A water power analog flood method, system, electronic device and storage medium
The two-dimensional hydrodynamic simulation flood method driven by high-point monitoring video data solves the problems of limited coverage, insufficient real-time performance and rigid boundary conditions in traditional flood simulation, and realizes high-precision and adaptive simulation of flood inundation range.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA TOWER CO LTD
- Filing Date
- 2026-01-30
- Publication Date
- 2026-06-05
AI Technical Summary
In traditional flood simulation technology, ground sensors have limited coverage, are easily damaged, lack real-time performance, and are costly. Existing video analysis technology lacks multi-dimensional parameter fusion and dynamic boundary condition optimization, making it difficult to meet the high-frequency and high-precision modeling requirements.
Geometric correction and preprocessing are performed using high-point monitoring video data to extract river flow elements, construct a two-dimensional hydrodynamic model, optimize the initial water depth field and downstream boundary conditions by combining historical video data, and dynamically adjust the model parameters through optimization algorithms to achieve real-time simulation of flood inundation range.
It achieves non-contact comprehensive monitoring, multi-dimensional hydrological parameter extraction, high-precision initial water depth field generation, and adaptive river channel changes, thereby improving the adaptability and simulation accuracy of the two-dimensional hydrodynamic model and forming a closed-loop feedback mechanism.
Smart Images
Figure CN122154527A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of flood simulation technology, and specifically relates to a hydrodynamic flood simulation method, system, electronic device and storage medium. Background Technology
[0002] In recent years, flood simulation technology has developed rapidly in the fields of hydrological monitoring and disaster prevention and mitigation. Two-dimensional hydrodynamic models, due to their ability to accurately depict the spatial distribution characteristics of flood evolution, have gradually become the mainstream tool. However, the accuracy and real-time performance of these models are highly dependent on the accuracy of initial conditions and boundary parameters. Traditional techniques mainly obtain key parameters such as river level and flow velocity through ground sensors (such as water level gauges, current meters, and radar flow measurement equipment), but these methods have the following problems: Data acquisition limitations: Ground sensors need to be deployed in fixed locations, making it difficult to cover complex river topography and flood-prone areas, and they are easily damaged by flood impacts, resulting in data loss.
[0003] Insufficient real-time performance: Sensor data transmission relies on wired or wireless communication networks, which may experience delays or interruptions in extreme weather conditions, making it difficult to provide dynamic updates for the model.
[0004] Cost and maintenance challenges: The construction and maintenance of sensor networks are costly, making them particularly uneconomical when applied to remote or large-scale watersheds.
[0005] To overcome these limitations, some studies have attempted to introduce remote sensing technologies (such as satellite imagery and drone aerial photography) to assist flood monitoring. However, these technologies are limited by low resolution, long revisit periods, and cloud cover, making it difficult to meet the high-frequency, high-precision modeling requirements. In recent years, hydrological parameter extraction technologies based on video analysis have gradually emerged, such as retrieving surface flow velocity or water level information from monitoring videos. However, existing technologies mostly focus on local measurements of single parameters (such as flow velocity), lacking deep integration with two-dimensional hydrodynamic models, and there is little research on optimizing model boundary conditions using historical video data.
[0006] While using drones to collect river images to generate inundation extent maps, the problem of dynamically assigning initial water depth field values to the model remains unresolved, and the reliance on manual flight control limits its timeliness. Using monitoring videos to train neural networks to estimate flow velocity fails to correlate with hydrodynamic model parameters and does not address the optimization of downstream boundary conditions. Initial conditions depend on measured or interpolated data, and boundary conditions require manually preset fixed relationship curves, making it difficult to adapt to the dynamic hydraulic responses caused by riverbed erosion and deposition. Summary of the Invention
[0007] To address the aforementioned problems, firstly, this application provides a method for simulating floods using hydrodynamics, comprising the following steps: Acquire video data from high-point monitoring, perform geometric correction and preprocessing on the video data, delineate the flood simulation area, and acquire the topographic elevation data of the flood simulation area; Based on the preprocessed video data, river flow element data is extracted, including the river water level, surface velocity and cross-sectional flow rate. Based on the flood simulation area, the simulation boundary is delineated, the computational domain is discretized into multiple triangular grids, and different Manning roughness coefficients are assigned to different land types. The model boundary conditions are preset, and a two-dimensional hydrodynamic model is constructed. By combining the extracted water level and topographic elevation data, the initial water depth field of the two-dimensional hydrodynamic model is generated, and the downstream boundary conditions of the two-dimensional hydrodynamic model are determined using historical video data. The initial water depth field, cross-sectional flow rate, and downstream boundary conditions are imported into a two-dimensional hydrodynamic model for initialization. The water level simulated by the two-dimensional hydrodynamic model is compared with the real-time water level observed in the video at set intervals, and the model parameters are optimized based on the comparison results. The two-dimensional hydrodynamic model is run based on the optimized model parameters, and the flood inundation range is determined after real-time calculation of key water flow variables.
