Flood control object multi-granularity simulation method and system based on scene driving

Abstract

The invention provides a scene-driven flood control object multi-granularity simulation method and system, and the method comprises the following steps: matching a current scene mode of a flood control object according to the real-time data of the flood control object; optimizing parameters of the drainage basin rainfall model, the flood control engineering model and the flood routing model according to the determined contextual model; updating the geographic space scene model according to the real-time data of the flood control object; simulating a spatial-temporal distribution change process of rainfall in a drainage basin where the flood control object is located through a drainage basin rainfall model; performing comprehensive simulation and evaluation on structural parameters and operating characteristics of flood control facilities of the flood control object through the flood control engineering model; the flood routing model takes a geographic space scene model as a basic framework, and simulates slope confluence and river channel confluence processes of flood on a flood control object terrain surface based on calculation results of the drainage basin rainfall model and the flood control engineering model. According to the method, the efficiency and scientificity of flood routing prediction and flood control scheduling are greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of water conservancy projects and flood control and disaster reduction, and particularly relates to a multi-granularity simulation method and system for flood control objects based on scenario driving. Background Technique

[0002] In the core field of water conservancy projects and flood control and disaster reduction, accurately simulating and effectively predicting the occurrence, development, and evolution process of floods has always been a top priority. However, existing flood control simulation technologies still face insurmountable challenges in the following key aspects:

[0003] 1. Limitations in model construction: Lack of refined description of complex terrain and vegetation characteristics

[0004] Most traditional flood control simulation methods rely on simplified mathematical models and empirical formulas, and it is difficult to deeply depict the influence of complex terrain features (such as undulations, valley orientations, and gully distributions) in mountainous and hilly areas on water flow convergence, diversion, acceleration, and blockage; at the same time, different vegetation types (forests, grasslands, farmlands, etc.) also have significant differences in processes such as rainfall interception, soil infiltration change, and surface runoff formation. Existing models are difficult to describe them finely, resulting in obvious deviations in the response of the simulation to changes in rainfall intensity and distribution, greatly reducing the accuracy and reliability of flood prediction.

[0005] 2. Insufficient utilization of data resources: Lack of dynamic fusion and parameter optimization of real-time monitoring information

[0006] Although modern hydrological monitoring systems can obtain rich real-time data such as water levels, flows, and early warning information, many simulation technologies have not established a perfect "data fusion and dynamic adjustment" mechanism. Real-time monitoring information cannot be injected into the model in time for updating, resulting in obvious lags or large deviations between the model results and the actual flood evolution situation, and it is difficult to provide accurate decision-making support for flood control scheduling at critical moments.

[0007] 3. The visualization presentation is not intuitive enough: Lack of high-resolution and multi-dimensional dynamic display

[0008] In the existing technology, the means for displaying the flood inundation range and evolution are relatively single, mostly staying at the level of simple two-dimensional static charts or low-precision three-dimensional models. It is difficult to intuitively display the spatial spread process of floods in the basin, the gradient distribution of water levels or flow velocities, and the interaction between floods and the surrounding environment (urban buildings, transportation networks, etc.), thus affecting the decision-makers' ability to quickly and accurately grasp the flood control situation and formulate scientific strategies.

[0009] 4. Single simulation scale and granularity: Lack of multi-level linkage of macro-meso-micro

[0010] At the macro scale, existing methods are unable to comprehensively and deeply simulate the flood evolution law of the entire basin, making it difficult to provide detailed basis for the site selection of large-scale water conservancy projects or the comprehensive planning of the basin; at the meso scale, the collaborative flood control effect and flood propagation characteristics among various water conservancy facilities (such as reservoirs, dams, sluice gates, pumping stations, etc.) within the region still lack refined simulation; at the micro scale, it is even more difficult to accurately depict the mechanical state, change in flow capacity or operation state of a single water conservancy facility under extreme flood conditions. Existing technologies generally cannot flexibly switch between different scales and granularities, restricting the depth of their application in multi-level flood control decision-making.

[0011] 5. Insufficient scenario-driven ability: Difficult to cope with the dynamic combination of diverse flood control scenarios

[0012] With the continuous intensification of climate change and human activities, the scenarios faced by flood control targets are becoming increasingly complex - including various rainfall patterns such as rainstorms, heavy rains, continuous rainfall, and local heavy rainfall, as well as the comprehensive influence of water conservancy facility regulation measures (such as reservoir flood discharge, sluice gate opening and closing, pumping station start and stop) and external environmental interferences (such as earthquake secondary disasters, ice jams blocking the river channel, etc.). Existing technologies mostly simulate in a static or single way, making it difficult to quickly adjust or deeply analyze according to specific scenario characteristics, resulting in rigidity when facing complex and changeable flood control requirements, and unable to timely and accurately adjust the model strategy and generate targeted flood control countermeasures. Summary of the Invention

[0013] The purpose of the present invention is to solve the deficiencies existing in the above background technology, and provide a multi-granularity simulation method and system for flood control targets based on scenario driving. By comprehensively using technical means such as geospatial scene models, high-precision basin rainfall models, flood control engineering models, and flood evolution models, it realizes refined, multi-scale, and multi-granularity real-time simulation of flood control targets under different scenarios. This method dynamically fuses and optimizes parameters of real-time monitoring data, and is supplemented by multi-scene visual presentation, providing high-precision, visual, and interactive simulation support for flood control decision-making under different complex scenarios, and greatly improving the efficiency and scientificity of flood evolution prediction and flood control scheduling.

[0014] The technical solution adopted by the present invention is: A multi-granularity simulation method for flood control targets based on scenario driving, including the following steps:

[0015] Match the scenario mode in which the flood control target is currently located according to the real-time data of the flood control target;

[0016] Optimize the parameters of the basin rainfall model, flood control engineering model, and flood evolution model according to the determined scenario mode;

[0017] Update the geospatial scene model according to the real-time data of the flood control target;

[0018] Based on the real-time data of the flood control object, simulate the spatio-temporal distribution change process of rainfall in the basin where the flood control object is located through the basin rainfall model;

[0019] Based on the real-time data of the flood control object, comprehensively simulate and evaluate the structural parameters and operation characteristics of the flood control facilities of the flood control object through the flood control project model, so as to reflect the regulation effect of the flood control facilities on the flood evolution and the overall flood control effect;

[0020] The flood evolution model takes the geospatial scene model as the basic framework, and simulates the overland flow and channel flow processes of the flood on the terrain surface of the flood control object based on the calculation results of the basin rainfall model and the flood control project model;

[0021] Visualize the specific calculation results of the basin rainfall model, the flood control project model and the flood evolution model in the current scenario mode based on the geospatial scene model.

[0022] In the above technical solution, the process of matching the scenario mode includes:

[0023] Assemble the real-time data into several scenario vectors according to the set feature dimensions; the scenario vectors are used to represent the key feature quantities of a certain scenario;

[0024] Score each scenario vector based on the threshold interval where each feature quantity in each scenario vector is located;

[0025] Select the scenario mode corresponding to the scenario vector with the highest score as the scenario mode in which the flood control object is currently located.

[0026] In the above technical solution, each scenario mode corresponds to a visualization template. Once the scenario mode is switched, the corresponding visualization template is automatically scheduled; the visualization feature quantities selected by the external instructions in each scenario mode are recorded in real time. When the selection frequency of any feature quantity in any scenario mode reaches the set threshold, the feature quantity is added to the visualization template corresponding to the scenario mode.

[0027] In the above technical solution, the basin rainfall model, the flood control project model and the flood evolution model are all set with initial model parameters; based on the mapping model, determine the parameters to be adjusted and their adjustment coefficients for the basin rainfall model, the flood control project model and the flood evolution model in the current scenario mode, and perform corresponding optimization based on the initial model parameters; the mapping model is used to reflect the mapping relationship between the scenario mode and the model parameter adjustment method.

[0028] In the above technical solution, the construction process of the mapping model includes:

[0029] Form historical scenario vectors based on historical data;

[0030] For each historical scenario vector, obtain the optimal model parameter adjustment amounts and corresponding adjustment coefficients for the basin rainfall model, flood control project model, and flood routing model after measured or simulated calibration;

[0031] Integrate the historical scenario vector with the corresponding parameter adjustment coefficients into training samples to form a mapping training set;

[0032] Train a machine learning or regression model using the mapping training set, and use the trained model as the mapping model.

[0033] In the above technical solution, the process of setting the initial model parameters of the basin rainfall model, flood control project model, and flood routing model includes:

[0034] Set initial values for the model parameters of the basin rainfall model, flood control project model, and flood routing model;

[0035] Calibrate the above models through historical flood events or a series of typical working conditions to make the model output consistent with historical observations;

[0036] Select historical flood events or data from different years that were not involved in calibration to test the simulation accuracy of the model under this event;

[0037] Through several calibration-validation iterations, gradually correct and optimize the model parameters;

[0038] Select the parameters that can optimize the error indicators of each model as the final initial model parameters.

[0039] In the above technical solution, the process of constructing the geospatial scenario model includes:

[0040] Import digital elevation model data into the GIS platform, and through terrain rendering technology, display the terrain and landform of the basin where the flood control object is located in three dimensions, set the visualization effect of the enhanced terrain to make it highly consistent with the actual terrain features;

[0041] Represent the land use type data of the flood control object in the model with polygon elements of different colors and textures;

[0042] Present the river system data of the flood control object with vector line elements having width and depth attributes, draw according to the actual shape of the river channel, and simulate the flowing state of the river by setting a water flow texture animation.

[0043] In the above technical solution, the process of constructing the basin rainfall model includes:

[0044] Use the first hydrological model to calculate the topographic index of each grid according to the topographic data in the geospatial scenario model, identify the distribution of soil saturation areas and runoff generation conditions; update the spatial distribution of soil moisture at the initial stage of the simulation or at each time period to determine potential rapid runoff areas;

[0045] Use the second hydrological model based on the water balance and energy balance equations to simulate the processes of atmospheric rainfall, vegetation interception and evapotranspiration, soil infiltration and water storage, surface runoff, and groundwater runoff; set corresponding parameters according to different vegetation types and soil layers in the basin;

[0046] Map the saturated area area and initial soil moisture information output by the first hydrological model to the grid cells of VIC, thereby affecting the runoff generation and concentration processes of each grid in the second hydrological model;

[0047] Ensure that the first hydrological model and the second hydrological model run at the same or compatible time steps, and regularly exchange key variables.

[0048] In the above technical solution, the construction process of the flood routing model includes:

[0049] Based on the geospatial scenario model as the basic framework, use the rainfall data generated by the basin rainfall model as the water source input for the flood;

[0050] In the simulation of overland flow concentration, use the overland flow model based on the kinematic wave theory, consider the influence of terrain slope, vegetation cover, and soil infiltration on the formation and velocity of overland runoff, and calculate the flow rate and flow direction of overland runoff;

[0051] In the simulation of river channel flow concentration, use the river channel hydrodynamic model based on the Saint-Venant equations, and combine the regulation effects of each flood control facility in the flood control project model to simulate the flood routing process in the river channel.

