A multi-model fusion flood evolution simulation system

Abstract

This invention relates to the field of flood evolution simulation technology, and more particularly to a multi-model fusion flood evolution simulation system, comprising: a model building module; a range determination module for determining the baseline matching setting or direction matching setting for a target viewport based on the linkage viewport fit and the number of linkage viewports; a loading analysis module for determining whether to adjust the loading method of the target viewport's pre-loading range from direct loading to time-domain layered loading based on the network degradation coefficient and viewport feature anomaly degree; a time-domain layering module for determining whether to perform uniform layering based on anomaly evaluation values ​​or clustering layering based on evolution correlation degree in time-domain layered loading based on evolution law coefficients and tile evolution intensity; and an optimization adjustment module for determining whether to perform optimization adjustment based on loading level heterogeneity and multi-viewport loading interference degree. This invention can optimize scene loading strategies to improve the scene loading construction effect of flood evolution digital twins.

Description

Technical Field

[0001] This invention relates to the field of flood evolution simulation technology, and in particular to a flood evolution simulation system that integrates multiple models. Background Technology

[0002] In the fields of flood control and disaster reduction, and watershed water resources management, flood evolution simulation is a core technology for predicting flood paths, extent, and severity, directly impacting the scientific rigor and timeliness of emergency decision-making. To intuitively present the flood evolution process and support precise decision-making, digital twin scenarios of flood evolution are typically used to visualize and simulate the dynamic evolution of floods. However, due to the massive amounts of data contained in these scenarios and the complex and variable network environments in real-world applications, problems such as excessively long scenario loading and construction times and unstable presentation effects arise. Therefore, optimizing scenario loading strategies to improve the loading and construction effectiveness of digital twin scenarios of flood evolution is a technical problem that urgently needs to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN121170204A discloses a system and method for dynamic construction and scheduling of digital twin water conservancy scenes, including: a data processing module for obtaining standard geospatial data services and standard 3D model data services; a preliminary scene construction module for obtaining a preliminary scene based on the standard geospatial data services and the standard 3D model data services in the scene template; a scene rendering scheduling module for obtaining triangular mesh data from vector data through an asynchronous computing thread, and then pushing the obtained triangular mesh data frame by frame to the rendering thread through the main thread to perform rendering operations to obtain a basic scene; and a scene editing and storage module for editing and structured storage of the basic scene. However, the above solution has the following problems: the rendering method is singular, it cannot adaptively load by combining the actual viewport state and the dynamic characteristics of the scene, and it is difficult to meet the requirements of real-time and accuracy rendering of flood evolution digital twin scenes, resulting in poor scene loading efficiency. Summary of the Invention

[0004] To address this, the present invention provides a multi-model fusion flood evolution simulation system to overcome the problems of existing technologies, such as the single rendering method, the inability to adaptively load by combining the actual viewport state and scene dynamic features, the difficulty in meeting the real-time and accuracy requirements of flood evolution digital twin scenarios, and the resulting poor scene loading efficiency.

[0005] To achieve the above objectives, the present invention provides a multi-model fusion flood evolution simulation system, comprising: The model building module is used to obtain flood spatiotemporal distribution data based on real-time meteorological data and watershed basic data, and through LSTM neural network precipitation prediction model, water flow propagation model and flood evolution model, and to build a digital twin model based on flood spatiotemporal distribution data and watershed basic data. The range determination module, which is connected to the model construction module, is used to determine the baseline matching setting or direction matching setting for the target viewport to obtain the preload range of the target viewport based on the linkage viewport fit degree and the number of linkage viewports. The loading analysis module, which is connected to the model building module and the range determination module respectively, is used to determine whether to adjust the loading method of the pre-loading range of the target viewport from direct loading to temporal layered loading based on the network degradation coefficient and the viewport feature anomaly degree; wherein, the viewport feature anomaly degree is determined based on the loading gradient change value and the watershed evolution distortion degree; The temporal layering module is connected to the range determination module and the loading analysis module respectively. It is used to determine, in the temporal layering loading, whether to perform uniform layering based on anomaly evaluation values ​​or clustering layering based on evolution correlation degrees to obtain several loading levels based on evolution law coefficients and tile evolution intensity, and to determine the loading priority coefficient of each loading level based on interval reference values. An optimization adjustment module, which is connected to the range determination module and the temporal layering module respectively, is used to determine whether to perform optimization adjustment based on the heterogeneity of the loading level and the multi-viewport loading interference. In the optimization adjustment, the pre-loading range is adjusted based on the comprehensive evaluation value, and the rendering reference value is adjusted based on the abnormal comparison value according to the feature tile change coefficient.

[0006] Furthermore, the range determination module performs a benchmark matching setting for target viewports where the linkage viewport fit degree is less than the preset linkage viewport fit degree or the number of linkage viewports is less than the preset number of linkage viewports. In the baseline matching settings, each tile in a circular area with the coordinates of the center point of the target viewport as the center and a preset length as the radius is used as the preload range of the target viewport.

[0007] Furthermore, the range determination module performs direction matching settings for target viewports whose linkage viewport matching degree is greater than or equal to a preset linkage viewport matching degree and whose number of linkage viewports is greater than or equal to a preset number of linkage viewports. In the orientation matching settings, the coordinates of the center point of the target viewport's screen are used as the fan vertex, the direction of the two reference viewport vectors is used as the direction of the fan axis, and the tiles in the fan constructed based on the length of the fan axis determined by the length threshold and the tiles in the rectangular area corresponding to the target viewport are used together as the preload range of the target viewport. The reference viewport vector starts from the coordinate point of the center point of the target viewport, and the angle between the reference viewport vector and the viewport vector of the left neighboring viewport of the target viewport is the angle threshold.

[0008] Furthermore, the loading analysis module adjusts the loading method from direct loading to temporal layered loading for the pre-loading range of target viewports where the network degradation coefficient is greater than or equal to the preset network degradation coefficient or the viewport feature anomaly degree is greater than or equal to the preset viewport feature anomaly degree.

[0009] Furthermore, the time-domain layering module responds to the condition that the evolution law coefficient is greater than or equal to the preset evolution law coefficient and the tile evolution intensity is less than the preset tile evolution intensity, and then performs uniform layering based on the abnormal evaluation value.