[0008] Furthermore, acquiring video data from high-point monitoring, and after geometric correction and preprocessing the video data, delineating the flood simulation area, includes the following steps: Ground control points are set up within the monitoring range of the high-point surveillance cameras, and the internal and external parameters of the cameras are calculated using the calibration method. The video data of high-point monitoring is obtained by using the calibrated internal and external parameters of the camera and decomposed into continuous video frames. Based on the perspective transformation model, the video frames are projected onto the river plane coordinate system for geometric correction. By identifying river surface areas in video frames using a semantic segmentation model, a dynamic mask is generated. This mask is then combined with the historical flood inundation range to expand the simulation range to the potential inundation area and delineate the flood simulation area. The geometrically corrected video data is preprocessed, including data cleaning, data interpolation, and format conversion.
[0009] Furthermore, the data cleaning includes cleaning outliers and noise from the video data; The data interpolation includes filling in missing values in video data using time series or spatial interpolation methods; The format conversion includes standardizing the video data into a uniform input format.
[0010] Furthermore, the extraction of river flow element data based on the preprocessed video data includes the following steps: Within the designated flood simulation area, virtual water level markers are set along the river cross-section. Based on edge detection and Hough transform, the intersections of the water surface line and the markers in the video frame are extracted, and the actual water level value is calculated. The pixel displacement of floating objects on the water surface is tracked using the sparse optical flow method or the surface velocity is calculated based on the RAFT deep learning model. The cross-sectional area of the river is calculated based on the actual water level, and the cross-sectional flow rate of the river is calculated by combining the cross-sectional area and the surface velocity.
[0011] Furthermore, the step of generating the initial water depth field of the two-dimensional hydrodynamic model by combining the extracted water level and topographic elevation data includes the following steps: The flood simulation area was divided into a triangular grid, and the terrain elevation data corresponding to each grid node was extracted. For network nodes covered by water levels, the water depth of the network node is calculated using the actual water level value and topographic elevation data of the network node; For network nodes that are not directly covered by water level, Gaussian process regression interpolation is used to estimate water level, and the interpolation smoothness is constrained by the river channel topographic slope. Set a reasonable water depth threshold and screen the water depth of all grid nodes. If the water depth of a certain network node exceeds the reasonable threshold, call a deep learning model pre-trained based on historical flood data to correct it.
[0012] Furthermore, the determination of downstream boundary conditions for the two-dimensional hydrodynamic model using historical video data includes the following steps: Collect and preprocess historical video data corresponding to the flood simulation area, and extract water level-flow time series data of the downstream section within the historical video period; A relational model is constructed, with water level data, historical upstream and downstream discharge, and historical river roughness coefficient from the water level-discharge time series data as inputs to the relational model, and the predicted discharge value as the output of the relational model. The relational model is trained with mean square error and the physical of the hydraulic continuity equation as constraints. The trained relational model is embedded into the downstream boundary of the two-dimensional hydrodynamic model, and the current flow prediction value is used as the downstream boundary condition by the relational model.
[0013] Furthermore, the initial water depth field, cross-sectional flow rate, and downstream boundary conditions are imported into a two-dimensional hydrodynamic model for initialization. The water level simulated by the two-dimensional hydrodynamic model is compared with the real-time video observation water level at set intervals, and the model parameters are optimized based on the comparison results, including the following steps: The initial water depth field is imported into the two-dimensional hydrodynamic model. The cross-sectional flow rate is set as the upstream boundary condition and the flow prediction value is set as the downstream boundary condition. The two-dimensional hydrodynamic model is then initialized. After the two-dimensional hydrodynamic model is initialized, a preliminary simulation is performed using historically calibrated parameters to output key flow variables, including water level and flow velocity. The interval time is set according to the speed of flood evolution, and the real-time observed water level is extracted from the video according to the set time interval; Calculate the residual between the water level output by the two-dimensional hydrodynamic model and the real-time observed water level, and determine whether the residual exceeds a preset threshold. If the residual does not exceed the threshold, the two-dimensional hydrodynamic model continues to run; if the residual exceeds the threshold, the adjoint equation method is used to adjust the roughness coefficient or the particle swarm optimization algorithm is used to dynamically optimize the model parameters.