[0052] The present invention also provides a multi-granularity simulation system for flood control objects based on scenario-driven, including:

[0053] A scenario mode matching module for matching the scenario mode in which the flood control object is currently located according to the real-time data of the flood control object;

[0054] A model parameter optimization module for optimizing the parameters of the basin rainfall model, flood control project model, and flood routing model according to the determined scenario mode;

[0055] A geospatial scenario update module for updating the geospatial scenario model according to the real-time data of the flood control object;

[0056] A rainfall simulation module for simulating the spatio-temporal distribution change process of rainfall in the basin where the flood control object is located based on the real-time data of the flood control object through the basin rainfall model;

[0057] A flood control project simulation module, which is used to comprehensively simulate and evaluate the structural parameters and operating characteristics of flood control facilities of a flood control object based on real-time data of the flood control object through a flood control project model, so as to reflect the regulation effect of flood control facilities on flood evolution and the overall flood control effect;

[0058] A flood evolution model, which is used to simulate the overland flow and channel flow processes of floods on the terrain surface of a flood control object based on a geographical space scenario model as the basic framework and the calculation results of a basin rainfall model and a flood control project model through the flood evolution model;

[0059] A visualization module, which is used to visualize the specific calculation results of the basin rainfall model, the flood control project model, and the flood evolution model in the current scenario mode based on the geographical space scenario model.

[0060] The beneficial effects of the present invention are as follows: Through steps such as scenario recognition, model parameter optimization, geographical scenario update, rainfall and flood control project simulation, flood evolution and visualization, the present invention forms an organic whole; it can automatically switch parameters and simulation strategies for different flood control scenarios (such as different rainfall patterns, water conservancy facility operating conditions, etc.); it collaborates at multiple levels of macro (the entire basin), meso (facility collaboration within a regional scope), and micro (operation of individual facilities) to achieve more delicate flood prediction and scheduling support; after obtaining the real-time data of the flood control object, it can quickly match the scenario and update the simulation results, reducing the deviation from the actual flood evolution; based on the geographical space scenario, the visualization of the simulation results is more intuitive, facilitating decision-makers to quickly grasp the overall situation and formulate reasonable strategies.

[0061] Furthermore, the present invention assembles real-time data into a scenario vector and scores it, which can quantitatively evaluate the matching degree of different scenarios; selects the scenario with the highest score to avoid conflicts caused by the coexistence of multiple scenarios, making the system more focused and clear in decision-making; based on thresholds or other judgment criteria, it can more flexibly distinguish various scenarios such as heavy rainstorms, heavy rains, and special scheduling conditions; when a new scenario appears, new feature dimensions and threshold intervals can be introduced to enhance the model's recognition ability for new or extreme scenarios.

[0062] Furthermore, when the scenario mode of the present invention switches, it automatically schedules the corresponding visualization template to quickly present the data and graphics most relevant to the current scenario; the record of the selection frequency of feature quantities by external instructions realizes the automatic improvement and expansion of the visualization template, gradually enhancing the practicality and pertinence of visualization; different scenarios have different focuses (such as rainfall distribution is concerned in case of heavy rainfall, and flow rate or gate status is concerned in flood control facility regulation, etc.), and the switching of visualization templates makes the information more focused on the current decision-making needs; there is no need to manually set the display content for each scenario, and the system can adjust the template configuration according to the usage frequency and instructions.

[0063] Furthermore, the present invention directly outputs “which parameters and adjustment coefficients need to be adjusted under the current scenario mode” through a mapping model, thereby reducing the blindness of manual parameter adjustment; after scenario determination, the parameters of the basin rainfall model, flood control project model and flood evolution model can be corrected in a timely manner to improve the real-time adaptability of the simulation; taking the initial model parameters as a benchmark and then superimposing appropriate adjustments is conducive to maintaining the stability of the model while taking into account the accuracy under different scenarios; the mapping model can train more parameters (such as gate opening rules for specific projects, pump station start and stop thresholds, etc.) to enhance the system's adaptability to changing flood control needs.

[0064] Furthermore, the present invention is based on historical data and historical scenario vectors, and can mine the optimal parameter adjustment amount in real scenarios with high credibility; a variety of algorithms (regression, neural network, etc.) can be selected for mapping training, and various types of data can be fully utilized to establish a training set of the "scenario vector-parameter adjustment" correspondence relationship, which can be quickly inferred during online operation after the training is completed; if the historical data is relatively abundant, the trained mapping model can also have better prediction and inference capabilities for new scenarios, thereby further improving the simulation accuracy.

[0065] Furthermore, the initial parameter settings of the basin rainfall model, flood control engineering model and flood evolution model of the present invention are calibrated and verified through historical flood events or typical working conditions to ensure that the initial parameters of the model have high accuracy; the optimal parameters are obtained after multiple calibrations and verifications, which can significantly reduce the simulation error and enhance the stable performance of the model in different time periods and scenarios; in the process of selecting independent data for verification, the deficiencies of the model in specific types of floods or regions can be discovered and repaired in a timely manner; the calibration process can be applicable to scenarios of different basins or different combinations of flood control facilities, as long as there is corresponding observation data for optimization and verification.

[0066] Furthermore, the present invention can intuitively restore the topography by integrating DEM with river systems, land use data, etc. into the GIS platform; different land use types, river width / depth attributes and other elements can be presented in a real or near-real form, which is convenient for subsequent model superposition; the setting of river water texture animation enhances the dynamic display effect, allowing users to intuitively understand the direction and speed of water flow and its interaction with the surrounding environment; it provides a spatial positioning benchmark for basin rainfall models, flood control engineering models and flood evolution models, greatly improving the accuracy and comprehensibility of data processing and result expression.

[0067] Furthermore, the present invention utilizes the topographic index of TOPMODEL to identify rapid runoff generation areas, and combines the VIC model to comprehensively consider the hydrological processes of vegetation, soil, and atmosphere, achieving a more refined simulation of rainfall-runoff: Regularly exchanging key variables (such as the area of soil saturation zone, soil water content) can reflect the actual rainfall and the dynamics of surface / groundwater, improving the response speed of the model to rainfall processes; Parameters can be set separately to more accurately capture the impacts of forests, grasslands, farmlands, etc. on rainfall interception and evapotranspiration; It supports running at the same or compatible time step as TOPMODEL and VIC, facilitating flexible adjustment of simulation accuracy according to rainfall characteristics and computing resources.

[0068] Furthermore, the present invention uses the kinematic wave theory for slope flow calculation, considering the impacts of terrain slope, vegetation cover, etc. on runoff formation and velocity; The Saint-Venant equations are used for high-precision hydrodynamic simulation of river confluence; During the river confluence process, flood control engineering models (such as dams, sluices, pumping stations, reservoirs, etc.) are introduced to regulate the flood process, enhancing the closeness of the simulation to real scenarios: Using the output of the basin rainfall model as the water source input can effectively reflect the impacts of rainfall spatio-temporal variations on flood evolution; Supported by three-dimensional terrain and river data, it is easier to visualize the evolution process and inundation range of floods in different terrain units, assisting in precise flood control decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 is a schematic flow chart of this embodiment;

[0070] Figure 2 is a schematic diagram of the application principle of this embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0071] The present invention will be further described in detail below in conjunction with the drawings and specific embodiments, which is convenient for clearly understanding the present invention, but they do not limit the present invention.

[0072] Embodiment 1

[0073] As Figure 1 shown, the present invention provides a multi-granularity simulation method for flood control objects based on scenario driving, including the following steps:

[0074] Match the current scenario mode of the flood control object according to the real-time data of the flood control object;

[0075] Optimize the parameters of the basin rainfall model, flood control engineering model, and flood evolution model according to the determined scenario mode;

[0076] Update the geospatial scene model according to the real-time data of the flood control object;

[0077] Based on the real-time data of the flood control object, simulate the spatio-temporal distribution change process of rainfall in the basin where the flood control object is located through the basin rainfall model;

[0078] Based on the real-time data of the flood control object, comprehensively simulate and evaluate the structural parameters and operation characteristics of the flood control facilities of the flood control object through the flood control project model, so as to reflect the regulation effect of the flood control facilities on the flood evolution and the overall flood control effect;

[0079] The flood evolution model is based on the geospatial scene model. Based on the calculation results of the basin rainfall model and the flood control project model, simulate the slope runoff and channel runoff processes of the flood on the terrain surface of the flood control object;

[0080] Based on the geospatial scene model, visualize the specific calculation results of the basin rainfall model, the flood control project model and the flood evolution model in the current scenario mode.

[0081] Before executing the above simulation method, it is first necessary to set the scenario mode based on the historical data of the flood control object, and establish a geospatial scene model, a basin rainfall model, a flood control project model and a flood evolution model.

[0082] The following further explains the process and principle of setting the scenario mode for a specific flood control object.

[0083] First of all, the flood control object can be a specific basin area (such as mountainous and hilly areas, plain areas, urban built-up areas, etc.), or a combination of multiple adjacent water conservancy project facilities (such as reservoirs, sluice dams, pumping stations, flood diversion channels, etc.).

[0084] When this "object" faces various scenarios in actual flood control (such as different rainfall patterns, sudden water level rise, change in the warning information level, external secondary disasters, etc.), the model needs to accurately identify and "switch" the corresponding simulation logic.

[0085] The scenario mode refers to abstracting several representative flood control states or working conditions from the massive real-time data and historical experience of the flood control object.

[0086] Typical scenario modes may include:

[0087] Rainfall scenarios (such as local heavy rain, widespread moderate rain, typhoon heavy rainfall, etc.);

[0088] Water level scenarios (such as rapid water level rise, water level peak crossing, overbank risk, etc.);

[0089] Warning scenarios (such as level 1-4 warnings, water level or flow mutation rate exceeding the threshold, etc.);

[0090] External interference scenarios (secondary floods caused by earthquakes, debris flows, ice jams, etc.).

[0091] The purpose of scenario mode setting is to enable the system to automatically determine which scenario mode the object is most likely in currently based on the real-time data of flood control objects and the predefined scenario feature dimensions, and perform subsequent operations such as model parameter optimization and simulation visualization switching based on this.

[0092] Establish scenario modes, set corresponding scenario vectors, and the characteristic quantities that the scenario vectors need to include for the flood control risks or key concerns that are likely to occur for this specific flood control object; for example:

[0093] For the rainfall scenario mode, construct a multi-dimensional rainfall scenario vector based on characteristics such as rainfall amount, rainfall intensity, rainfall duration, and rainfall spatial distribution. Let the rainfall amount be P (unit: mm), the rainfall intensity be I (unit: mm / hour), the rainfall duration be Y (unit: hour), and the rainfall spatial distribution characteristics can be represented by a specific spatial distribution index S. Then the rainfall scenario vector can be expressed as

[0094] For the water level scenario mode, construct a water level scenario vector with the water level rising speed V (unit: m / hour), water level peak H (unit: m), etc.