[0010] Furthermore, if the temporal hierarchical module responds to an evolution law coefficient that is less than a preset evolution law coefficient or the tile evolution intensity is greater than or equal to a preset tile evolution intensity, clustering and hierarchical processing is performed based on the evolutionary correlation degree.

[0011] Furthermore, the optimization adjustment module determines to perform optimization adjustment when the loading level heterogeneity is greater than the preset loading level heterogeneity or the multi-viewport loading interference is greater than the preset multi-viewport loading interference.

[0012] Furthermore, the optimization adjustment module reduces the preload range based on the comprehensive evaluation value; The overall evaluation value is positively correlated with both the heterogeneity of the loading level and the multi-viewport loading interference.

[0013] Furthermore, the optimization adjustment module responds to the condition that the feature tile change coefficient is greater than the preset feature tile change coefficient, and determines to reduce the rendering reference value based on the abnormal comparison value; The decrease in the rendering reference value is positively correlated with the anomaly comparison value.

[0014] Furthermore, the optimization adjustment module determines the feature tiles based on the pixel change index and feature gradient.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: In the technical solution of the present invention, the consistency of linkage matching, the continuity of viewport operation, and the sufficiency of linkage loading between the target viewport and each linkage viewport are effectively reflected by the linkage viewport fit degree and the number of linkage viewports. Then, based on the linkage viewport fit degree and the number of linkage viewports, the reference matching setting or direction matching setting is adaptively selected, which is conducive to accurately defining the preloading range of the target viewport, avoiding resource waste caused by an excessively large preloading range, improving the accuracy and smoothness of viewport loading in the watershed digital twin model, and ensuring the continuity of the visualization display of the flood evolution simulation scene.

[0016] Furthermore, in this invention, when the network degradation coefficient or viewport feature anomaly is large, it indicates that the current network transmission status has significantly deteriorated, the data presentation rate cannot meet the real-time loading requirements, or the data loading pressure in the target viewport pre-loading area is too high, or there are abnormalities in the evolution of watershed parameters. If the direct loading method is continued, it is easy to cause problems such as data loading delay, rendering stutter, and incomplete scene presentation. Adjusting the loading method from direct loading to time-domain layered loading is beneficial to perform layered scheduling and orderly loading of data in the pre-loading range according to data evolution characteristics and loading priority, avoiding the load pressure on the network caused by loading a large amount of data at once, and avoiding loading instability problems caused by abnormal loading gradient and evolution distortion. This improves the smoothness, stability, and real-time performance of viewport loading in the watershed digital twin model, ensuring the accurate presentation of the flood evolution simulation scene.

[0017] Furthermore, this invention effectively reflects the stability of the interval evolution curve, the drastic changes of characteristic time-domain tiles in the time dimension, and the evolutionary stability of watershed parameters during flood evolution by using evolution law coefficients and tile evolution intensity. This allows for the adaptive selection of uniform stratification based on anomaly assessment values ​​or clustering stratification based on evolutionary correlation, which is beneficial for combining the degree of anomaly in the loading environment with data evolution characteristics to achieve accurate adaptation of time-domain stratification. This avoids loading redundancy or lag in key data loading caused by mismatch between stratification methods and data evolution trends, thereby optimizing the rationality and efficiency of time-domain stratified loading, improving the orderliness of data loading at each loading level, and ensuring the smoothness and real-time performance of viewport loading in the watershed digital twin model.

[0018] Furthermore, in this invention, when the heterogeneity of the loading level or the interference of multi-viewport loading is high, it indicates that there is obvious heterogeneity anomaly in the current loading level, the loading gradient switching between levels is unstable, or the mutual interference generated by parallel loading of multiple viewports has exceeded the reasonable range. If no optimization adjustment is made, it is easy to cause data loading lag and discontinuous transition of loading levels. Adjusting the preloading range based on the comprehensive evaluation value is beneficial to accurately match the comprehensive anomaly degree of the loading environment, reasonably reduce the amount of preloaded data, reduce loading pressure and multi-viewport loading interference, and avoid the instability problems caused by heterogeneity of loading levels. When the feature tile change coefficient is large, it indicates that after the preloading range is adjusted, the distribution density of feature tiles in the unpresented area has significantly increased compared with before the adjustment. If the original rendering reference value is maintained, it is easy to cause excessive rendering load and decreased rendering smoothness. Adjusting the rendering reference value based on the anomaly comparison value is beneficial to adapt to the change range of feature tile distribution density, reasonably reduce the rendering load, and improve the stability and smoothness of viewport loading and rendering of the watershed digital twin model. Attached Figure Description

[0019] Figure 1This is a module connection diagram of the flood evolution simulation system with multi-model fusion according to the present invention; Figure 2 This is a flowchart illustrating the process of determining the baseline matching setting or direction matching setting for the target viewport based on the linkage viewport fit and the number of linkage viewports in this invention. Figure 3 This is a flowchart illustrating the process of determining whether to perform uniform stratification based on anomaly evaluation values ​​or clustering stratification based on evolutionary correlation degrees according to the evolutionary law coefficient and the intensity of tile evolution in this invention. Figure 4 This is a flowchart illustrating the present invention for determining whether to adjust the rendering reference value based on the abnormal comparison value according to the feature tile change coefficient. Detailed Implementation

[0020] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0021] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0022] Please see Figures 1 to 4 As shown, the present invention provides a flood evolution simulation system with multi-model fusion, comprising: The model building module is used to obtain flood spatiotemporal distribution data based on real-time meteorological data and watershed basic data, and through LSTM neural network precipitation prediction model, water flow propagation model and flood evolution model, and to build a digital twin model based on flood spatiotemporal distribution data and watershed basic data. The range determination module, which is connected to the model construction module, is used to determine the baseline matching setting or direction matching setting for the target viewport to obtain the preload range of the target viewport based on the linkage viewport fit degree and the number of linkage viewports. The loading analysis module, which is connected to the model building module and the range determination module respectively, is used to determine whether to adjust the loading method of the pre-loading range of the target viewport from direct loading to temporal layered loading based on the network degradation coefficient and the viewport feature anomaly degree; wherein, the viewport feature anomaly degree is determined based on the loading gradient change value and the watershed evolution distortion degree; The temporal layering module is connected to the range determination module and the loading analysis module respectively. It is used to determine, in the temporal layering loading, whether to perform uniform layering based on anomaly evaluation values ​​or clustering layering based on evolution correlation degrees to obtain several loading levels based on evolution law coefficients and tile evolution intensity, and to determine the loading priority coefficient of each loading level based on interval reference values. An optimization adjustment module, which is connected to the range determination module and the temporal layering module respectively, is used to determine whether to perform optimization adjustment based on the heterogeneity of the loading level and the multi-viewport loading interference. In the optimization adjustment, the pre-loading range is adjusted based on the comprehensive evaluation value, and the rendering reference value is adjusted based on the abnormal comparison value according to the feature tile change coefficient.