[0014] Furthermore, the process of running a two-dimensional hydrodynamic model based on optimized model parameters, calculating key flow variables in real time, and determining the flood inundation range includes the following steps: A two-dimensional hydrodynamic model is run using the optimized model parameters to calculate key flow variables in real time, including water level and flow velocity. By combining digital elevation data with water levels, the inundation boundary is delineated using the contour line method or the grid overlay method, and the flood inundation range is determined.
[0015] Furthermore, the process of running the two-dimensional hydrodynamic model based on the optimized model parameters also includes: continuously receiving real-time data from hydrological stations, and if the output value of the two-dimensional hydrodynamic model deviates from the observation value again by exceeding the threshold, continuing to optimize the model parameters.
[0016] Furthermore, the method also includes visualizing the simulation results after determining the flood inundation area, including visually displaying the flood inundation area and the changing trends of key variables; The flood inundation area includes displaying the real-time flood-affected area, highlighting key inundated areas, and showing the changing trends; Key variable trends include dynamically updated curves for water level, flow rate, and flow velocity changes.
[0017] Secondly, this application proposes a hydrodynamic flood simulation system, comprising: The video data acquisition module is used to acquire video data from high-point monitoring, perform geometric correction and preprocessing on the video data, delineate the flood simulation area, and acquire the topographic elevation data of the flood simulation area. The video data extraction module is used to extract river flow element data based on preprocessed video data. The river flow element data includes the river's water level, surface velocity, and cross-sectional flow rate. The two-dimensional hydrodynamic model construction module is used to delineate the simulation boundary based on the flood simulation area, discretize the computational domain into multiple triangular grids, assign different Manning roughness coefficients to different land types, preset model boundary conditions, and construct a two-dimensional hydrodynamic model. The model initialization module is used to combine the extracted water level and topographic elevation data to generate the initial water depth field of the two-dimensional hydrodynamic model, and to determine the downstream boundary conditions of the two-dimensional hydrodynamic model using historical video data. The model optimization module is used to import the initial water depth field, cross-sectional flow rate and downstream boundary conditions into the two-dimensional hydrodynamic model for initialization, compare the water level simulated by the two-dimensional hydrodynamic model with the real-time video observation water level at set intervals, and optimize the model parameters based on the comparison results. The flood simulation module is used to run a two-dimensional hydrodynamic model based on optimized model parameters, calculate key water flow variables in real time, and determine the flood inundation range.
[0018] Thirdly, this application proposes an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, which stores computer programs; The processor, when executing the program stored in the memory, implements the described hydrodynamic flood simulation method.
[0019] Fourthly, this application proposes a computer-readable storage medium storing a computer program, which, when run, executes the described hydrodynamic flood simulation method.
[0020] Compared with the prior art, this application has the following advantages: This application utilizes widely deployed high-point monitoring video resources to replace traditional ground sensors, achieving non-contact monitoring. This avoids the limited coverage and susceptibility to flood damage caused by fixed sensor deployments, and comprehensively captures hydrological information from complex river topography and flood-prone areas. Furthermore, it simultaneously extracts multi-dimensional hydrological parameters such as water level, flow velocity, and flow rate through intelligent video analysis technology, solving the problem of single parameters in existing video analysis techniques and improving the richness and accuracy of parameter acquisition. Based on water level data-DEM elevation calculation, Gaussian process regression interpolation, and deep learning model correction, a high-precision, spatially continuous initial water depth field is generated, avoiding the over-smoothing defects of traditional interpolation methods and ensuring a high degree of matching between the model's initial state and actual hydrological conditions. Using historical video data to drive the construction of a downstream boundary relationship model to obtain predicted flow, replacing the traditional preset water level-flow rate curve, it can adapt to changes in hydraulic characteristics caused by riverbed scouring and deposition, achieving dynamic updates of downstream boundary conditions. This solves the problem of rigid traditional boundary conditions and significantly improves the adaptability of the two-dimensional hydrodynamic model to complex hydrological scenarios.
[0021] This application also uses a dynamic correction mechanism with fixed time intervals to automatically trigger parameter optimization by comparing real-time water level data extracted from real-time video with simulation results, thus forming a closed-loop feedback.