[0095] For the hydrological station warning scenario mode, construct a vector based on the warning level L (for example, levels 1 - 4 correspond to different degrees of danger) and the warning information change rate C (such as the change rate of water level or flow rate).

[0096] For the facility operation scenario mode, the corresponding scenario vector can include the following characteristics: gate opening, pump station output rate, flood diversion channel flow rate, initial water level, flood discharge flow rate, etc.

[0097] Set several intervals for each characteristic quantity of each scenario vector. The number of intervals and the upper and lower limits can be determined comprehensively based on historical data and expert experience.

[0098] For example, for the rainfall scenario mode, the intervals can be divided on the rainfall amount (mm) as [0 - 20) (light rain), [20 - 50) (moderate rain), [50 - 100) (heavy rain), [100, ∞) (rainstorm), and different grade scores are assigned when scoring.

[0099] The historical data includes:

[0100] Meteorological monitoring data: rainfall amount, rainfall intensity, rainfall spatial distribution, temperature, wind speed, etc.;

[0101] Hydrological monitoring data: water level, flow rate, flow velocity, warning information;

[0102] Flood control facility operation data: reservoir flood discharge gate opening, pump station startup status, dam seepage monitoring, etc.;

[0103] External event information: geological disaster notifications (earthquakes, landslides), abnormal urban drainage systems, ice jam reports, etc.

[0104] When new special situations are indicated by historical data or expert experience (such as new extreme rainfall patterns, major engineering renovations, changes in urban underlying surfaces, etc.), new feature quantities can be added or thresholds modified in the system, and corresponding scenario names can be defined.

[0105] During subsequent operation, once the feature quantities of real-time data exceed the old range, the system can identify and match the new scenario pattern.

[0106] In this embodiment, diverse scenario patterns can be set according to actual flood control requirements and possible risk situations. For example, considering various combinations of different rainfall patterns (such as changes in the position of the rainstorm center, increasing or decreasing rainfall intensity, different rainfall durations, etc.) and flood control project regulation schemes (such as different combinations of reservoir flood discharge flows, different opening degrees and opening time sequences of sluice gates, different start-stop strategies of pumping stations, etc.); at the same time, the influence of external environmental factors can also be incorporated (such as sudden flood discharge caused by damage to the reservoir dam due to an earthquake, flood caused by the sudden thawing of an ice jam blocking a river channel, flood and tide superposition in coastal areas caused by typhoon rainstorms, etc.) to construct a complex and variable flood control scenario pattern library. In addition, for different flood warning levels (such as blue warning, yellow warning, orange warning, red warning), corresponding scenario parameters such as the scope of personnel evacuation, material allocation plan, and operation priority of flood control project facilities are set to provide comprehensive simulation support for flood control decision-making at different stages.

[0107] The process and principle of establishing a geospatial scenario model, a basin rainfall model, a flood control project model, and a flood evolution model for specific flood control objects will be further described below.

[0108] The first step is data collection:

[0109] 1. Collection of geographical data:

[0110] ① Use high-resolution satellite remote sensing technology (such as satellite images of WorldView-3, GeoEye-1, etc.) to obtain the topographic data of the basin, generate a digital elevation model (DEM) with a precision of up to 0.5 meters or even higher, clearly and accurately reflecting the micro-topographic features such as the terrain undulation, mountain ranges, valleys, and gullies within the basin; at the same time, collect land use type data to accurately identify the vegetation cover (forests, grasslands, farmlands, etc.), urban construction (building distribution, road networks, etc.), water areas (rivers, lakes, reservoirs, etc.) in different regions; as well as river system distribution data to determine in detail information such as the source, tributaries, river width, depth, and curvature of the river.

[0111] ②For key areas or complex terrain areas, use unmanned aerial vehicle (UAV) aerial surveying and mapping technology for supplementary collection to obtain more detailed topographic and ground feature information, such as local topographic changes around flood control engineering facilities like dams and sluice gates, and the distribution of obstacles in the river channel, etc., to further improve the accuracy and integrity of geographical data.

[0112] ③Field measure the geographical coordinates and elevation data of some key control points, such as the locations of hydrological stations and key parts of large-scale water conservancy project facilities, to calibrate and verify the data collected by satellite remote sensing and UAVs, and ensure the absolute accuracy of geographical data.

[0113] 2. Meteorological data collection:

[0114] ①Establish a comprehensive data sharing and cooperation mechanism with national meteorological departments, local meteorological observation stations, and professional meteorological data service providers to obtain meteorological data with long time series (at least over 30 years), high temporal resolution (such as every 10 minutes or even shorter time intervals), and spatial resolution (covering the entire basin area, with an accuracy of up to 1 square kilometer or higher), including information such as rainfall, rainfall intensity, rainfall duration, rainfall type (such as heavy rain, moderate rain, light rain, shower, thunderstorm, etc.), temperature, humidity, wind speed, wind direction, air pressure, etc.

[0115] ②Install small meteorological observation stations at key locations in the basin (such as mountainous rainfall concentration areas, urban flood-prone areas, etc.) to supplement the collection of meteorological data in local areas, and focus on monitoring local meteorological anomalies (such as local heavy rainfall, small-scale convective weather, etc.), providing data support for accurately simulating the impact of local rainfall on flood formation.

[0116] 3. Flood control project data collection:

[0117] ①Organize a professional water conservancy project exploration team to conduct a detailed exploration and data collection of various flood control engineering facilities within the basin. For dams, record their detailed geometric dimensions (length, height, crest width, thickness of each part of the dam body, etc.), building materials (earth dams, concrete dams, earth-rock mixed dams, etc.), slopes (inner slope and outer slope gradients), permeability coefficients (permeation characteristics of different dam body materials), foundation geological conditions (stratum structure, soil type, groundwater level, etc.) and other parameters; for sluices, collect information such as their model specifications, number and size of sluice gates, discharge coefficients, opening and closing methods (electric, manual, hydraulic, etc.), gate material and structural characteristics, water stop performance parameters, river channel morphology and dimensions of the upstream and downstream connection sections of the sluice; for pumping stations, obtain their installed capacity, head, flow rate, operating efficiency, motor characteristics, pipe diameter and length, etc.; for reservoirs, record in detail their storage capacity curves (correspondence between water level and storage capacity), types of flood discharge facilities (spillways, flood discharge tunnels, etc.) and parameters (dimensions, flow characteristics, gate control methods, etc.), dam structure parameters (dam height, dam length, dam type, dam body material, etc.), historical operation records of the reservoir (annual water level changes, water storage changes, flood discharge event records, etc.); for flood diversion channels, measure their width, depth, length, roughness coefficient, flood carrying capacity (flow rate passing through at different water levels), distribution of obstacles in the river channel, etc.

[0118] ②Establish a flood control engineering database to classify, organize, store and manage the collected data, facilitate data query, retrieval and update, and ensure the integrity and timeliness of the data.

[0119] 4. Hydrological station data collection:

[0120] ①Install high-precision and highly reliable water level and flow monitoring equipment at key hydrological stations within the basin, such as ultrasonic water level gauges (accuracy up to 0.01 m), radar water level gauges (accuracy up to 0.02 m), acoustic Doppler current profilers (ADCP, flow measurement accuracy within 5%), etc., to collect real-time data such as water level, flow velocity and flow rate. The data collection frequency can be set to high frequency according to flood warning requirements, such as once every 5 minutes or even shorter, and the data is transmitted to the data processing center in a timely and stable manner through wired or wireless communication technologies (such as 4G / 5G networks, fiber optic communication, etc.).

[0121] ②Regularly calibrate, maintain and overhaul the monitoring equipment at hydrological stations to ensure the accuracy and continuity of data collection. At the same time, establish a data quality monitoring and auditing mechanism to conduct real-time quality inspections on the collected data, eliminate abnormal data, and ensure the reliability of the data.

[0122] The second step, data preprocessing:

[0123] 1. Data cleaning:

[0124] ①Conduct a comprehensive and in-depth inspection and elimination of error data for various types of collected data. For geographical data, check and remove significantly abnormal terrain elevation values (such as isolated points with a large difference in terrain elevation from the surrounding areas), incorrect land use type markings (such as marking water areas as land), unreasonable river water system connection relationships, etc.; for meteorological data, eliminate data such as negative rainfall amounts that do not conform to physical laws, air temperatures outside the reasonable range (such as below absolute zero or much higher than the local historical extreme values), and abnormally large wind speeds; for flood control project data, correct incorrect engineering parameters (such as incorrect dimension markings, unreasonable flow coefficients, etc.); for hydrological station data, identify and remove abnormal water level and flow data (such as sudden water level changes, flow jumps, etc.) caused by equipment failures, electromagnetic interference, etc.

[0125] ②Use data interpolation algorithms to supplement a small amount of missing data. For missing points in geographical data, estimate and supplement them using Kriging interpolation method, inverse distance weighted interpolation method, etc. based on the elevation values of surrounding known points, land use type distribution rules, etc.; for missing meteorological data, conduct comprehensive interpolation and supplementation by combining historical data of the same period, data from surrounding meteorological stations, and meteorological model prediction values; for missing flood control project data, supplement it by consulting engineering design documents, historical exploration reports, or conducting on-site re-measurements; for missing hydrological station data, conduct interpolation and supplementation based on the correlation between upstream and downstream hydrological station data and the water level-flow relationship curve to ensure the continuity and integrity of the data.

[0126] 2. Data format conversion:

[0127] ①Use professional data conversion software and tools to uniformly convert data from different sources and in different formats into a format suitable for subsequent processing and analysis. For example, convert satellite remote sensing image data from the original image format (such as TIFF, JPEG, etc.) to a raster data format (such as GRID format) recognizable by a geographic information system (GIS), and establish a geographic coordinate reference system; convert meteorological data provided by the meteorological department from text formats (such as CSV, XML, etc.) to a structured data format (such as a table form in a relational database) that can be stored and queried in a database, facilitating data management and retrieval; organize flood control project data into a standardized table format (such as an Excel table or a database table), and standardize the naming of data fields to facilitate data retrieval and processing; convert real-time data collected by hydrological stations into a specific data format (such as JSON format) to maintain data consistency and compatibility during data transmission and storage.

[0128] The third step is to establish a geospatial scene model.

[0129] 1. Prepare and clean the following data:

[0130] High-precision terrain data (DEM): Using a DEM with a resolution better than 5m as the basis, the undulating characteristics of mountainous and hilly areas and the terrain differences in plain areas are completely retained; preprocessing such as noise reduction, stitching, and coordinate unification is performed on the collected DEM data, and abnormal high values or depressions are removed.

[0131] Land use data: Obtaining vector or raster data classified as urban, farmland, forest, water area, etc.; if there are multi-source differences, fusion and reclassification processing are performed.

[0132] River system data: Including the width, depth, and river section division of the main river channel and its tributaries; recorded in the form of vector line elements, and measured cross-section data of the river channel is supplemented when necessary.