[0023] The application scenario of this invention is the real-time visualization and adaptive loading rendering of digital twin scenarios of flood evolution, in order to solve the problems of high data density, rapid temporal changes, easy lag in multi-viewport interactive browsing, and insufficient loading stability under network fluctuations.

[0024] Based on real-time meteorological data and watershed baseline data, and using LSTM neural network precipitation prediction models, flow propagation models, and flood evolution models, spatiotemporal distribution data of floods are obtained, including: Real-time meteorological data is input into the LSTM neural network precipitation prediction model to output precipitation prediction parameters; The precipitation forecast parameters and basic watershed data are input into the flow propulsion model to output the dynamic change parameters of the river hydraulic elements. The flood evolution model is based on precipitation prediction parameters, watershed basic data, and dynamic changes in river hydraulic elements to output flood spatiotemporal distribution data.

[0025] The target basin is the river and surrounding area where flood evolution simulation is conducted; Real-time meteorological data refers to meteorological parameters collected in real time within the target watershed, including rainfall, temperature, humidity, and air pressure. Basic watershed data includes river geometry, distribution and structural parameters of structures along the river, DEM / DOM topographic data, administrative maps, satellite imagery, DLG digital line maps, hydraulic engineering contour data, and initial gate parameters. River geometry includes, but is not limited to, the cross-sectional dimensions, direction, curvature, and slope of the river. Distribution and structural parameters of structures along the river include the location, scale, structural strength, and flood control level of bridges, dams, houses, municipal facilities, and other structures within the watershed. Hydraulic engineering contour data includes spatial data such as the three-dimensional contours, planar locations, and project scope of hydraulic engineering projects such as reservoirs, sluices, pumping stations, and dikes. Initial gate parameters include the initial opening degree, closing status, and control range of each sluice and gate within the watershed. This is content easily understood by those skilled in the art and will not be elaborated upon further. The precipitation prediction parameters are the precipitation probability of each tile in the target watershed at each evolution time point within a preset time period, and the precipitation amount corresponding to each evolution time point of each tile in the target watershed within a preset time period; the preset time period is 6 hours long. The dynamic change parameters of the river hydraulic elements are the flow rate through the sluice gates of the structures along the river, the gate control parameters, and the real-time water level and cross-sectional flow rate of each section in the target basin at each evolution time point within the preset time period; the flow rate through the sluice gates of the structures along the river is the flow rate through the outlets of the sluice gates, dams and other hydraulic structures in the basin at each evolution time point within the preset time period; the gate control parameters are the recommended opening degree and control rate of each gate in the basin at each evolution time point within the preset time period. This is content that is easy for those skilled in the art to understand, and will not be elaborated on in detail. The spatiotemporal distribution data of the flood includes the flood level, flow velocity, discharge, inundation depth, and advance velocity corresponding to each evolution time point of each tile in the target watershed within a preset time period. The inundation depth is the difference between the flood level and the ground elevation of the area, and the advance velocity is the spatial advance velocity of the flood front in the target watershed. The geographic projection space of the target watershed is divided into several tiles, with each tile corresponding to a spatial size of 1m × 1m. The geographic projection space is the conversion of the real three-dimensional geographic space of the target watershed into a two-dimensional geographic space in a planar coordinate system using Mercator projection. Several evolution time points are set within the preset time period, with an evolution time point set every 1 second starting from the beginning of the preset time period. The time interval between two adjacent evolution time points can be adaptively adjusted according to the simulation accuracy and hardware performance.

[0026] The establishment of the LSTM neural network precipitation prediction model includes: data preparation: collecting historical meteorological forecast data within the watershed, with a time span of no less than 10 years; data preprocessing: performing preprocessing operations such as data cleaning, format conversion, and normalization to adapt to the input requirements of the LSTM model; model training: training the LSTM model using historical data, adjusting model parameters, and improving prediction accuracy; and model validation: using a portion of historical data as a validation set to evaluate the model's prediction performance and robustness.

[0027] The establishment of the water flow propulsion model includes: data import and initialization: importing the river geometry, distribution and structural parameters of along-line structures, and initial state parameters of gates from the basic watershed data to complete the initial configuration of the basic model data; mathematical generalization of physical objects: segmenting the river according to hydrological characteristics, and generalizing the nodes of along-line structures into virtual channel segments according to their physical functions, or directly calculating the gate flow rate using the gate flow rate formula; selection and solution of core equations: selecting the one-dimensional Saint-Venant equations as the basic equations describing water flow motion, and using the PreissWann four-point eccentric implicit scheme in the finite difference method to numerically solve the one-dimensional Saint-Venant equations; model verification and optimization: using historical hydrological monitoring data of the watershed to verify the accuracy of the model, comparing the deviation between the simulation results and the actual monitoring data, iteratively adjusting the model parameters, improving the model's calculation accuracy and robustness, and completing the establishment of the water flow propulsion model.