[0022] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description
[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0024] Figure 1 A flowchart of a hydrodynamic flood simulation method according to an embodiment of this application is shown; Figure 2 A schematic diagram of a hydrodynamic flood simulation system according to an embodiment of this application is shown; Figure 3 A schematic diagram of an electronic device according to an embodiment of this application is shown. Detailed Implementation
[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0026] This application proposes a two-dimensional hydrodynamic flood simulation method based on high-point monitoring video. By driving model parameterization and dynamic correction with video data, it achieves high-precision real-time simulation of flood evolution. Figure 1 As shown, the specific technical solution includes the following steps: S1: Acquire video data from high-point monitoring, perform geometric correction and preprocessing on the video data, delineate the flood simulation area, and acquire the topographic elevation data of the flood simulation area; S2: Based on the preprocessed video data, extract river flow element data, which includes the river water level, surface velocity and cross-sectional flow rate. S3: Based on the flood simulation area, the simulation boundary is delineated, the computational domain is discretized into multiple triangular grids, and different Manning roughness coefficients are assigned to different land types. The model boundary conditions are preset, and a two-dimensional hydrodynamic model is constructed. S4: Combine the extracted water level and topographic elevation data to generate the initial water depth field of the two-dimensional hydrodynamic model, and use historical video data to determine the downstream boundary conditions of the two-dimensional hydrodynamic model. S5: Initialize the initial water depth field, cross-sectional flow rate and downstream boundary conditions by importing them into the two-dimensional hydrodynamic model. Compare the water level simulated by the two-dimensional hydrodynamic model with the real-time water level observed in the video at set intervals, and optimize the model parameters based on the comparison results. S6: Run a two-dimensional hydrodynamic model based on the optimized model parameters, calculate key water flow variables in real time, determine the flood inundation range, and visualize the simulation results.
[0027] In step S1, video data from high-point monitoring is acquired. After geometric correction and preprocessing of the video data, the flood simulation area is delineated. This specifically includes the following steps: Ground control points are set up within the field of view of the high-point surveillance camera, and the internal and external parameters of the camera are calculated using the calibration method. The video data from high-point monitoring is acquired using the calibrated internal and external parameters of the camera and decomposed into continuous video frames. Based on the perspective transformation model, the video frames are projected onto the river plane coordinate system for geometric correction. This step can eliminate viewpoint distortion and achieve video geometric correction. By using a semantic segmentation model to identify river surface areas in video frames, a dynamic mask is automatically generated. This mask is then combined with the historical flood inundation range to expand the simulation range to the potential inundation area and delineate the flood simulation area. Real-time video data is uploaded to a central data receiving server using wireless communication technologies (such as 4G and 5G) or wired networks. The data receiving center performs multi-step preprocessing on the video data, including data cleaning, data interpolation, and format conversion, to ensure data integrity and accuracy. Data cleaning removes outliers and noise from the video data; data interpolation fills in missing values using time-series or spatial interpolation methods; and format conversion standardizes multi-source data into a unified input format. Real-time data is the core input for the operation and optimization of the two-dimensional hydrodynamic model, and its integrity and reliability directly determine the quality of subsequent simulations and optimizations.
[0028] In step S2, based on the preprocessed video data, river flow element data is extracted, including the following steps: Within the designated flood simulation area, virtual water level markers are set along the river cross-section. Based on edge detection and Hough transform, the intersections of the water surface line and the markers in the video frame are extracted, and the actual water level value is calculated. An improved sparse optical flow method is employed to track the pixel displacement of floating objects (such as tree branches) on the water surface. This displacement is then converted into actual flow velocity using calibration coefficients to calculate surface velocity. Innovations in feature point selection, tracking strategies, motion compensation, and post-processing address core challenges in water surface scenarios, such as feature instability, complex motion, and numerous interferences. This allows for the stable and accurate extraction of pixel displacement data for surface velocity calculation from ordinary surveillance videos. Alternatively, a deep learning model based on RAFT (Recurrent All-Pairs Field Transforms) can be trained to directly output a dense velocity field, which is then denoised using median filtering to calculate the surface velocity.
[0029] The cross-sectional area of the river is calculated based on the actual water level. The cross-sectional flow rate of the river is then calculated by combining the cross-sectional area and the surface velocity (multiplied by the vertical average coefficient).
[0030] Step S2 involves synchronously acquiring multi-dimensional hydrological parameters such as water level, flow velocity, and flow rate from video data, replacing traditional sensor data. This provides real-time observation data required for the operation and parameter calibration of the two-dimensional hydrodynamic model, enabling dynamic input of real-time hydrological conditions and laying the foundation for subsequent parameter adjustments.