[0133] 2. Importing the GIS platform and 3D visualization:

[0134] DEM data rendering: Importing the DEM into ArcGIS, QGIS or other 3D GIS software, setting lighting, shadows, and texture mapping to realistically represent the terrain undulations and make it highly consistent with the actual terrain features.

[0135] Land use polygons: Assigning different colors and textures to different land use types (such as light green for farmland, dark green for forest, gray for urban areas, etc.) to clearly distinguish areas such as urban, farmland, forest, and water area, so as to intuitively understand the distribution pattern of different land use types in the basin and their potential impact on flood evolution.

[0136] Dynamic representation of the river system: Matching the width and depth of the river vector line elements along the river cross-section in a 3D environment; simulating the water flow movement by setting water flow texture animations to lay the foundation for subsequent flood evolution visualization.

[0137] 3. Data integration and metadata storage:

[0138] Establishing a unified geographical database (such as Geodatabase) to store DEM, land use, river channel information, as well as the corresponding coordinate system and metadata information.

[0139] Assuming that the cross-section of a certain urban section of the river uses measured river cross-sections, digitize its cross-section curve and perform precise calibration in the GIS with reference to the DEM to ensure that the river channel position coincides with the elevation model.

[0140] Step 4: Construct a basin rainfall model.

[0141] 1. Prepare the following data:

[0142] Meteorological data: Historical and real-time rainfall, rainfall intensity, raindrop size distribution, air temperature, wind speed, etc., collected from multiple channels such as meteorological observatories, automatic weather stations, and radar remote sensing.

[0143] Vegetation data: vegetation coverage, vegetation type (such as coniferous forest, broad-leaved forest, grassland, farmland, etc.), leaf area index (LAI), etc.

[0144] Topographic data (from the aforementioned geospatial scene model): derivative factors such as slope and aspect, used to identify the impact of topography on runoff generation.

[0145] 2. In this embodiment, the first hydrological model uses TOPMODEL, and the second hydrological model uses VIC; coupling TOPMODEL and VIC:

[0146] TOPMODEL identifies the distribution of saturated areas and initial runoff conditions through the topographic index:

[0147] Calculate the catchment area and slope of each grid cell from the DEM to form a topographic index distribution map; thereby determining the spatial distribution pattern of soil moisture and the initial runoff generation conditions. Judge which areas are more likely to be saturated, determine the initial soil moisture; output the saturated area and initial runoff to the subsequent links.

[0148] The VIC model comprehensively considers the entire hydrological cycle such as rainfall, vegetation interception, evapotranspiration, and soil layer infiltration:

[0149] Initialize the soil moisture output by TOPMODEL into the VIC grid; input meteorological data (rainfall, rainfall intensity, temperature), combined with vegetation parameters (such as leaf area index), and simulate surface runoff, evapotranspiration, and multi-layer soil moisture changes under the water balance and energy balance equations; generate gridded surface runoff, deep infiltration, soil moisture, etc. data for each time period (such as 1h or 6h).

[0150] 3. Spatial interpolation and accuracy improvement:

[0151] Optimize the spatial interpolation algorithm for rainfall, adopt the co-kriging interpolation method based on topographic and meteorological factors, establish a variogram by correlating rainfall with topographic factors (such as altitude, slope, wind direction), and improve the capture accuracy of local heavy rain in mountainous and hilly areas. After evaluating the effects in the test area and comparing with other interpolation methods (such as IDW, ordinary kriging), select the optimal parameter set to improve the spatial resolution and accuracy of rainfall data under complex terrain conditions, so as to better reflect the unevenness of local rainfall and its impact on flood formation.

[0152] If a heavy rain gathers in a valley and causes local floods, co-kriging interpolation can accurately reflect the rainfall center in the valley area and avoid ignoring the gathering effect of the narrow valley terrain on the rain cloud cluster.

[0153] Step 5, construct a flood control engineering model.

[0154] For facilities such as dams, sluices, pumping stations, reservoirs, flood diversion channels, etc. within the basin, establish mathematical models respectively and integrate them into a unified flood control engineering model system.

[0155] 1. Establish a dam model (finite element mechanical analysis):

[0156] Discretize the geometric shape and material properties of the dam into finite element meshes, and construct a mechanical model using the finite element analysis method; consider the hydrostatic pressure, hydrodynamic impact force, scouring force of the flood on the dam body, as well as the self-weight and friction of the dam, and simulate the stability and seepage process of the dam through the mechanical equilibrium equation.

[0157] 2. Establish a sluice model (flow regulation):

[0158] According to the gate type, orifice size, gate opening, and water level difference between upstream and downstream, etc., calculate the flow through the sluice using a flow regulation model of hydraulics formulas (such as weir flow or orifice flow); combine variables such as gate opening and water level difference between upstream and downstream to accurately calculate the flow rate through the sluice and the water level regulation effect under different working conditions.

[0159] 3. Establish a pumping station model (pumping and drainage):

[0160] Based on the installed capacity, head, flow rate characteristic curve, efficiency curve, and operating parameters of the motor and pump, construct a pumping capacity function of the pumping station; set the start / stop rules and factors such as the change of operating efficiency of the pumping station at different water levels and different time periods, dynamically calculate the drainage volume of the pumping station, and accurately simulate the pumping volume and drainage effect of the pumping station during the flood control process.

[0161] 4. Establish a reservoir model (water balance and operation):

[0162] Taking the reservoir storage - water level curve as the core, consider the inflow (including rainfall runoff, upstream inflow, etc.), outflow (such as flood discharge flow, power generation water flow, irrigation water flow, etc.), reservoir evaporation, seepage, etc. to construct a water balance model; combine operation strategies (such as pre - flood drawdown, flood storage during the flood season, cascade linkage, etc.), update the reservoir water level and storage data in real - time, and calculate the flood discharge / generation / irrigation outflow volume to simulate the flood storage and flood discharge processes of the reservoir under different flood control scenarios and its regulatory effect on the downstream flood evolution.

[0163] 5. Establish a flood diversion channel model (based on the Saint - Venant equations):

[0164] Parameterize the morphology, roughness coefficient, and flood - passing capacity of the flood diversion channel, and use one - dimensional or two - dimensional hydrodynamic models to simulate the water flow evolution in the flood diversion channel, including water level changes, velocity distribution, and water flow exchange with the surrounding area, etc. Analyze the reduction effect of the flood diversion channel on the downstream flood peak under normal flood discharge and emergency flood diversion and other working conditions.

[0165] If a sluice gate has an automated monitoring system that records the gate opening and upstream and downstream water level difference in real time, the data can be dynamically bound to the model to more realistically reflect the impact of sluice gate regulation on the flood process.

[0166] Step 6: Construction and overall coupling of flood evolution model:

[0167] Taking the geospatial scene model as the spatial benchmark;

[0168] The basin rainfall model is used to provide external water source input for runoff generation and confluence;

[0169] Use flood control engineering models to provide operating boundaries and rules for dams, sluice gates, pumping stations, reservoirs, and flood diversion channels;

[0170] A hydrodynamic model combining slope confluence and river confluence is used to simulate the flood evolution of the entire basin.

[0171] Among them, the slope confluence simulation (kinematic wave theory) maps the surface runoff output from the VIC model (or TOPMODEL) to the geographic space grid; in the three-dimensional terrain scene, the slope flow model based on the kinematic wave theory is used to consider the influence of factors such as terrain slope, vegetation cover, soil infiltration on the formation and velocity of slope runoff, and calculate the slope flow velocity and flow direction according to the kinematic wave equation to obtain the flow convergence process on the slope to the river channel.

[0172] River confluence simulation (Saint-Venant equations) uses one-dimensional or two-dimensional Saint-Venant equations to discretely solve the main river and tributaries; at the same time, the regulation effect of flood control projects is superimposed:

[0173] Dams: block floods and generate seepage,

[0174] Water gate: flow regulation changes,

[0175] Pumping station: pumps low-lying surface water into or out of the river.

[0176] Reservoir: flood storage and flood discharge,

[0177] Flood diversion channel: divert part of the flood water to other areas.

[0178] Dynamically output spatiotemporal series data such as water level fluctuation, flow velocity distribution, and flooding range.

[0179] The above process can accurately simulate the evolution of floods in the river channel, including the rise and fall of water levels, changes in flow velocity, expansion of the inundation range, and the interaction between floods and the surrounding environment (such as urban areas, farmland, etc.).

[0180] Furthermore, depending on the region size and simulation accuracy, the finite difference method (FDM) or the finite volume method (FVM) can be selected:

[0181] Initial conditions: initial water level, initial flow rate, soil water content, etc.;

[0182] Boundary conditions: water level or flow control at the basin outlet, operation rules of flood control engineering facilities, etc.

[0183] Assume that for a specific rainstorm process, the time-varying inflow process curve generated by the basin rainfall model is fed into the flood routing model; if the reservoir pre-releases water before the peak, the dynamic change of the reservoir storage capacity and the reduction amount of the downstream flood peak can be simulated when the flood peak arrives.

[0184] Step 7: Train the constructed basin rainfall model, flood control engineering model and flood routing model to determine their initial parameters.

[0185] The initial parameters of the basin rainfall model include:

[0186] Soil permeability coefficient, soil water content: typical values can be referred to from soil type data, field tests or literature queries;

[0187] Vegetation evapotranspiration coefficient: can be based on the vegetation types in the basin or published research;

[0188] Rainfall spatial distribution parameters (such as variogram parameters): spatial interpolation variogram fitting can be done by combining historical observations of rain gauges / radars.

[0189] The initial parameters of the flood control engineering model include:

[0190] Dike model: geometric parameters (dike height, slope, material properties) usually come from engineering design drawings or measured data;

[0191] Sluice model: sluice opening, gate height, discharge coefficient, etc., according to the design manual or existing operation records;

[0192] Pump station model: pump station head, power, efficiency curve, obtained from equipment manufacturers or actual operation records;

[0193] Reservoir model: reservoir storage capacity curve, discharge coefficient of the reservoir outlet, evaporation loss rate, etc., can be obtained from historical operation data or design data.

[0194] The initial parameters of the flood routing model include:

[0195] River channel geometric parameters: such as cross-section data, Manning roughness coefficient, river channel longitudinal slope, etc., usually obtained from measured river channel cross-sections and historical hydrological analysis;

[0196] Diversion channel parameters: such as roughness coefficient, upper limit of diversion flow rate, river channel geometric shape;

[0197] Numerical solution grid: If the finite difference or finite volume method is adopted, the grid scale needs to be preset according to the river channel length and basin area.

[0198] Specifically, after the above initial parameters are set based on literature references, expert experience or publicly available engineering design data, the model needs to be calibrated and verified through historical flood events or a series of typical working conditions, so as to gradually approach the real situation and make the initial parameters more in line with the actual situation:

[0199] 1. Group the historical data collected and preprocessed in the first step:

[0200] Group according to time or flood events (such as selecting 5 - 10 typical flood events), and further divide them into training set, validation set and test set:

[0201] Training set: Used to adjust model parameters;

[0202] Validation set: Check the model performance in real time during training to avoid overfitting;

[0203] Test set: Independently check the generalization ability of the final model on unseen data.