[0028] The establishment of the flood evolution model includes: Modeling parameter integration: collecting and integrating precipitation prediction parameters output by the LSTM neural network precipitation prediction model, dynamic change parameters of river hydraulic elements output by the flow propagation model, and basic data of the entire watershed, determining the initial boundary conditions and core calculation parameters for model numerical calculation; Hydrodynamic model construction: based on the hydrogeological and topographical characteristics of the target watershed and the integrated modeling parameters, constructing a river hydrodynamic model specific to the target watershed, and using numerical calculation methods adapted to the flow propagation model to simulate the propagation process of floodwater in the target watershed channel, calculating the hydrodynamic parameters of each river within the target watershed. The model generates flood hydrodynamic parameters for each watershed at various evolution time points within a preset time period; it generates spatiotemporal data by compiling and outputting spatiotemporal distribution data of floods in the target basin based on the calculation results of the river hydrodynamic model. The spatiotemporal distribution data includes the flood level, flow velocity, flow rate, inundation depth, and propulsion speed corresponding to each watershed at various evolution time points within the preset time period; it verifies and optimizes the model by using historical flood monitoring data of the target basin to verify the accuracy of the model, compares the deviation between the simulation results and the actual monitoring data, iteratively adjusts the model parameters, improves the model's calculation accuracy and robustness, and completes the establishment of the flood evolution model.

[0029] The construction of a digital twin model based on flood spatiotemporal distribution data and watershed basic data includes: spatial tile mapping: converting the real three-dimensional geographic space of the target watershed into a two-dimensional geographic projection space in a plane coordinate system using Mercator projection, and dividing the geographic projection space into several tile units according to a preset spatial size to form a unified tiled geographic space base; multi-source data fusion and registration: spatially registering and gridding the river geometry, distribution and structural parameters of buildings along the river, DEM / DOM topographic data, administrative maps, satellite imagery, DLG digital line maps, water conservancy project outline data, and gate initial state parameters from the watershed basic data according to tile units, and establishing a basic geographic attribute library corresponding to each tile unit. Flood spatiotemporal data mounting: The spatiotemporal distribution data of floods output by the flood evolution model is mounted sequentially according to tile units and evolution time points, so that each tile unit has dynamic hydrodynamic parameters such as flood level, flow velocity, flow rate, inundation depth, and propulsion speed at each evolution time point within a preset time period; Digital twin scene instantiation: Using the tile unit as the smallest rendering and calculation unit, a digital twin model that is real-time mapped and highly faithfully replicated with the physical space of the target watershed is constructed by integrating the basic geographic attribute library and dynamic hydrodynamic parameters. This enables the digital twin model to output the corresponding tile visualization image and dynamic hydrological parameters at each evolution time point, realizing full-element, full-time, and high-precision digital twin mapping of the flood evolution process in the target watershed.

[0030] Once the digital twin model is constructed, it can output visualized images of the tiles corresponding to each evolution time point within a preset time period; the tile data corresponding to a single tile includes images of each tile corresponding to each evolution time point within a preset time period, as well as the flood level, flow rate, flow volume, inundation depth, and advance speed corresponding to that tile at each evolution time point within the preset time period.

[0031] Specifically, the range determination module performs a baseline matching setting for target viewports whose linkage viewport fitting degree is less than the preset linkage viewport fitting degree or whose number of linkage viewports is less than the preset number of linkage viewports. In the baseline matching settings, each tile in a circular area with the coordinates of the center point of the target viewport as the center and a preset length as the radius is used as the preload range of the target viewport.

[0032] Specifically, the viewport is the display area of ​​the user terminal interface used to display the visualization image of the digital twin model, and the spatial display range can be adjusted in real time with translation, scaling, and rotation operations; the target viewport is the display area that needs to be loaded and rendered, which is a rectangular area.

[0033] The initial time at which the target viewport in the digital twin model needs to be displayed is denoted as the target time. The linked viewports corresponding to the target viewport are the various viewports loaded within the [reference time, target time]. It should be noted that if a time earlier than the target time by a preset duration is earlier than or equal to the initial time when the digital twin model is completed, then the reference time is the initial time when the digital twin model is completed; if a time earlier than the target time by a preset duration is later than the initial time when the digital twin model is completed, then the reference time is the time earlier than the target time by a preset duration. The greater the accuracy requirement for timely loading of the digital twin viewport, the smaller the value of the preset duration. One preset duration value is provided, which is 5 minutes. The number of linked viewports is the total number of viewports that are loaded within the [reference time, target time] period; The method for confirming the alignment of linked viewports is as follows: Linked viewports are sorted in ascending order of display time and recorded as a reference sequence. The linked viewport with the earliest display time is designated as the target linked viewport, and the other linked viewports are designated as reference linked viewports. Each reference linked viewport corresponds to a left neighboring viewport. The left neighboring viewport for a single reference linked viewport is the linked viewport adjacent to and preceding it in the reference sequence. Each reference linked viewport corresponds to a viewport vector. The viewport vector for a single reference linked viewport is a directed line segment originating from the center point of the screen of its corresponding left neighboring viewport and ending at the center point of the screen of the reference linked viewport itself. The angular alignment of a single reference linked viewport is calculated as 1 - the angle of alignment of the reference linked viewport. =The angle between the viewport vector of the linked viewport and the viewport vector of the left neighboring viewport corresponding to the reference linked viewport / standard angle, where the standard angle is 360°. The distance fit degree corresponding to a single reference linked viewport = max(x0, 0), x0 = 1 - the actual geographical length of the target watershed mapped by the line connecting the coordinate point of the center point of the screen of the reference linked viewport and the coordinate point of the center point of the screen of the left neighboring viewport corresponding to the reference linked viewport / length threshold. The length threshold is the average value of the actual geographical length of the target watershed mapped by the diagonals of each linked viewport. Linked viewport fit degree = minimum value of the angle fit degree corresponding to each reference linked viewport × first weight coefficient + minimum value of the distance fit degree corresponding to each reference linked viewport × second weight coefficient. The first weight coefficient and the second weight coefficient are both 0.5. The preset length is calculated as follows: preset length = average display length of each linked viewport × first proportional coefficient + display length of the target viewport. In this embodiment, the first proportional coefficient is 0.3. The display length of a single viewport is half of the actual geographical length of the target watershed mapped by the diagonal of that viewport.