[0031] In step S4, the initial water depth field of the two-dimensional hydrodynamic model is generated by combining the extracted water level and topographic elevation data, including the following steps: The flood simulation area is divided into a triangular grid, and the terrain elevation data corresponding to each grid node is extracted; the grid size is set according to user needs to achieve spatial discretization. For network nodes covered by water levels, the water depth of the network node is calculated using the actual water level value and topographic elevation data of the network node; the water depth of the grid node = the actual water level value of the network node - the topographic elevation data (DEM elevation value) of the network node; if the actual water level value ≤ the DEM elevation value, the water depth of the network node is recorded as 0.
[0032] For grid nodes that are not directly covered by the water level, Gaussian process regression interpolation is used to estimate the water level, and the river channel topographic slope is used as a covariate to constrain the smoothness of the interpolation results. Set a reasonable water depth threshold and screen the water depth values of all grid nodes. If the water depth of a certain network node exceeds the reasonable threshold, use a deep learning model to predict and correct it (based on a model pre-trained from historical flood data) and remove outliers.
[0033] In step S4, the downstream boundary conditions of the two-dimensional hydrodynamic model are determined using historical video data, including the following steps: Following steps S1-S2, extract the water level-flow time series data of the downstream section (i.e., the water level and the corresponding flow at different times, forming a time series dataset). A relational model is constructed, taking water level data, historical upstream and downstream discharge flow, and historical river roughness coefficient from the water level-flow time series data as inputs, the predicted flow value as output, and the mean square error and the physical properties of the hydraulic continuity equation as constraints, and the relational model is trained; the relational model is a common deep learning model.
[0034] The trained relational model is embedded into the downstream boundary of a two-dimensional hydrodynamic model, and the current flow prediction value is used as the downstream boundary condition. The current flow prediction value can replace the preset QH curve (water level-flow curve) to improve the model.
[0035] Step S5 involves initializing the initial water depth field, cross-sectional flow rate, and downstream boundary conditions into a two-dimensional hydrodynamic model. The water level simulated by the two-dimensional hydrodynamic model is compared with the real-time video observation water level at set intervals. Based on the comparison results, the model parameters are optimized, including the following steps: The initial water depth field generated in step S4 is imported into the two-dimensional hydrodynamic model. The upstream boundary condition is set as the cross-sectional flow rate extracted in step S2, and the predicted flow rate is set as the downstream boundary condition. The two-dimensional hydrodynamic model is then initialized. The model initialization process includes parameter setting, boundary condition loading, and error function definition. Parameter setting involves setting initial values, such as the roughness coefficient, bed friction coefficient, channel width, and slope, and defining the range of parameter values. Boundary condition loading involves applying inflow and outflow boundary conditions to the model to ensure that the initial state of the model matches the actual situation. Error function definition involves selecting functions such as mean squared error (MSE) and Nash-Sutcliffe efficiency coefficient (NSE) to quantify the deviation between simulated and measured values.
[0036] After the two-dimensional hydrodynamic model is initialized, a preliminary simulation is performed using historically calibrated parameters to output key flow variables (such as water level and flow velocity) to verify the basic applicability of the model. The interval time is set according to the speed of flood evolution, and the real-time observed water level is extracted from the video stream every few minutes (e.g., 10 minutes); The residual between the output water level of the calculation model and the measured water level is used to determine whether the residual exceeds a preset threshold. If the residual does not exceed the threshold, the two-dimensional hydrodynamic model continues to run; if the residual exceeds the threshold (e.g., ±0.2m), parameter inversion is automatically triggered, the roughness coefficient is adjusted, the adjoint equation method is used to optimize and minimize the residual, the model parameters are updated and the simulation is restarted, forming a closed-loop feedback.
[0037] Specifically, Particle Swarm Optimization (PSO), based on its global optimization capabilities and fast convergence characteristics, can be used to dynamically adjust the model parameters of the two-dimensional hydrodynamic model. The optimization process includes parameter updates, model iteration, and convergence checks. Parameter updates generate new parameter combinations within the allowable parameter range. Model iteration runs the hydrodynamic model using the updated parameters and calculates the error function value. Convergence checks stop optimization if the error is less than a set threshold or the maximum number of iterations is reached; otherwise, iteration continues. The optimized parameters are immediately updated in the model to ensure that the model adapts to the current hydrological conditions and improves simulation accuracy.