[0204] 2. Model parameter optimization and training

[0205] After the model is constructed (including the basin rainfall model, flood control project model, flood routing model), a series of key parameters (such as interception coefficient, infiltration coefficient, roughness coefficient, hydraulic conductivity coefficient, flood control project discharge coefficient, efficiency parameter, etc.) need to be trained and calibrated. The core steps are as follows:

[0206] ① Set the objective function to minimize the simulation error.

[0207] Common error measurement indicators include RMSE and NSE (Nash efficiency coefficient). In this embodiment, the South China Sea can calculate the errors for multiple indicators such as water level, flow rate, and inundation range respectively, and form a comprehensive objective function by weighting them, or adopt multi - objective optimization technology.

[0208] Optional indicator combinations include: water level simulation error, correlation coefficient of water level hydrograph, absolute error of flow rate simulation, Nash efficiency coefficient (NSE) of flow rate, area error of inundation range, flood routing time node error, etc.

[0209] ② Select any one of the following optimization algorithms to perform model parameter optimization calculation

[0210] Genetic algorithm (GA): Iteratively search through mechanisms such as selection, crossover, and mutation to minimize the error between model simulation and historical observation data.

[0211] Particle Swarm Optimization (PSO): By searching and updating through multiple particles (candidate solutions) in the parameter space, it can quickly approach the global optimum.

[0212] Simulated Annealing Algorithm: During the random search process from high temperature to low temperature, it can jump out of the local optimum and conduct a global search for model parameters.

[0213] Other possible algorithms: Such as Bayesian optimization, differential evolution, etc., as long as they can effectively search in the parameter space, they can be applied.

[0214] ③ Parameter Search Space and Iteration

[0215] 1) Parameter range setting: Based on literature, field measurements, and model experience, reasonable upper and lower limits are given for each parameter to be optimized (such as the interception coefficient / infiltration coefficient in the watershed rainfall model, the discharge coefficient / efficiency parameter in the flood control engineering model, the roughness coefficient in the flood routing model, etc.);

[0216] Avoid a too large range causing difficulties in algorithm convergence, or a too small range resulting in missing the true optimum.

[0217] 2) Iteration process:

[0218] Initialization: Generate a set of parameter vectors randomly or according to a distribution;

[0219] Evaluate fitness: Substitute each set of parameters into the model and calculate the error between the simulation results and the historical data of the training set;

[0220] Update operation:

[0221] The genetic algorithm will perform selection, crossover, and mutation;

[0222] Particle Swarm Optimization updates the velocity and position according to the global optimal particle and the individual optimal particle;

[0223] Simulated Annealing controls the random jump probability according to the temperature.

[0224] Multiple iterations: Continuously approach the global optimal solution with the minimum error or the highest efficiency coefficient;

[0225] At the same time, use the validation set to monitor in real time. If signs of overfitting appear (the training error decreases but the validation error increases), it is necessary to appropriately adjust the algorithm hyperparameters or reduce the number of iterations.

[0226] 3) Cross-validation:

[0227] Split the historical data into k parts, take turns as the training set / validation set, let the model iterate multiple times and statistically calculate the average performance, further prevent overfitting and improve the robustness and generalization ability of the model.

[0228] 3. Verify the initial parameters of the model

[0229] ① Select independent historical flood events as verification data: These do not participate in the training process and have a certain difference in time period or basin conditions from the training set (such as different rainfall distributions, different reservoir operation methods, etc.), which can comprehensively test the generalization ability of the model.

[0230] Select verification metrics: Similar to the error metric used in the training stage, including water level simulation error, flow simulation error, area error of inundation extent (AAE), time to peak flood, etc.; if the verification results maintain high accuracy on multiple metrics, the model is considered to perform well on this event.

[0231] ② Visualize the verification process:

[0232] Plot the water level hydrograph: Compare the simulated curve with the measured curve;

[0233] Inundation extent comparison map: Overlay the inundated area output by the model with actual remote sensing / aerial photography data to observe the coincidence degree and shape difference between the two;

[0234] ③ Quantitative error assessment

[0235] If the error metrics (such as relative water level error, Nash efficiency coefficient of flow, etc.) reach the set thresholds (such as water level error < 5%, NSE > 0.8), it indicates that the model already has acceptable accuracy; if the error is too large or there is an obvious systematic deviation, it is necessary to analyze whether there are problems with the model structure, data quality, or algorithm settings, and perform backtracking adjustments (such as modifying the model parameter range, improving the boundary condition settings, etc.).

[0236] 4. Evaluate the model

[0237] Before further applying it to the actual flood control scenario, the trained model can be comprehensively evaluated based on new flood events or high-precision monitoring data.

[0238] ① Select accuracy evaluation metrics

[0239] Water level simulation accuracy: MAE (mean absolute error), MRE (relative error), correlation coefficient of water level hydrograph (R), etc., which evaluate the accuracy of the model's water level simulation from aspects such as error magnitude, relative error ratio, and consistency of water level change trend;

[0240] Flow simulation accuracy: FAE (mean absolute error), FRE (relative error), Nash efficiency coefficient (NSE), which reflect the model's ability to simulate the magnitude and change process of flow and its fitting degree with the actual flow;

[0241] Inundation extent accuracy: AAE (inundation area error), SSIM (structural similarity index), which evaluate the accuracy of the model's prediction of the flood inundation area and the similarity degree with the actual inundation shape;

[0242] Flood routing time: such as the peak arrival time error (PTTE) and the recession time error (WTTE), which measure the prediction accuracy of the model for flood routing time nodes and comprehensively evaluate the performance of the model from multiple dimensions.

[0243] ② Evaluation data preparation

[0244] High-precision monitoring data: It can include high-precision water level and flow monitoring data from hydrological stations (time resolution can reach the minute level or even the second level), flood inundation range data obtained from high-resolution satellite remote sensing images or drone aerial photography (spatial resolution can reach the meter level), field survey records of flood routing time (including the time of peak flood appearance, the time when the flood starts to recede, etc.), and relevant social and economic impact data (such as population density and economic asset distribution in the affected area).

[0245] Independent flood events: Different from the aforementioned training / validation floods to test the adaptability to new scenarios.

[0246] ③ Analysis of evaluation results

[0247] Comparison of index values: If the water level MAE is less than 0.1m, R > 0.9, the flow NSE > 0.8, and AAE < 10% (for example only), it indicates that the model can accurately restore the evolution of floods in space and time;

[0248] If the evaluation results meet the standards: It means that the model meets the acceptable level of actual flood control work in terms of water level, flow, inundation range, peak arrival time, etc. and can be put into normal operation;

[0249] If the evaluation results are not satisfactory: It is necessary to trace the source and analyze:

[0250] 1) Check whether the model structure assumption is too simplified,

[0251] 2) Whether the parameter search range or the number of iterations is insufficient,

[0252] 3) Whether the data quality or the characteristics of the new scenario are too different from the historical training scenario.

[0253] After readjusting or optimizing, conduct the evaluation again until the model meets the application requirements.

[0254] Through the complete process of the above historical data collection and preparation → parameter optimization (training) → model verification → evaluation, the initial values and adjustment methods of various key parameters in the basin rainfall model, flood control project model, and flood routing model can be determined or optimized. This process has the following advantages:

[0255] Multi-dimensional precision guarantee: Optimize and verify multiple indicators such as water level, flow rate, inundation range, and flood evolution time simultaneously to ensure the all-round simulation ability of the model.

[0256] Data-driven global optimal search: Relying on advanced algorithms (genetic algorithm, particle swarm algorithm, simulated annealing, etc.), it can break through the limitations of traditional manual parameter tuning or local search and quickly approach the global optimal solution.

[0257] Avoid overfitting: Use the validation set and cross-validation techniques during training to ensure that the model also has good performance on more unseen events.

[0258] Facing complex flood control scenarios: Whether it is different rainfall patterns, the joint operation of various flood control projects, or the flood evolution under complex terrain conditions, the optimized model can accurately depict them.

[0259] When the evaluation indicators show that the model performance has reached the expected accuracy requirements, the trained model can be used for actual flood prediction and flood control scheduling support, such as:

[0260] Before the coming of future rainstorm events, predict the possible flood peaks and inundation ranges;

[0261] Provide real-time simulation calculations in the scheduling of facilities such as levees, reservoirs, and pumping stations;

[0262] Quickly simulate the flood evolution speed and range when dealing with emergency flood diversion, urban drainage and other working conditions.

[0263] So far, the initial parameters (or optimal parameters) of the model have been established and it has the basis for being applied in actual scenarios at any time. By continuously introducing new flood observation data and engineering scheduling information, the model can also be updated and recalibrated regularly to ensure that it always remains accurate and stable, providing sustainable scientific support for basin flood control work.

[0264] The basin rainfall model, flood control project model, and flood evolution model are all trained using the above process to determine their initial parameters.

[0265] The basin rainfall model focuses on the rainfall-runoff process (such as the impact of terrain, vegetation, and rainfall spatio-temporal distribution on the surface / groundwater cycle), and its core is to calibrate hydrological elements (such as interception coefficient, infiltration coefficient, soil saturation, etc.). Usually, independent hydrological calibration is first carried out based on the historical rainfall-runoff relationship, or runoff simulation calibration is carried out in the upstream area with less engineering scheduling influence (natural boundary conditions).

[0266] The core of the flood control engineering model lies in simulating the operation and structural characteristics of facilities such as dams, sluices, pumping stations, reservoirs, and flood diversion channels, involving flow coefficients, efficiency parameters, mechanical properties, etc. Generally, it can also be relatively independently calibrated for initial parameters with historical operation data (such as gate opening - flow relationship, reservoir water level - outflow discharge records, etc.).

[0267] The flood routing model includes overland flow concentration (kinematic wave theory) and channel flow concentration (Saint - Venant equations), and comprehensively considers the runoff generated from the basin rainfall model and the regulation effects of various flood control engineering facilities. Its parameters such as roughness coefficient, hydraulic conductivity, etc. can also be initially calibrated through independent channel hydrodynamics experiments or local flow - water level monitoring data.

[0268] In the final flood simulation framework, these three types of models will be coupled with each other - the runoff generated by the basin rainfall model will become the inflow boundary condition of the flood routing model, and various controls of the flood control engineering model (such as gate opening, reservoir discharge, etc.) will also affect the flood routing process in real - time. However, in the model training (parameter calibration) stage, the core parameters are often calibrated separately first, and then iterative coupling optimization is carried out in the integration stage.

[0269] In this embodiment, the basin rainfall model is calibrated first. First, lock the runoff relationship under meteorological and hydrological elements to obtain a more accurate surface / subsurface runoff output, reduce the inflow uncertainty of the subsequent flood routing model, enable it to accurately generate runoff under various rainfall patterns (heavy rain, rainstorm, continuous rainfall, etc.), and make the output runoff more credible as the input of the flood routing model.