[0034] The values ​​of preset linkage viewport fit and preset linkage viewport number are adaptively determined based on the scale of the watershed simulation, data density, and interaction smoothness requirements. The linkage viewport fit and the number of linkage viewports effectively reflect the linkage matching and sufficiency of linkage loading between the target viewport and the linkage viewports. The greater the requirement for the viewport linkage display effect and preloading accuracy in the flood evolution simulation scenario, the larger the values ​​of preset linkage viewport fit and preset linkage viewport number. In this embodiment, the preset linkage viewport fit is 0.8 and the preset linkage viewport number is 10. Each tile in the circular area is a tile that spatially intersects with that circular area.

[0035] Specifically, the range determination module performs direction matching settings for target viewports whose linkage viewport matching degree is greater than or equal to the preset linkage viewport matching degree and whose number of linkage viewports is greater than or equal to the preset number of linkage viewports. In the orientation matching settings, the coordinates of the center point of the target viewport's screen are used as the fan vertex, the direction of the two reference viewport vectors is used as the direction of the fan axis, and the tiles in the fan constructed based on the length of the fan axis determined by the length threshold and the tiles in the rectangular area corresponding to the target viewport are used together as the preload range of the target viewport. The reference viewport vector starts from the coordinate point of the center point of the target viewport, and the angle between the reference viewport vector and the viewport vector of the left neighboring viewport of the target viewport is the angle threshold.

[0036] Specifically, the vertex of the sector is the intersection of two axes, and the direction of the sector axis is the direction of the two rays extending outward from the vertex; the rectangular area corresponding to the target viewport is the visualization range of the target watershed in the target viewport; the included angle threshold is the average value of the reference angles corresponding to each reference linked viewport; the reference angle corresponding to a single reference linked viewport is the angle between the viewport vector of the reference linked viewport and the viewport vector of the left neighboring viewport corresponding to the reference linked viewport; the length of the sector axis = length threshold × second scaling factor, and the second scaling factor is 1.0.

[0037] Each tile in the sector and each tile in the rectangular region corresponding to the target viewport are respectively the tiles that have spatial intersection with the sector region and the tiles that have spatial intersection with the rectangular region corresponding to the target viewport.

[0038] It should be noted that the viewport vector corresponding to the left neighboring viewport of the target viewport passes through the aforementioned sector.

[0039] Specifically, the loading analysis module adjusts the loading method from direct loading to temporal layered loading for the pre-loading range of target viewports where the network degradation coefficient is greater than or equal to the preset network degradation coefficient or the viewport feature anomaly degree is greater than or equal to the preset viewport feature anomaly degree.

[0040] Specifically, the loading analysis module does not need to change the loading method from direct loading to temporal layered loading for the preloading range of target viewports where the network degradation coefficient is less than the preset network degradation coefficient and the viewport feature anomaly degree is less than the preset viewport feature anomaly degree.

[0041] Understandably, after constructing a digital twin model based on flood spatiotemporal distribution data and basin basic data, it is necessary to load the tile data corresponding to each tile onto the front-end display terminal, which can be displayed in the viewport. The front-end display terminal can be a computer, tablet computer, or industrial control display device.

[0042] The network degradation coefficient is the average of the degradation reference values ​​corresponding to each time period within the [reference time, target time]. The greater the demand for network monitoring accuracy, the smaller the interval between two time points. A method for setting time points is provided, which sets an interval point every 1 second starting from the reference time. The starting point and each interval point are recorded as time points. The time range between two adjacent time points is recorded as a time period. The degradation reference value corresponding to a single time period is 1 - the ratio of the presentation rate corresponding to the time period to the presentation rate threshold. The presentation rate corresponding to a single time period is the ratio of the amount of tile data corresponding to the tile that starts loading and is presented to the front-end display terminal within the time period to the time length of the time period, in GB / s. The presentation rate threshold is 0.5 GB / s. Viewport feature anomaly degree = Loading gradient change value / Preset loading gradient change value × First weighting coefficient + Watershed evolution distortion degree / Preset watershed evolution distortion degree × Second weighting coefficient, where both the first and second weighting coefficients are 0.5; The loading gradient change value is the ratio of the data volume of the tile data corresponding to the un-rendered area in the pre-loaded range of the target viewport to a data volume threshold of 5GB. The un-rendered area refers to the tiles whose tile data has not yet been loaded onto the front-end display terminal. It is understood that since different viewports each have their own pre-loaded areas, and these pre-loaded areas overlap in spatial range, there may be overlapping parts between the pre-loaded areas of different viewports.

[0043] The watershed evolution distortion degree is the standard deviation of the interval comparison degree corresponding to each evolution time point. The interval comparison degree corresponding to a single evolution time point is the average value of the comparison reference value corresponding to each evolution parameter at that evolution time point. The comparison reference value of a single evolution parameter at a single evolution time point = the standard deviation of the value of the evolution parameter of each tile in the unpresented area at that evolution time point / the average value of the value of the evolution parameter of each tile in the unpresented area at that evolution time point. Evolution parameters include flood level, flow velocity, flow rate, inundation depth, and propulsion speed. The preset loading gradient change value and preset watershed evolution distortion degree are adaptively determined based on the real-time rendering load, network status, and the severity of flood evolution. The loading gradient change value and watershed evolution distortion degree effectively reflect the data loading pressure and the degree of abnormality in watershed parameter evolution in the target viewport pre-loading area. The greater the demand for smoothness and real-time loading stability in the digital twin scene, the smaller the preset loading gradient change value and preset watershed evolution distortion degree. In this embodiment, the preset loading gradient change value is 0.6 and the preset watershed evolution distortion degree is 0.3.

[0044] The preset network degradation coefficient and preset viewport feature anomaly are adaptively determined based on the real-time rendering load, network status, and the severity of flood evolution. The network degradation coefficient and preset viewport feature anomaly effectively reflect the degree of network transmission degradation and the degree of viewport loading feature anomaly. The greater the demand for real-time stable presentation and efficient loading of the watershed digital twin model, the smaller the preset network degradation coefficient and preset viewport feature anomaly values ​​will be. In this embodiment, the preset network degradation coefficient is 0.6 and the preset viewport feature anomaly is 0.5.

[0045] Direct loading means loading the tile data of the unpresented areas in the preloaded range to the front-end display terminal all at once.