[0038] This step involves dynamically optimizing a 2D hydrodynamic model using real-time video data, significantly improving simulation accuracy and timeliness. The step runs the 2D hydrodynamic model based on real-time data, outputting simulated values (such as water level and flow rate). These simulated values are compared with real-time observation data to calculate the error and determine if parameter optimization is necessary. If the error exceeds a threshold, the next parameter optimization process is triggered; otherwise, the model continues running without adjustment. The optimization algorithm adjusts the model parameters to minimize the error between simulated and measured values, outputting optimized parameters to provide a more accurate computational basis for subsequent simulations. This step receives the error results from the real-time simulation and data comparison step as optimization input, outputting optimized parameters to the model to improve the accuracy of flood inundation simulation.
[0039] Step S6: Based on the optimized model parameters, run the two-dimensional hydrodynamic model to calculate key flow variables in real time and determine the flood inundation range, including: The two-dimensional hydrodynamic model is run in combination with the optimized model parameters to calculate key variables of water flow in real time, including water level and flow velocity. By combining digital elevation data with water levels, the inundation boundary is delineated using the contour line method or the grid overlay method, and the flood inundation range is determined.
[0040] This process also includes continuous adjustment of model parameters through dynamic adjustment, anomaly response, and process simulation. Dynamic adjustment involves continuously receiving real-time data from hydrological stations and dynamically updating the model based on new input conditions. Anomaly response restarts the parameter optimization process if the deviation between simulation and observation exceeds a threshold again. Process simulation refines the flood propagation process and captures the hydrodynamic changes in key areas.
[0041] The output of this step can be used to support decision-making, such as flood warning, flood control scheduling, and emergency response. This step uses optimized model parameters to simulate the flood process, outputting key results such as inundation range and water level changes, achieving real-time dynamic simulation of the flood process and providing data support for flood control scheduling and emergency response. This step uses the results of the parameter optimization step for subsequent calculations, continuously receiving real-time data as input. If the error exceeds the threshold again, parameter optimization can be restarted.
[0042] In step S6, the simulation results are visualized, including the flood inundation area and the trends of key variables. The flood inundation area shows the real-time flood-affected area, indicating key inundated areas and their changing trends. The trends of key variables are dynamically updated curves showing changes in water level, flow rate, and flow velocity.
[0043] This step dynamically displays results such as the flood inundation area and water level change trends, providing real-time data support to assist in flood control scheduling, emergency response, and smart water management decisions. This step is the endpoint of the entire process, showcasing the combined results of the previous four steps.
[0044] In summary, steps S1-S2 constitute the data acquisition layer, replacing traditional sensors to achieve non-contact monitoring. Steps S3-S4 constitute the model parameterization layer, transforming video data into the core input of the hydrodynamic model. Steps S5-S6 constitute the dynamic optimization layer, forming a monitoring-simulation-correction closed loop.
[0045] Based on the above method, this application proposes a hydrodynamic flood simulation system, such as... Figure 2 As shown, it includes: The video data acquisition module is used to acquire video data from high-point monitoring, perform geometric correction and preprocessing on the video data, delineate the flood simulation area, and acquire the topographic elevation data of the flood simulation area. The video data extraction module is used to extract river flow element data based on preprocessed video data. The river flow element data includes the river's water level, surface velocity, and cross-sectional flow rate. The two-dimensional hydrodynamic model construction module is used to delineate the simulation boundary based on the flood simulation area, discretize the computational domain into multiple triangular grids, assign different Manning roughness coefficients to different land types, preset model boundary conditions, and construct a two-dimensional hydrodynamic model. The model initialization module is used to combine the extracted water level and topographic elevation data to generate the initial water depth field of the two-dimensional hydrodynamic model, and to determine the downstream boundary conditions of the two-dimensional hydrodynamic model using historical video data. The model optimization module is used to import the initial water depth field, cross-sectional flow rate and downstream boundary conditions into the two-dimensional hydrodynamic model for initialization, compare the water level simulated by the two-dimensional hydrodynamic model with the real-time video observation water level at set intervals, and optimize the model parameters based on the comparison results. The flood simulation module is used to run a two-dimensional hydrodynamic model based on optimized model parameters, calculate key water flow variables in real time, and determine the flood inundation range.
[0046] Another exemplary embodiment of this application provides an electronic device. For example... Figure 3 As shown, the electronic device includes at least one processor 301, at least one communication interface 302, at least one memory 303, and at least one communication bus 304; wherein the processor 301, communication interface 302, and memory 303 communicate with each other through the communication bus 304. Memory 303 stores computer programs; The processor 301 is used to execute the program stored in the memory 303 to implement the hydrodynamic flood simulation method.