[0270] The flood control engineering model is debugged in parallel or separately in the same stage. For example, the flow coefficient of the sluice model is corrected with the historical gate opening - flow relationship, and the efficiency curve of the pumping station model is corrected with the pumping and drainage records. According to the historical engineering scheduling data, independent parameter optimization is carried out for single facilities or facility groups such as dams, sluices, pumping stations, and reservoirs to ensure that the flow coefficients or scheduling response characteristics of each facility are accurately reflected in the model.

[0271] Then calibrate the flood routing model: Use the results of the first two stages (runoff input, engineering scheduling rules) as boundaries or conditions, and then calibrate the roughness coefficient, hydraulic conductivity, etc. of the flood routing model; compare with the measured river water level - flow data, inundation area images, etc. to evaluate the accuracy of the routing model.

[0272] Integrate the basin rainfall model and the flood control project model into the flood routing model and then conduct joint optimization: First, input the results of the updated basin rainfall model into the flood routing model. If there are large errors in some flood events, trace back to check whether the parameters of the basin rainfall model need to be fine-tuned. Similarly, if there are large deviations in the scheduling rules or efficiency parameters of the flood control project model in simulating the effect of flood peak reduction, conduct local optimization again. After repeating several times, ensure that the three models are in good consistency when jointly simulating floods.

[0273] At this stage, generally only fine-tune the hydraulic parameters such as Manning roughness in the flood routing model, or make secondary corrections to the local efficiency parameters in the flood control project model; when matching the observed water level process line and inundation area, make the overall error reach an acceptable threshold. After completing the step-by-step + coupled calibration, obtain a set of initial or optimal parameters that perform well under multiple scenarios in the entire basin.

[0274] The following further describes the execution process of the scenario-driven multi-granularity simulation method for flood control objects in this embodiment.

[0275] The first step is to match the current scenario pattern based on the real-time data of the current flood control object.

[0276] 1. Align the real-time data to a unified time step (such as every hour or shorter), and interpolate or aggregate the spatial data (if the flood control object covers multiple stations or grids).

[0277] 2. Extract (or calculate) the corresponding indicators from the original observed data according to the aforementioned feature dimensions, and assemble them into a scenario vector: V = [v 1 , v 2 , v 3 ,..., v n

[0278] Among them, each v i corresponds to a key feature quantity (such as rainfall, water level rising rate, gate opening, etc.).

[0279] 3. Calculate the scenario score.

[0280] First, set the scoring rules for different feature quantities. Common practices include:

[0281] Piecewise assignment: If the rainfall R total < 20mm, the score is 1; 20 - 50mm gets a score of 2; ≥ 50mm gets a score of 3;...

[0282] Linear or non-linear function: For example, for the water level rising rate, a weighted function can be defined to map it to the 0 - 10 score interval;

[0283] Then calculate the total score of each scenario vector: ​

[0284]

[0285] where f i () is the scoring function for the i-th characteristic quantity.

[0286] 4. Compare the scores of different scenario vectors

[0287] In practical applications, scenario vectors are often constructed separately for different types of scenarios (rainfall type, water level type, warning type, etc.) and scored in sequence;

[0288] It is also possible to use a unified scenario vector, but include characteristic quantities that distinguish different types in the vector and reflect weighted differences when scoring;

[0289] Select the one with the highest score as the "only current scenario mode" to avoid parallel or conflicting identifications.

[0290] 5. Generate the scenario mode determination result

[0291] Once the total score corresponding to a certain scenario vector is the highest, the system will output that the current flood control object is in "Scenario X";

[0292] If the score difference is not significant, additional rules can be set for secondary determination (such as giving priority to water level safety levels, warning levels, etc.) to ensure the most reliable decision-making in critical states.

[0293] If the system finds that there are too many misjudgments or delays in scoring in certain scenarios, the scoring thresholds of the corresponding characteristic quantities can be appropriately lowered or raised by comparing the model simulation results with the real observation data;

[0294] Machine learning methods can also be combined to perform clustering or classification mining on a large number of historical samples to further optimize the scenario classification rules.

[0295] The system can record the deviation between the model and the actual measurement as well as the user decision-making effect after each scenario switch; if there is a large deviation, automatic correction can be performed or experts can be prompted to manually correct the scenario threshold / scoring function to make the matching of the scenario mode closer to the actual working conditions of the flood control object.

[0296] In the second step, optimize the parameters of the basin rainfall model, flood control project model, and flood routing model according to the determined scenario mode.

[0297] Once the current scenario mode is locked, call the mapping model (or the pre-defined parameter mapping relationship) to fine-tune or update the weights of the relevant parameters of the basin rainfall model, flood control project model, and flood routing model;

[0298] For example: a high-intensity rainfall scenario will increase the infiltration coefficient or change the reserved gate scheduling volume, while a low water level scenario may lower the pump station start / stop threshold, etc.

[0299] The mapping model is used to provide parameter adjustment directions and adjustment coefficients for the basin rainfall model, flood control project model, and flood routing model under different scenario modes, so as to achieve dynamic and accurate flood simulation. The entire training process can be summarized as the following steps:

[0300] 1. Data collection and preparation

[0301] Multi-source observation data: Collect meteorological, hydrological, engineering operation and other data related to flood control over the years, covering different rainfall scenarios, water level scenarios, warning scenarios, etc.

[0302] Scenario vector construction: Digitalize the scenario characteristics (such as rainfall intensity, duration, water level rising rate, warning level, etc.) in different time periods (or events) to form a series of scenario vectors.

[0303] Optimal parameter recording: For each historical scenario, obtain the "optimal" model parameter adjustment amount and corresponding adjustment coefficient after actual measurement or simulation calibration. For example, the infiltration coefficient correction amount for the basin rainfall model, the gate opening control amount for the flood control project model, the Manning roughness coefficient increase or decrease amount for the flood routing model, etc.

[0304] Matching relationship: Integrate the "scenario vector" and the "corresponding parameter adjustment amount / coefficient" into training samples

[0305] Missing value and outlier handling: For example, delete abnormal samples, interpolate or fill in missing values.

[0306] Feature engineering: Process the scenario vectors by standardization, normalization, or dimensionality reduction (PCA), etc., to improve the training efficiency and stability.

[0307] 2. Model selection and construction

[0308] 1) Mapping model input: Each dimension of the scenario vector (for example: rainfall amount, water level rising rate, warning change rate, etc.).

[0309] Mapping model output: Parameter adjustment amount (or adjustment coefficient), that is, the numerical value that the corresponding parameters of each sub-model (basin rainfall model, flood control project model, flood routing model) need to increase or decrease.

[0310] 2) The mapping model can be selected as follows:

[0311] Regression models: Such as linear regression, support vector regression (SVR), random forest regression, gradient boosting regression (GBDT), etc.

[0312] Neural networks: Such as multi-layer perceptron (MLP), LSTM, etc., which are suitable for processing complex non-linear mappings.

[0313] Rule-based or expert system: If the adjustment amount has an obvious threshold pattern, a simple rule-based mapping can also be adopted.

[0314] 3. Model training

[0315] Divide the overall dataset into a training set, a validation set, and a test set (common ratios such as 6:2:2 or 7:2:1, etc.) for model fitting, hyperparameter tuning, and final evaluation.

[0316] If the parameter adjustment amount is output in a regression manner, mean squared error (MSE), mean absolute error (MAE), etc. can be used as loss functions.

[0317] Combine methods such as gradient descent, stochastic gradient descent (SGD), Adam, LBFGS, etc., and gradually iteratively adjust the internal parameters of the model.

[0318] If the validation set error no longer decreases or there is a trend of overfitting, terminate the training.

[0319] Automatically search for the optimal configuration (such as the depth of the random forest tree, learning rate, etc.) among a series of hyperparameter combinations. Use Bayesian optimization / BayesianOptimization: More efficiently search for the global optimal combination in the hyperparameter space. Use K-fold cross-validation to evaluate the robustness of the model and prevent overfitting.

[0320] 4. Model evaluation and validation

[0321] Numerical error metrics include: RMSE, MAE, R 2 etc., which are used to evaluate the accuracy of the mapping model in predicting the parameter adjustment amount.

[0322] Scenario classification accuracy includes (optional): If there is a classification part in the model (such as identifying which scenario triggers which adjustment strategy), accuracy, F1-score, etc. can be used.

[0323] Conduct scenario simulation comparison based on external validation:

[0324] Apply the parameter adjustment amount output by the mapping model to the basin rainfall model, flood control project model, and flood routing model for a round of simulation.

[0325] Compare the simulation effects of the "no adjustment" or "previous adjustment" schemes, such as the flood peak flow error, inundation area comparison, water level process comparison, etc., and comprehensively measure the actual improvement degree.

[0326] The third step is simulation calculation.

[0327] Based on the input parameters, the model combines the geospatial scene model, the basin rainfall model, the flood control project model, and the flood routing model, and calculates the flood routing process under different scenarios according to the predetermined calculation process and algorithm. During the calculation process, the changes in rainfall data, the operation status data of flood control projects, and basic data such as topography and landform are updated in real time to ensure the accuracy and timeliness of the simulation results. Through large-scale parallel computing technology and high-performance computing platforms, the speed and efficiency of simulation calculations are improved, and the simulation calculations of complex flood control scenarios can be completed within a short time (such as a few minutes to dozens of minutes), and various forms of simulation results are generated, including water level change curves, flow hydrographs, dynamic change maps of inundation areas, flood routing animations, etc., to intuitively and vividly display the flood routing process in the basin and its impact on the surrounding environment.

[0328] Specifically, the geospatial scene model provides spatial basic information such as three-dimensional topography and landform, river system morphology, and land use distribution. Precise geographical and geometric boundary conditions are set for the basin rainfall model, the flood control project model, and the flood routing model, such as river cross-sections, dam positions and geometric dimensions, and urban building distributions.

[0329] During the simulation calculation process, if basic data such as topography, land use, or river morphology is corrected during the simulation (such as newly measured river cross-sections), the geospatial scene model will be updated synchronously; during subsequent flood routing simulations, the calculation results of each model (such as water level, flow velocity, inundation area) will be mapped back to the three-dimensional scene for subsequent dynamic display.

[0330] The basin rainfall model is responsible for simulating the temporal and spatial distribution and intensity of rainfall within the basin, including the coupled effects of topography, vegetation, and meteorology. Through distributed hydrological models such as TOPMODEL and VIC, surface or near-surface runoff, soil moisture distribution, etc. are output, providing "rainfall source input" for the flood routing model.