[0046] Specifically, the time-domain layering module responds to the condition that the evolution law coefficient is greater than or equal to the preset evolution law coefficient and the tile evolution intensity is less than the preset tile evolution intensity, and then performs uniform layering based on the abnormal evaluation value.

[0047] Specifically, based on the interval comparison degree corresponding to each evolution time point in the preset time period, an interval evolution curve is drawn, with time as the horizontal axis and interval comparison degree as the vertical axis. The evolutionary regularity coefficient is determined as follows: the time between two evolutionary time points is recorded as an evolutionary segment, and the evolutionary regularity coefficient is the standard deviation of the slope variation corresponding to each evolutionary segment in the interval evolution curve; the slope variation corresponding to a single evolutionary segment in the interval evolution curve = |slope of the curve corresponding to the evolutionary segment in the interval evolution curve - slope of the curve corresponding to the evolutionary segment that is adjacent to and earlier than the evolutionary segment in the interval evolution curve| / [|slope of the curve corresponding to the evolutionary segment that is adjacent to and earlier than the evolutionary segment in the interval evolution curve| + 1]; if a single evolutionary segment does not have an evolutionary segment that is adjacent to and earlier than the evolutionary segment, then the slope variation corresponding to the evolutionary segment is recorded as 0; The tile evolution intensity is the standard deviation of the sub-intensity corresponding to each evolution segment. The sub-intensity corresponding to a single evolution segment is the larger of 1 - the number of identical feature time-domain tiles corresponding to two evolution time points in that evolution segment / the number of feature time-domain tiles corresponding to two evolution time points in that evolution segment. The feature time-domain tile corresponding to a single evolution time point is a tile whose time-domain feature coefficient is greater than the preset time-domain feature coefficient. The time-domain feature coefficient of a single tile at a single evolution time point is the ratio of the standard deviation of the pixel value corresponding to each pixel point of the tile at that evolution time point to the standard deviation threshold, and the standard deviation threshold is 127.5. The value of the preset time-domain feature coefficient can be adaptively set according to the actual application scenario. The time-domain feature coefficient effectively reflects the degree of fluctuation of the pixel value of the tile within a single evolution time point and is the core threshold for distinguishing ordinary tiles from feature time-domain tiles. The greater the need for the precision of key spatiotemporal feature identification and the accuracy of loading priority determination during flood evolution, the smaller the value of the preset time-domain feature coefficient. In this embodiment, the preset time-domain feature coefficient is 0.4. The values ​​of the preset evolution law coefficient and the preset tile evolution intensity are adaptively determined according to the actual application scenario. The evolution law coefficient and the tile evolution intensity effectively reflect the stability of the interval evolution curve and the intensity of change of the characteristic time domain tiles in the time dimension. The greater the demand for the visualization stability and loading smoothness of the flood evolution process, the larger the value of the preset evolution law coefficient and the smaller the value of the preset tile evolution intensity. In this embodiment, the preset evolution law coefficient is 0.75 and the preset tile evolution intensity is 0.30. When performing uniform stratification based on abnormal assessment values, the preset time period is evenly divided into several time segments of equal length. The number of time segments is equal to the number of abnormal assessment values ​​ / the preset abnormal assessment value × the number threshold. In this embodiment, the number threshold is 8. Anomaly assessment value = network degradation coefficient / preset network degradation coefficient × third weight coefficient + viewport feature anomaly degree / preset viewport feature anomaly degree × fourth weight coefficient, where the third weight coefficient and the fourth weight coefficient are both 0.5; The preset anomaly evaluation value is adaptively determined based on the actual application scenario. The anomaly evaluation value effectively reflects the degree of anomaly in the current loading environment and the necessity of temporal layered loading. The greater the demand for temporal layer refinement and rendering stability, the smaller the preset anomaly evaluation value. In this embodiment, the preset anomaly evaluation value is 0.5.

[0048] Specifically, the time-domain hierarchical module clusters and hierarchically performs clustering based on the evolution correlation degree if the response evolution law coefficient is less than the preset evolution law coefficient or the tile evolution intensity is greater than or equal to the preset tile evolution intensity.

[0049] When performing clustering hierarchy based on evolutionary correlation, each clustering time period is used as the loading level. A single clustering time period contains several evolutionary segments, each of which is a continuous time series segment. The evolutionary correlation between any two evolutionary segments within the clustering time period is greater than or equal to the preset evolutionary correlation. Furthermore, there are evolutionary segments within the clustering time period whose evolutionary correlation with the evolutionary segments on both sides of the clustering time period (i.e., the evolutionary segments adjacent to the clustering time period and located before and after it) is less than the preset evolutionary correlation. Evolutionary correlation degree between any two evolutionary segments = [1 - absolute value of the difference in slope change degree between the two evolutionary segments / (larger value of slope change degree between the two evolutionary segments + 1)] × 0.5 + [1 - absolute value of the difference in sub-intensity degree between the two evolutionary segments / (larger value of sub-intensity degree between the two evolutionary segments + 1)] × 0.5; The preset evolution correlation degree is adaptively determined according to the actual application scenario. The evolution correlation degree effectively reflects the similarity between the evolution trend and tile change characteristics between different evolution segments. The greater the demand for clustering hierarchical accuracy and temporal segment regularity, the larger the preset evolution correlation degree value. In this embodiment, the preset evolution correlation degree is 0.68.

[0050] When determining the loading priority coefficient of each loading level based on the interval reference value, the loading priority coefficient of a single loading level is -0.1 × the interval reference value. The larger the loading priority coefficient of a single loading level, the higher the priority of loading for that loading level. The interval reference value corresponding to a single loading level is the time length between the earliest evolution time point in that loading level and the initial time of the preset time period.

[0051] Specifically, the optimization adjustment module determines to perform optimization adjustment when the loading level heterogeneity is greater than the preset loading level heterogeneity or the multi-viewport loading interference is greater than the preset multi-viewport loading interference.

[0052] Specifically, the optimization adjustment module determines that no optimization adjustment is needed when the loading level heterogeneity is less than or equal to the preset loading level heterogeneity and the multi-viewport loading interference is less than or equal to the preset multi-viewport loading interference.