[0047] Optionally, the communication interface can be an interface of a communication module, such as the interface of a GSM module; the processor may be a CPU, an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The memory may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device. The memory stores a program, and the processor calls the program stored in the memory to execute some or all of the above-described method embodiments.
[0048] Based on the same inventive concept, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed, implements some or all of the above-described method embodiments. Optionally, the storage medium may be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage battery device, etc.
[0049] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A method for simulating floods using hydrodynamics, characterized in that, Includes the following steps: Acquire video data from high-point monitoring, perform geometric correction and preprocessing on the video data, delineate the flood simulation area, and acquire the topographic elevation data of the flood simulation area; Based on the preprocessed video data, river flow element data is extracted, including the river water level, surface velocity and cross-sectional flow rate. Based on the flood simulation area, the simulation boundary is delineated, the computational domain is discretized into multiple triangular grids, and different Manning roughness coefficients are assigned to different land types. The model boundary conditions are preset, and a two-dimensional hydrodynamic model is constructed. By combining the extracted water level and topographic elevation data, the initial water depth field of the two-dimensional hydrodynamic model is generated, and the downstream boundary conditions of the two-dimensional hydrodynamic model are determined using historical video data. The initial water depth field, cross-sectional flow rate, and downstream boundary conditions are imported into a two-dimensional hydrodynamic model for initialization. The water level simulated by the two-dimensional hydrodynamic model is compared with the real-time water level observed in the video at set intervals, and the model parameters are optimized based on the comparison results. The two-dimensional hydrodynamic model is run based on the optimized model parameters, and the flood inundation range is determined after real-time calculation of key water flow variables.
2. The hydrodynamic flood simulation method according to claim 1, characterized in that, Acquiring video data from high-altitude surveillance, and then delineating the flood simulation area after geometric correction and preprocessing of the video data, includes the following steps: Ground control points are set up within the monitoring range of the high-point surveillance cameras, and the internal and external parameters of the cameras are calculated using the calibration method. The video data of high-point monitoring is obtained by using the calibrated internal and external parameters of the camera and decomposed into continuous video frames. Based on the perspective transformation model, the video frames are projected onto the river plane coordinate system for geometric correction. By identifying river surface areas in video frames using a semantic segmentation model, a dynamic mask is generated. This mask is then combined with the historical flood inundation range to expand the simulation range to the potential inundation area and delineate the flood simulation area. The geometrically corrected video data is preprocessed, including data cleaning, data interpolation, and format conversion.
3. The hydrodynamic flood simulation method according to claim 2, characterized in that, The data cleaning includes removing outliers and noise from the video data; The data interpolation includes filling in missing values in video data using time series or spatial interpolation methods; The format conversion includes standardizing the video data into a uniform input format.
4. The hydrodynamic flood simulation method according to claim 1, characterized in that, The extraction of river flow element data based on preprocessed video data includes the following steps: Within the designated flood simulation area, virtual water level markers are set along the river cross-section. Based on edge detection and Hough transform, the intersections of the water surface line and the markers in the video frame are extracted, and the actual water level value is calculated. The pixel displacement of floating objects on the water surface is tracked using the sparse optical flow method or the surface velocity is calculated based on the RAFT deep learning model. The cross-sectional area of the river is calculated based on the actual water level, and the cross-sectional flow rate of the river is calculated by combining the cross-sectional area and the surface velocity.
5. The method for simulating floods using hydrodynamics according to claim 1, characterized in that, The process of generating the initial water depth field for a two-dimensional hydrodynamic model by combining the extracted water level and topographic elevation data includes the following steps: The flood simulation area was divided into a triangular grid, and the terrain elevation data corresponding to each grid node was extracted. For network nodes covered by water levels, the water depth of the network node is calculated using the actual water level value and topographic elevation data of the network node; For network nodes that are not directly covered by water level, Gaussian process regression interpolation is used to estimate water level, and the interpolation smoothness is constrained by the river channel topographic slope. Set a reasonable water depth threshold and screen the water depth of all grid nodes. If the water depth of a certain network node exceeds the reasonable threshold, call a deep learning model pre-trained based on historical flood data to correct it.