[0331] During the simulation calculation process, if new strong rainfall units or information such as raindrop size distribution are observed by meteorological stations or radars, the model can cooperate with Kriging or other interpolation techniques to project rainfall data onto the DEM grid; based on TOPMODEL, use the topographic index to identify saturated areas and determine the initial runoff conditions; based on VIC, combine vegetation interception, soil infiltration, and atmospheric elements to continuously output the runoff and soil moisture that change over time. At the same time, interface with the flood routing model: transfer the time-series runoff data to the flood routing model as the upstream inflow boundary or areal rainfall distribution of channel or slope confluence.

[0332] The flood control engineering model simulates the structure and operation status of flood control facilities (such as dams, sluices, pumping stations, reservoirs, flood diversion channels, etc.), and gives their flood control effects (such as flood interception, flood discharge, pumping drainage, etc.) under various working conditions. The parameters of these facilities (gate opening, pumping station operation efficiency, reservoir flood discharge curve, etc.) directly affect the water level, flow rate and inundation range during the flood evolution process.

[0333] During the simulation calculation process, the dam model judges in real time whether there is an overloading impact and seepage risk; the sluice model calculates the flow rate through the sluice according to the water level difference between upstream and downstream and the gate opening, etc., and plays a role in reducing or diverting the flood peak; the pumping station model pumps water for drainage in low-lying areas, and dynamically calculates the pumping drainage capacity with the change of water level or motor efficiency; the reservoir model dynamically outputs the outflow from the reservoir and updates the water level in the reservoir area according to the storage capacity-water level curve and the operation strategy (such as flood storage, flood discharge, power generation water use, etc.); the flood diversion channel model accurately simulates the flood diversion operation under certain flood scenarios based on the river channel morphology and flood carrying capacity, and judges the flood control effect downstream.

[0334] The simulation results of each facility (such as the actual flow rate of the sluice, the water level in the reservoir area, the flow rate of the pumping station) are then input into the flood evolution model to form the linkage effect of the complete flood control system.

[0335] The flood evolution model takes the runoff from the basin rainfall model and the boundaries / operations of water conservancy facilities from the flood control engineering model as comprehensive inputs, and uses a method combining the kinematic wave theory (overland flow) and the Saint-Venant equations (river channel hydrodynamics) to simulate the dynamic evolution of floods on the basin surface and in the river channels;

[0336] Outputs core results such as water level change curves, velocity distributions, inundation ranges and flood evolution times.

[0337] During the simulation calculation process, based on the slope, soil permeability, etc. in the geospatial scene, the surface runoff provided by the basin rainfall model is converted into the overland flow direction and velocity; the Saint-Venant equations are discretely solved for the main river channel and its tributaries, and the regulation of dams, sluices, reservoirs, pumping stations, and flood diversion channels is coupled; if a certain sluice suddenly opens or a pumping station starts, the model refreshes the river channel flow rate and water level according to the operation status of the facilities during that period; parallel algorithms and high-performance computing platforms are used in large basins or complex terrains to complete the multi-dimensional discrete solution of the hydrodynamic equations in a short time; innovative spatial analysis algorithms and models (such as a flood inundation model based on an improved seed filling algorithm) are adopted, combined with terrain data and the principles of water flow dynamics, to accurately present the dynamic changes of the boundaries and depth degrees of the inundation range. This improved seed filling algorithm takes into account the influence of factors such as the water flow velocity u (unit: m / s) and the terrain slope θ on the inundation propagation based on the traditional seed filling algorithm. Let the inundation propagation velocity V satisfy the relationship v = u×cosθ + k 1 ×sinθ, where k1 is a coefficient determined according to the terrain resistance characteristics. In this way, the flooding process of floods under complex terrains can be simulated more accurately.

[0338] Specifically, the simulation calculation process has the following real-time update and dynamic adjustment mechanisms:

[0339] When the meteorological station / radar detects the coming of a new round of heavy rainfall, the basin rainfall model immediately receives it and performs spatial interpolation; the runoff yield / temporal rainfall intensity is output to the flood routing model to refresh the overland and channel inflows.

[0340] The gate opening changes in real time according to the dispatching order, and the operating conditions of the pumping station or reservoir (such as motor failure / water level exceeding the limit) will also trigger the adjustment of the model boundary or parameters; if local seepage or over-standard water level is found in the dam, the model will record and correct the force parameters or leakage volume of the dam accordingly.

[0341] If local terrain changes such as construction, emergency dam heightening, or temporary soil stacking occur, the vector and DEM data can be updated in the geospatial scene model and the flood routing model can be notified to perform local remeshing or dispatching correction.

[0342] If the high-precision water level monitoring equipment detects that the water level rising rate is greater than the predicted value, the system can automatically trigger the "dynamic correction" mechanism of the flood routing model to fine-tune the roughness coefficient or other parameters to make the simulation closer to the measured values.

[0343] Specifically, the simulation calculation process uses a high-performance computing cluster or GPU parallel acceleration to allocate the basin grid (or river channel discrete units) to multiple cores or nodes for simultaneous solution; the basin rainfall model and the flood routing model can process different time periods or grids in parallel, and the dispatching operations of flood control engineering facilities can also be independently executed and then integrated.

[0344] Large-scale flood routing simulations can be completed within a few minutes to dozens of minutes, which is convenient for meeting the requirements of emergency command; if a two-dimensional or three-dimensional hydrodynamic model based on the Saint–Venant equations is used and combined with an improved seed filling algorithm for flood inundation range search, the calculation efficiency can also be guaranteed after numerical discretization and parallelization.

[0345] Specifically, the simulation results in the embodiment are output in a variety of ways:

[0346] Water level change curve: Generate a water level process line that changes with time at key cross-sections or monitoring points;

[0347] Discharge process line: Show the discharge magnitude and the time of peak occurrence at the main stream and tributary hydrological stations;

[0348] Dynamic change map of flood inundation range: Intuitively show in the geospatial scene which areas are being inundated and the gradient of inundation depth;

[0349] Flood evolution animation: simulates the process of flood peak propagation in the river channel, and the real-time impact of flood diversion channel or reservoir operation on flood peak reduction.

[0350] Specifically, under different flood control scenario modes (such as heavy rainfall scenarios, sluice emergency flood discharge scenarios, reservoir peak scheduling scenarios, etc.), in order to meet the decision-making needs at the macro, meso and micro levels, the system provides multi-granularity visualization solutions.

[0351] 1. Macro level: understanding the situation of the entire river basin

[0352] Using the geospatial scene model as the base map, the water depth distribution output by the flood evolution model is plotted over time to create a “water level stratified coloring map” with different shades of color.

[0353] The blue to dark blue gradient is used to visually show the evolution of the flood peak from upstream to downstream.

[0354] The flow / water level process lines of each key station are superimposed on one side of the screen to assist in determining the flood peak arrival time, peak value and the impact of the confluence of each tributary on the main stream.

[0355] 2. Meso-level: key node analysis

[0356] Locate important embankment sections, sluice gates, pumping stations or reservoirs for partial zooming to view real-time water level differences, gate openings, pumping station flow, etc.

[0357] In conjunction with the velocity vector display, observe the impact of downstream water acceleration areas after the sluice gate releases water or the reservoir discharges floodwater on the downstream flood diversion channel.

[0358] Highlight the velocity distribution and lateral flooding spread in the floodway in the form of velocity vectors or contour maps;

[0359] At the same time, the impact on surrounding farmland and villages is observed to help decision makers determine whether further flood diversion or capacity expansion is needed.

[0360] 3. Micro level: detailed operation simulation of a single water conservancy facility

[0361] Display stress distribution of dam sections, seepage line analysis, or velocity distribution of gate water flowing through gate sections and gate holes;

[0362] If water level monitoring finds a sudden increase in seepage on the outer slope of the dam, the model can call the corresponding mechanical calculation module to evaluate changes in the dam's safety factor.

[0363] Dynamically visualize the pump station operating status through real-time power, head, and pumping efficiency curves to identify potential failure risks due to high load or overload.

[0364] 4. Switch between different scene modes

[0365] 1) Heavy rainfall pattern

[0366] Use co-kriging interpolation and TOPMODEL+VIC model to conduct real-time simulation of the spatio-temporal evolution of heavy rainstorms, and dynamically display the runoff generation situation in the basin and the water inflow of each tributary.

[0367] Visualization key points: Heavy rainstorm center, urban waterlogging risk points, slope confluence paths.

[0368] 2) Emergency flood discharge mode of sluices

[0369] After increasing the opening of the sluice gate, the changes in the main river channel flow velocity and downstream water level in the flood routing model need to be updated in real time.

[0370] Visualization key points: Flood carrying capacity of key river reaches downstream and flood diversion channels, surrounding inundation risks.

[0371] 3) Reservoir peak shaving operation mode

[0372] The timing matching of upstream water inflow and reservoir flood discharge has a significant impact on the downstream water level and inundation range;

[0373] Visualization key points: Reservoir water level curve, dynamic changes of reservoir area inundation, impact on towns due to changes in the arrival time of downstream flood peaks.

[0374] 4) Comprehensive linkage mode of multiple facilities

[0375] Dispatch multiple sluices, pumping stations, and flood diversion channels simultaneously, and view the contributions of each facility and the local water level evolution at macro-meso-micro scales;

[0376] Facilitate emergency command to quickly evaluate the overall flood control effect under complex working conditions.

[0377] This embodiment can provide the simulation results to flood control decision-making departments in a timely manner, providing strong technical basis for formulating scientific and reasonable flood control decisions. For example, according to the simulated inundation range and depth information, determine the specific scope and priority of the personnel and property to be evacuated, and formulate detailed evacuation routes and resettlement plans for the personnel; based on the predicted results of water level changes and flow rate changes, arrange the maintenance, reinforcement and emergency rescue work of flood control engineering facilities in advance, such as determining the heightening and reinforcement positions and quantities of dikes, adjusting the opening and closing times of sluice gates, optimizing the operation strategies of pumping stations, etc.; according to the simulation results under different flood control engineering regulation schemes, compare and analyze the control effects of various schemes on flood evolution and the possible negative impacts, and select the optimal regulation strategy, such as determining the best flood discharge timing, flood discharge flow rate and flood discharge duration of the reservoir, so as to achieve the purpose of minimizing flood disaster losses, ensuring the safety of people's lives and property and social and economic stability. At the same time, by using the real-time simulation ability of the model, adjust the simulation parameters and scenarios in a timely manner according to the changes in the actual situation during the flood control process (such as real-time update of rainfall conditions, accidental failures of flood control engineering facilities, etc.), provide dynamic and continuous technical support for emergency decision-making, ensure that flood control work can flexibly respond to various emergencies, and improve the overall efficiency of flood control and disaster reduction work.

[0378] In terms of flood simulation accuracy, by virtue of the in-depth integration of multi-source data (topography, vegetation, meteorology, flood control projects, hydrological station data, etc.) and the application of advanced model algorithms (TOPMODEL, VIC model, Saint-Venant equation solving model, etc.), the present invention can extremely accurately reflect the formation mechanism and evolution process of floods, significantly improving the reliability and practicality of the simulation results, providing highly accurate data support for flood control decisions, and helping to accurately deploy flood control resources in advance, such as the allocation of sandbags, emergency rescue personnel, rescue equipment, etc.