[0053] Loading level heterogeneity is the ratio of the duration of an abnormal loading level to the duration of a preset time period. An abnormal loading level is a loading level whose level change value is greater than the preset level change value, reflecting the smoothness of the loading transition between levels. The level change value corresponding to a single loading level = the average of the abnormal reference values ​​corresponding to each evolution segment in that loading level - the average of the abnormal reference values ​​corresponding to each evolution segment in the loading levels adjacent to that loading level and earlier in time. The abnormal reference value corresponding to a single evolutionary segment = slope change degree / preset slope change degree × change degree weighting coefficient + sub-intensity degree / preset sub-intensity degree × intensity degree weighting coefficient, where both the change degree weighting coefficient and the intensity degree weighting coefficient are 0.5; The preset level change value is adaptively determined according to the actual application scenario. The level change value effectively reflects the degree of difference in data loading gradient and the degree of switching anomaly between loading levels. The greater the need for a smooth transition of loading levels and a continuous and stable evolution process, the smaller the preset level change value. In this embodiment, the preset level change value is 0.6. The values ​​of the preset slope variation degree and the preset sub-intensity degree are adaptively determined according to the actual application scenario. The slope variation degree and the sub-intensity degree effectively reflect the degree of slope fluctuation of the interval evolution curve and the degree of change of characteristic time-domain tiles within the evolution segment. The greater the demand for the temporal stability, clustering layering accuracy and loading smoothness of flood evolution simulation, the smaller the values ​​of the preset slope variation degree and the preset sub-intensity degree. In this embodiment, the preset slope variation degree is 0.8 and the preset sub-intensity degree is 0.4. The multi-viewport loading interference is the ratio of the number of linked viewports to the preset number of linked viewports; The preset values ​​of multi-viewport loading interference and preset loading level heterogeneity can be determined based on the stability requirements of the loading level and the smoothness requirements of digital twin visualization. The multi-viewport loading interference and loading level heterogeneity effectively reflect the degree of mutual interference between multi-viewport parallel loading and the degree of heterogeneity and anomaly in the evolution of the loading level. The greater the requirement for the stability of flood evolution simulation, the smaller the values ​​of preset multi-viewport loading interference and preset loading level heterogeneity. In this embodiment, the preset multi-viewport loading interference is 0.45 and the preset loading level heterogeneity is 0.35.

[0054] Specifically, the optimization adjustment module reduces the preload range based on a comprehensive evaluation value; The overall evaluation value is positively correlated with both the heterogeneity of the loading level and the multi-viewport loading interference.

[0055] Specifically, the comprehensive evaluation value = (load level heterogeneity / preset load level heterogeneity × 0.5) + (multi-viewport load interference / preset multi-viewport load interference × 0.5); When adjusting the preload range based on the comprehensive evaluation value, if the target viewport is set to a baseline match, the preset length is adjusted based on the comprehensive evaluation value. If the comprehensive evaluation value is greater than or equal to the preset comprehensive evaluation value, the first proportional coefficient is adjusted to 0.12; if the comprehensive evaluation value is less than the preset comprehensive evaluation value, the first proportional coefficient is adjusted to 0.23. If the target viewport is set to orientation matching, the length of the sector axis is reduced based on the comprehensive evaluation value. If the comprehensive evaluation value is greater than or equal to the preset comprehensive evaluation value, the second proportional coefficient is adjusted to 0.4. If the comprehensive evaluation value is less than the preset comprehensive evaluation value, the second proportional coefficient is adjusted to 0.77.

[0056] The preset comprehensive evaluation value is adaptively determined according to the actual application scenario. The comprehensive evaluation value effectively reflects the degree of comprehensive anomaly after the superposition of loading layer heterogeneity and multi-viewport interference. The greater the demand for visual smoothness, the smaller the preset comprehensive evaluation value. In this embodiment, the preset comprehensive evaluation value is 0.9.

[0057] Specifically, the optimization adjustment module responds to the condition that the feature tile change coefficient is greater than the preset feature tile change coefficient, and determines to reduce the rendering reference value based on the abnormal comparison value; The decrease in the rendering reference value is positively correlated with the anomaly comparison value.

[0058] The optimization adjustment module determines that there is no need to reduce the rendering reference value based on the abnormal comparison value when the response feature tile change coefficient is less than or equal to the preset feature tile change coefficient. The preload range after reducing the preload range based on the comprehensive evaluation value is denoted as the adjusted preload range. The feature tile change coefficient = number of feature tiles in the unpresented area of ​​the adjusted preload range / total number of tiles in the unpresented area of ​​the adjusted preload range - number of feature tiles in the unpresented area of ​​the unadjusted preload range / total number of tiles in the unpresented area of ​​the unadjusted preload range. The value of the preset feature tile change coefficient is adaptively determined according to the actual application scenario. The feature tile change coefficient effectively reflects the distribution density of feature tiles and the proportion of rendering key areas within the preloaded range of the target viewport. The greater the demand for volume rendering smoothness, the smaller the value of the preset feature tile change coefficient. In this embodiment, the preset feature tile change coefficient is 0.3. Anomaly comparison value = (feature tile change coefficient - preset feature tile change coefficient) / preset feature tile change coefficient; The decrease in the rendering reference value = anomaly comparison value × initial rendering reference value; For a single evolution time point, the initial rendering reference value is negatively correlated with the display length corresponding to the target viewport. The initial rendering reference value = max(A1, A2), where A1 = rendering reference value threshold - display length corresponding to the target viewport / display length threshold × rendering reference value threshold × reference coefficient. The reference coefficient is 0.35, the rendering reference value threshold is 1.0, the display length threshold is 1920m, and A2 is 0.4. When A1 is less than A2, the initial rendering reference value is set to 0.4 to avoid a negative rendering reference value and ensure rendering accuracy in low display length scenarios.

[0059] The rendering reference value for a single evolution time point = the number of pixels actually generated during the loading process in the preload range of the target viewport at that evolution time point / the total number of pixels in the preload range of the target viewport at that evolution time point; Specifically, the optimization adjustment module determines feature tiles based on pixel change index and feature gradient.