6. The hydrodynamic flood simulation method according to claim 1, characterized in that, The method of determining the downstream boundary conditions of a two-dimensional hydrodynamic model using historical video data includes the following steps: Collect and preprocess historical video data corresponding to the flood simulation area, and extract water level-flow time series data of the downstream section within the historical video period; A relational model is constructed, with water level data, historical upstream and downstream discharge, and historical river roughness coefficient from the water level-discharge time series data as inputs to the relational model, and the predicted discharge value as the output of the relational model. The relational model is trained with mean square error and the physical of the hydraulic continuity equation as constraints. The trained relational model is embedded into the downstream boundary of the two-dimensional hydrodynamic model, and the current flow prediction value is used as the downstream boundary condition by the relational model.
7. The hydrodynamic flood simulation method according to claim 6, characterized in that, The initial water depth field, cross-sectional flow rate, and downstream boundary conditions are imported into a two-dimensional hydrodynamic model to complete the initialization. The water level simulated by the two-dimensional hydrodynamic model is compared with the real-time video observation water level at set intervals, and the model parameters are optimized based on the comparison results, including the following steps: The initial water depth field is imported into the two-dimensional hydrodynamic model. The cross-sectional flow rate is set as the upstream boundary condition and the flow prediction value is set as the downstream boundary condition. The two-dimensional hydrodynamic model is then initialized. After the two-dimensional hydrodynamic model is initialized, a preliminary simulation is performed using historically calibrated parameters to output key flow variables, including water level and flow velocity. The interval time is set according to the speed of flood evolution, and the real-time observed water level is extracted from the video according to the set time interval; Calculate the residual between the water level output by the two-dimensional hydrodynamic model and the real-time observed water level, and determine whether the residual exceeds a preset threshold. If the residual does not exceed the threshold, the two-dimensional hydrodynamic model continues to run; if the residual exceeds the threshold, the adjoint equation method is used to adjust the roughness coefficient or the particle swarm optimization algorithm is used to dynamically optimize the model parameters.
8. The method for simulating floods using hydrodynamics according to claim 1, characterized in that, The process of running a two-dimensional hydrodynamic model based on optimized model parameters, calculating key flow variables in real time, and determining the flood inundation range includes the following steps: A two-dimensional hydrodynamic model is run using the optimized model parameters to calculate key flow variables in real time, including water level and flow velocity. By combining digital elevation data with water levels, the inundation boundary is delineated using the contour line method or the grid overlay method, and the flood inundation range is determined.
9. The hydrodynamic flood simulation method according to claim 8, characterized in that, The process of running the two-dimensional hydrodynamic model based on the optimized model parameters also includes: continuously receiving real-time data from hydrological stations; if the output value of the two-dimensional hydrodynamic model deviates from the observation value again by exceeding the threshold, the model parameters are further optimized.
10. The method for simulating floods using hydrodynamics according to any one of claims 1-9, characterized in that, The method also includes visualizing the simulation results after determining the flood inundation area, including visually displaying the flood inundation area and the changing trends of key variables; The flood inundation area includes displaying the real-time flood-affected area, highlighting key inundated areas, and showing the changing trends; Key variable trends include dynamically updated curves for water level, flow rate, and flow velocity changes.
11. A hydrodynamic flood simulation system, characterized in that, include: The video data acquisition module is used to acquire video data from high-point monitoring, perform geometric correction and preprocessing on the video data, delineate the flood simulation area, and acquire the topographic elevation data of the flood simulation area. The video data extraction module is used to extract river flow element data based on preprocessed video data. The river flow element data includes the river's water level, surface velocity, and cross-sectional flow rate. The two-dimensional hydrodynamic model construction module is used to delineate the simulation boundary based on the flood simulation area, discretize the computational domain into multiple triangular grids, assign different Manning roughness coefficients to different land types, preset model boundary conditions, and construct a two-dimensional hydrodynamic model. The model initialization module is used to combine the extracted water level and topographic elevation data to generate the initial water depth field of the two-dimensional hydrodynamic model, and to determine the downstream boundary conditions of the two-dimensional hydrodynamic model using historical video data. The model optimization module is used to import the initial water depth field, cross-sectional flow rate and downstream boundary conditions into the two-dimensional hydrodynamic model for initialization, compare the water level simulated by the two-dimensional hydrodynamic model with the real-time video observation water level at set intervals, and optimize the model parameters based on the comparison results. The flood simulation module is used to run a two-dimensional hydrodynamic model based on optimized model parameters, calculate key water flow variables in real time, and determine the flood inundation range.
12. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, which stores computer programs; A processor, when executing a program stored in a memory, implements the hydrodynamic flood simulation method according to any one of claims 1-10.
13. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is run, it performs the hydrodynamic flood simulation method as described in any one of claims 1-10.