[0379] Based on the visualization effect of the digital twin engine, the present invention converts complex and abstract flood data into vivid, easy-to-understand three-dimensional dynamic images and animations, which can clearly display details such as the evolution path of floods in the geographical space scene, the dynamic changes of the inundation range, and the interaction with the surrounding environment. This enables decision-makers to intuitively observe the subtle changes in the flood control situation, greatly shortening the decision-making time (the decision-making time can be shortened by more than 30%) and improving the decision-making quality, and reducing losses caused by decision-making delays or mistakes.

[0380] The multi-granularity and multi-scale simulation capabilities of the present invention comprehensively cover and seamlessly connect the full range of requirements from macro watershed planning to micro water conservancy facility operation management. At the macro level, it can provide scientific and detailed simulation analysis data for the strategic layout of large-scale water conservancy projects (such as determining the location of large reservoirs, the regulation and planning of backbone river channels within the watershed, etc.); at the meso level, it can accurately simulate the collaborative flood control effects among multiple water conservancy facilities within a region (such as the control effect of the joint operation of multiple sluice gates on the regional water level, the water volume allocation relationship between pumping stations and reservoirs, etc.) and the propagation characteristics of floods within the region (such as the spreading path and inundation sequence of floods in urban areas, etc.); at the micro level, it can accurately simulate the fine operation status of a single water conservancy facility (such as the force distribution on the gate of a small sluice under flood impact, the motor load change of a pumping station during high water level operation, the flow capacity and structural stability of a culvert under different water flow conditions, etc.), providing a strong basis for the refined management and maintenance of water conservancy facilities, extending the service life of water conservancy facilities, and reducing maintenance costs.

[0381] The characteristic of scenario-driven endows the simulation process with a high degree of pertinence and flexibility, enabling it to quickly switch simulation strategies and generate practical simulation results for different complex flood control scenarios such as different rainfall patterns, water conservancy project regulation, and external environmental factors. For example, in the face of the scenario of the superposition of heavy rain and reservoir flood discharge, it can quickly adjust the model parameters and accurately simulate the evolution of floods in the downstream river channel, effectively improving the adaptability and application value of the simulation technology in actual flood control work and enhancing the ability to cope with complex and changing flood control situations.

[0382] At the same time, the present invention can respond in real time to the warning information of hydrological stations and provide a scientific and rigorous basis for emergency decision-making in the first time (the simulation can be started within 5 minutes after the warning information is released), thereby effectively reducing the threat of flood disasters to the safety of people's lives and property (the proportion of affected population can be reduced by more than 20%) and the impact on social and economic order (the economic loss can be reduced by more than 30%), laying a solid foundation for ensuring social stability and sustainable development.

[0383] Embodiment 2

[0384] The present invention provides a scenario-driven multi-granularity simulation system for flood control objects, including:

[0385] A scenario mode matching module, configured to match the scenario mode in which the flood control object is currently located according to the real-time data of the flood control object;

[0386] A model parameter optimization module, configured to optimize the parameters of the basin rainfall model, flood control project model, and flood evolution model according to the determined scenario mode;

[0387] A geospatial scene update module, configured to update the geospatial scene model according to the real-time data of the flood control object;

[0388] A rainfall simulation module, configured to simulate the spatio-temporal distribution change process of rainfall in the basin where the flood control object is located through a basin rainfall model based on the real-time data of the flood control object;

[0389] A flood control project simulation module, configured to comprehensively simulate and evaluate the structural parameters and operation characteristics of the flood control facilities of the flood control object through a flood control project model based on the real-time data of the flood control object, so as to reflect the regulation effect of the flood control facilities on flood evolution and the overall flood control effect;

[0390] A flood evolution model, configured to simulate the overland flow and channel flow processes of flood on the terrain surface of the flood control object based on the calculation results of the basin rainfall model and the flood control project model with a geospatial scene model as the basic framework through the flood evolution model;

[0391] A visualization module, configured to visualize the specific calculation results of the basin rainfall model, the flood control project model, and the flood evolution model in the current scenario mode based on the geospatial scene model.

[0392] Embodiment 3

[0393] The present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the multi-granularity simulation method for flood control objects based on scenario-driven described in the above technical solution is implemented.

[0394] Contents not described in detail in this specification belong to the prior art well-known to those skilled in the art.

Claims

1. A scenario-driven multi-granularity simulation method for flood control objects, characterized by: The following steps are involved: According to the real-time data of flood control objects, match the current scenario mode of flood control objects; Optimize the parameters of the basin rainfall model, flood control engineering model and flood evolution model according to the determined scenario model; Update geospatial scenario models based on real-time data of flood control objects; Based on the real-time data of flood control objects, the temporal and spatial distribution of rainfall in the basin where the flood control object is located is simulated through the basin rainfall model; Based on the real-time data of flood control objects, the structural parameters and operating characteristics of the flood control facilities of the flood control objects are comprehensively simulated and evaluated through the flood control engineering model to reflect the regulatory role of flood control facilities on flood evolution and the overall flood control effect; The flood evolution model uses the geospatial scenario model as the basic framework and simulates the flood confluence process on the slope and river confluence of the terrain surface of the flood control object based on the calculation results of the basin rainfall model and the flood control engineering model. Based on the geospatial scenario model, the specific calculation results of the basin rainfall model, flood control engineering model and flood evolution model under the current scenario mode are visualized.

2. The method according to claim 1, characterized in that: The process of matching scenario mode includes: Assemble the real-time data into a number of scenario vectors according to the set feature dimensions; the scenario vectors are used to represent the key feature quantities of a scenario; Scoring each scenario vector based on the threshold interval of each feature quantity in each scenario vector; The scenario mode corresponding to the scenario vector with the highest score is selected as the current scenario mode of the flood control object.

3. The method according to claim 1, characterized in that: Each scenario mode corresponds to a visualization template. Once the scenario mode is switched, the corresponding visualization template is automatically scheduled. The visualization feature quantity selected by the external instruction in each scenario mode is recorded in real time. When the frequency of any feature quantity selected by the external instruction in any scenario mode reaches the set threshold, the feature quantity is added to the visualization template corresponding to the scenario mode.

4. The method according to claim 2, characterized in that: The basin rainfall model, flood control engineering model and flood evolution model are all provided with initial model parameters; based on the mapping model, the parameters and adjustment coefficients that need to be adjusted for the basin rainfall model, flood control engineering model and flood evolution model under the current scenario mode are determined, and corresponding optimization is performed based on the initial model parameters; the mapping model is used to reflect the mapping relationship between the scenario mode and the model parameter adjustment method.

5. The method according to claim 4, characterized in that: The process of building a mapping model includes: Forming a historical scenario vector based on historical data; For each historical scenario vector, obtain the model parameter adjustment amount and corresponding adjustment coefficient that are considered to be the optimal for the basin rainfall model, flood control engineering model and flood evolution model after actual measurement or simulation calibration; Integrate historical scenario vectors and corresponding parameter adjustment coefficients into training samples to form a mapping training set; The mapping training set is used to train the machine learning or regression model, and the trained model is used as the mapping model.

6. The method according to claim 4, characterized in that: The initial model parameter setting process of the basin rainfall model, flood control engineering model and flood evolution model includes: Initialize the model parameters of the basin rainfall model, flood control engineering model and flood evolution model; Calibrate the above models through historical flood events or a series of typical operating conditions to make the model output consistent with historical observations; Select historical flood events or data from different years that are not involved in the calibration to test the simulation accuracy of the model under the event; Through several calibration-validation iterations, the model parameters are gradually corrected and optimized; The parameters that can optimize the error indicators of each model are selected as the final initial model parameters.

7. The method according to claim 1, characterized in that: The construction process of the geospatial scene model includes: Import the digital elevation model data into the GIS platform, and use terrain rendering technology to display the topography of the basin where the flood control object is located in three-dimensional form, and set up enhanced terrain visualization effects to make it highly consistent with the actual terrain features; The land use type data of flood control objects are represented in the model as polygonal elements with different colors and textures; The river system data of the flood control object is presented as vector line elements with width and depth attributes. It is drawn according to the actual shape of the river channel, and the flow state of the river is simulated by setting water flow texture animation.

8. The method according to claim 7, characterized in that: The construction process of the basin rainfall model includes: The first hydrological model is used to calculate the terrain index of each grid based on the terrain data in the geospatial scene model to identify the distribution of soil saturated areas and runoff generation conditions; the spatial distribution of soil moisture is updated at the beginning of the simulation or in each period to determine the potential rapid runoff generation area; The second hydrological model is used to simulate atmospheric precipitation, vegetation interception and evapotranspiration, soil infiltration and water storage, and surface runoff and underground runoff processes based on water balance and energy balance equations; corresponding parameters are set according to different vegetation types and soil stratifications in the basin; The saturated area and initial soil moisture information output by the first hydrological model are mapped to the grid cells of the VIC, thereby affecting the runoff generation and confluence process of each grid in the second hydrological model; Ensure that the first and second hydrological models are run at the same or matching time steps, exchanging key variables regularly.

9. The method according to claim 1, characterized in that: The construction process of the flood evolution model includes: Based on the geospatial scenario model as the basic framework, the rainfall data generated by the basin rainfall model is used as the water source input for floods; In the simulation of slope runoff, a slope flow model based on motion wave theory is used to consider the effects of terrain slope, vegetation cover, and soil infiltration on the formation and velocity of slope runoff, and to calculate the flow and direction of slope runoff. In the river confluence simulation, a river hydrodynamic model based on the Saint-Venant equations is used, combined with the regulatory role of various flood control facilities in the flood control engineering model to simulate the evolution of floods in the river.

10. A scenario-driven multi-granularity simulation system for flood control objects, characterized in that include: A scenario mode matching module is used to match the scenario mode of the flood control object according to the real-time data of the flood control object; Model parameter optimization module, used to optimize the parameters of the basin rainfall model, flood control engineering model and flood evolution model according to the determined scenario mode; A geospatial scene update module, used to update the geospatial scene model according to the real-time data of flood control objects; The rainfall simulation module is used to simulate the temporal and spatial distribution change process of rainfall in the basin where the flood control object is located through the basin rainfall model based on the real-time data of the flood control object; The flood control engineering simulation module is used to comprehensively simulate and evaluate the structural parameters and operating characteristics of the flood control facilities of the flood control objects through the flood control engineering model based on the real-time data of the flood control objects, so as to reflect the regulatory role of the flood control facilities on the evolution of floods and the overall flood control effect; The flood evolution model is used to simulate the slope confluence and river confluence process of floods on the terrain surface of flood control objects based on the calculation results of the basin rainfall model and the flood control engineering model, using the geospatial scenario model as the basic framework; The visualization module is used to visualize the specific calculation results of the basin rainfall model, flood control engineering model and flood evolution model under the current scenario mode based on the geospatial scene model.

Images