[0060] Specifically, the feature tile is a tile in an unpresented area where the pixel change index is greater than a preset pixel change index or the feature gradient is greater than a preset feature gradient.

[0061] The pixel change index corresponding to a single tile is the average value of the temporal feature coefficients of that tile at each evolution time point within a preset time period; The feature gradient corresponding to a single tile is the average of the gradient reference values ​​corresponding to each evolution time point of the tile within a preset time period; the gradient reference value corresponding to a single tile at a single evolution time point is the average of the gradient comparison degree between the tile and each neighboring tile, and the neighboring tiles corresponding to a single tile are each tile adjacent to the tile in the adjustment preloading range; the gradient comparison degree between the tile and a single neighboring tile = the absolute value of the difference between the temporal feature coefficients corresponding to the two tiles / the larger value among the temporal feature coefficients corresponding to the two tiles.

[0062] The values ​​of the preset pixel change index and preset feature gradient can be adaptively set according to the actual application scenario. The pixel change index and feature gradient effectively reflect the significance of the temporal evolution characteristics of the tile itself and the degree of difference in spatial characteristics between tiles. The greater the demand for smooth rendering of the flood evolution digital twin scene, the smaller the values ​​of the preset pixel change index and preset feature gradient. In this embodiment, the preset pixel change index is 0.6 and the preset feature gradient is 0.5.

[0063] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A multi-model fusion flood evolution simulation system, characterized in that, include: The model building module is used to obtain flood spatiotemporal distribution data based on real-time meteorological data and watershed basic data, and through LSTM neural network precipitation prediction model, water flow propagation model and flood evolution model, and to build a digital twin model based on flood spatiotemporal distribution data and watershed basic data. The range determination module, which is connected to the model construction module, is used to determine the baseline matching setting or direction matching setting for the target viewport to obtain the preload range of the target viewport based on the linkage viewport fit degree and the number of linkage viewports. The loading analysis module, which is connected to the model building module and the range determination module respectively, is used to determine whether to adjust the loading method of the pre-loading range of the target viewport from direct loading to temporal layered loading based on the network degradation coefficient and the viewport feature anomaly degree; wherein, the viewport feature anomaly degree is determined based on the loading gradient change value and the watershed evolution distortion degree; The temporal layering module is connected to the range determination module and the loading analysis module respectively. It is used to determine, in the temporal layering loading, whether to perform uniform layering based on anomaly evaluation values ​​or clustering layering based on evolution correlation degrees to obtain several loading levels based on evolution law coefficients and tile evolution intensity, and to determine the loading priority coefficient of each loading level based on interval reference values. An optimization adjustment module, which is connected to the range determination module and the temporal layering module respectively, is used to determine whether to perform optimization adjustment based on the heterogeneity of the loading level and the multi-viewport loading interference. In the optimization adjustment, the pre-loading range is adjusted based on the comprehensive evaluation value, and the rendering reference value is adjusted based on the abnormal comparison value according to the feature tile change coefficient.

2. The flood evolution simulation system with multi-model fusion according to claim 1, characterized in that, The range determination module performs a baseline matching setting for target viewports where the linkage viewport fit degree is less than the preset linkage viewport fit degree or the number of linkage viewports is less than the preset number of linkage viewports. In the baseline matching settings, each tile in a circular area with the coordinates of the center point of the target viewport as the center and a preset length as the radius is used as the preload range of the target viewport.

3. The flood evolution simulation system with multi-model fusion according to claim 2, characterized in that, The range determination module performs direction matching settings for target viewports whose linkage viewport fitting degree is greater than or equal to the preset linkage viewport fitting degree and whose linkage viewport number is greater than or equal to the preset linkage viewport number. In the orientation matching settings, the coordinates of the center point of the target viewport's screen are used as the fan vertex, the direction of the two reference viewport vectors is used as the direction of the fan axis, and the tiles in the fan constructed based on the length of the fan axis determined by the length threshold and the tiles in the rectangular area corresponding to the target viewport are used together as the preload range of the target viewport. The reference viewport vector starts from the coordinate point of the center point of the target viewport, and the angle between the reference viewport vector and the viewport vector of the left neighboring viewport of the target viewport is the angle threshold.

4. The flood evolution simulation system with multi-model fusion according to claim 1, characterized in that, The loading analysis module adjusts the loading method from direct loading to temporal layered loading for the pre-loading range of target viewports where the network degradation coefficient is greater than or equal to the preset network degradation coefficient or the viewport feature anomaly degree is greater than or equal to the preset viewport feature anomaly degree.

5. The flood evolution simulation system with multi-model fusion according to claim 4, characterized in that, The time-domain stratification module responds to the condition that the evolution law coefficient is greater than or equal to the preset evolution law coefficient and the tile evolution intensity is less than the preset tile evolution intensity, and then performs uniform stratification based on the abnormal evaluation value.

6. The flood evolution simulation system with multi-model fusion according to claim 5, characterized in that, The time-domain hierarchical module performs clustering and hierarchical division based on the condition that the evolution law coefficient of the response is less than the preset evolution law coefficient or the tile evolution intensity is greater than or equal to the preset tile evolution intensity.

7. The flood evolution simulation system with multi-model fusion according to claim 1, characterized in that, The optimization and adjustment module determines to perform optimization and adjustment when the loading level heterogeneity is greater than the preset loading level heterogeneity or the multi-viewport loading interference is greater than the preset multi-viewport loading interference.

8. The flood evolution simulation system with multi-model fusion according to claim 7, characterized in that, The optimization adjustment module reduces the preload range based on a comprehensive evaluation value. The overall evaluation value is positively correlated with both the heterogeneity of the loading level and the multi-viewport loading interference.

9. The flood evolution simulation system with multi-model fusion according to claim 8, characterized in that, The optimization adjustment module responds to the condition that the feature tile change coefficient is greater than the preset feature tile change coefficient, and determines to reduce the rendering reference value based on the abnormal comparison value. The decrease in the rendering reference value is positively correlated with the anomaly comparison value.

10. The flood evolution simulation system with multi-model fusion according to claim 9, characterized in that, The optimization adjustment module determines feature tiles based on pixel change index and feature gradient.

Images