulation state field, and the observation residual vector is calculated. Based on the observation residual vector, the model parameters of the two-dimensional unsteady hydrodynamic calculation model are updated using the ensemble Kalman filter method. The updated model parameters are substituted into the two-dimensional unsteady hydrodynamic calculation model to continue the subsequent time step solution. The calculation results of the subsequent time steps are used to overwrite and update the water depth, lateral depth average velocity, longitudinal depth average velocity and water surface elevation values ​​at the corresponding time steps and spatial locations in the hydrodynamic simulation state field.

6. The method for dynamic flood digital twin simulation of navigation and hydropower hubs based on the fusion of hydrodynamic models according to claim 5, characterized in that, The specific method for interactive visualization deduction and data rereading driven by the 3D twin rendering scene is as follows: The hydrodynamic simulation state field is mapped onto a three-dimensional terrain mesh to generate a three-dimensional twin rendering scene. The generated three-dimensional twin rendering scene associates the hydraulic parameters of the state field with the three-dimensional water surface rendering mesh through an index mapping table, dynamically updates the water surface elevation and flow velocity texture, and superimposes the three-dimensional structural model of the navigation and power hub to form a three-dimensional twin rendering scene. Based on a 3D twin rendering scene-driven interactive visualization simulation and data playback, the interactive visualization simulation overlays an engineering operation status layer on the 3D twin rendering scene, supports interactive section selection and equipment location clicking, and real-time playback of quantitative hydraulic data and generation of curves and charts.

7. A digital twin simulation system for dynamic floods in navigation and hydropower hubs integrating hydrodynamic models, used to implement the digital twin simulation method for dynamic floods in navigation and hydropower hubs integrating hydrodynamic models as described in any one of claims 1-6, characterized in that, The system includes: Multi-source data acquisition and hydrodynamic boundary parameter construction module: used to acquire multi-source observation data, form a spatiotemporally encoded raw dataset, and construct a hydrodynamic boundary parameter set based on the spatiotemporally encoded raw dataset; Hydrodynamic Model Construction and Online Assimilation Module: This module is used to construct a two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic boundary parameter set, assimilate the latest monitoring data online to obtain the hydrodynamic simulation state field, and update the model parameters of the two-dimensional unsteady hydrodynamic calculation model based on the hydrodynamic simulation state field. 3D Twin Rendering and Interactive Inference Module: Used to map the hydrodynamic simulation state field onto a 3D terrain mesh to generate a 3D twin rendering scene, and drive interactive visualization inference and data rereading based on the 3D twin rendering scene; The scheduling optimization and instruction issuance module is used to calculate flooding indicators and scheduling efficiency indicators based on the 3D twin rendering scene, generate a set of scheduling optimization schemes, and generate execution control instructions based on the set of scheduling optimization schemes for users to refer to and make decisions.

Images