Drainage basin digital twinning construction method and system based on multi-source data fusion

Abstract

The invention provides a drainage basin digital twinning construction method and system based on multi-source data fusion. The method is applied to the technical field of drainage basin digital twinning construction, and comprises the following steps: acquiring water level monitoring data, surface flood routing path data, weather forecast rainfall data and terrain slope change data; rainfall and topographic slope data are fused to generate a runoff distribution relation, and a high-risk area is identified in combination with a water level sudden change amplitude. A digital twinning preliminary framework is established, and related characteristic parameters are generated by utilizing reverse feedback of a flood path and water level data. And inputting meteorological data, constructing a twinborn model based on the boundary condition of the catchment area, and simulating flood direction abnormity and flooding early warning signals. And according to the early warning space-time distribution and the terrain elevation data, generating emergency schemes such as flood dividing gate scheduling and flood storage and detention area starting, and realizing risk hierarchical management and control. According to the technical scheme provided by the invention, the flood inundation prediction precision and the emergency response efficiency are improved, and precise decision support is provided for flood control of a drainage basin.

Description

Technical Field

[0001] This application relates to the technical field of basin digital twin construction, and particularly to a method and system for constructing a basin digital twin based on multi-source data fusion. Background Art

[0002] In basin flood control and water resources management, it is necessary to grasp the flood evolution trend in real time, accurately predict the inundation range, and optimize emergency dispatching. There is an urgent need to construct a high-precision digital twin system that can integrate multi-source heterogeneous data (such as meteorology, terrain, hydrology), dynamically simulate the interaction process between the river channel and surface runoff, and quickly generate adaptive control strategies to cope with the spatio-temporal uncertainty of floods under extreme weather.

[0003] Currently, the mainstream solution uses a satellite remote sensing data-driven static hydrological model. By integrating multi-spectral satellite images (such as Sentinel-2) and historical hydrological databases, a two-dimensional flood diffusion model with fixed parameters is constructed. Based on the remotely sensed inundation range and rainfall forecast data, the hydrodynamic equations (such as the shallow water equations) are used to simulate flood evolution, and the prediction results are visualized through a GIS platform. Some systems introduce machine learning algorithms (such as LSTM) to optimize the rainfall-runoff relationship and improve short-term forecasting accuracy.

[0004] This solution has three core defects. First, it relies on the periodic update of satellite images (usually 6 - 12 hours), making it difficult to capture the rapid changes in floods (such as the sudden change in flow velocity at the moment of levee breach). Second, the fixed parameters cannot dynamically adapt to topographic micro-geomorphology (such as river channel siltation and vegetation cover changes), resulting in the accumulation of simulation errors. Third, there is a lack of a real-time monitoring data feedback mechanism, unable to achieve a closed-loop of "monitoring, simulation, and control", and the generation of emergency plans lags behind the actual disaster process. Summary of the Invention

[0005] This application provides a method and system for constructing a basin digital twin based on multi-source data fusion to solve the problem of the lack of dynamic fusion of multi-source data and accurate early warning ability in the prior art.

[0006] In the first aspect, this application provides a method for constructing a basin digital twin based on multi-source data fusion, including: Based on the interaction relationship between the river channel structure and surface runoff in the river hydrological interaction scenario, obtain the water level monitoring data of the river channel, the surface flood evolution path data, the meteorological forecast rainfall data, and the terrain slope change data; Fuse the meteorological forecast rainfall data with the terrain slope change data to generate the distribution relationship between rainfall and flood runoff based on the terrain slope change, and determine the high-risk area of flood evolution in the distribution relationship according to the abnormal water level change characteristics of the water level monitoring data. The abnormal water level change characteristics include the water level mutation amplitude at the key monitoring section in the river channel; Establish a preliminary framework of the watershed digital twin model corresponding to the high-risk area of flood evolution, generate associated characteristic parameters through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and adjust the flood interaction parameters between the river channel structure and surface runoff in the preliminary framework based on the associated characteristic parameters; Input the meteorological forecast rainfall data into the adjusted preliminary framework, and combine the flood interaction parameters with the watershed catchment boundary conditions in the terrain slope change data to construct a watershed digital twin model including abnormal change data of surface flood evolution direction and flood inundation warning signals in the river channel; Generate a flood emergency regulation plan according to the spatio-temporal distribution characteristics of the flood inundation warning signals in the watershed digital twin model, in combination with the river channel terrain elevation data.

[0007] Optionally, the establishment of the preliminary framework of the watershed digital twin model corresponding to the high-risk area of flood evolution, generating associated characteristic parameters through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and adjusting the flood interaction parameters between the river channel structure and surface runoff in the preliminary framework based on the associated characteristic parameters includes: Based on the spatial distribution range of the high-risk area of flood evolution, establish a preliminary framework of the watershed digital twin model including the river channel topological structure and the surface flood evolution path; Reverse track the continuous change process of the flood evolution path data in chronological order, and at the same time match the spatial distribution state of the water level monitoring data at the river channel monitoring section. According to the reverse tracking result and the matching result, construct the spatio-temporal association characteristics between the flood flow velocity mutation points in the flood evolution path data and the abnormal fluctuation points in the water level monitoring data; Perform multi-scale combination of the distribution density of the flood flow velocity mutation points and the water level fluctuation amplitude in the spatio-temporal association characteristics, establish an association characteristic parameter reflecting the correlation between the flood evolution path change pattern and the water level response intensity, and map the association characteristic parameter to the flood interaction parameters between the river channel structure and surface runoff in the preliminary framework; According to the mapping result, combine the change rate of the flood evolution path data at the boundary of the flood high-risk area with the continuous fluctuation amplitude of the water level monitoring data, dynamically update the flood interaction parameters, and perform spatial compensation on the updated flood interaction parameters through the spatial distribution range to adjust the preliminary framework.

[0008] Optionally, the fusion of the meteorological forecast rainfall data and the terrain slope change data to generate the distribution relationship between rainfall and flood runoff based on the terrain slope change, and determining the high-risk area of flood evolution in the distribution relationship according to the abnormal water level change characteristics of the water level monitoring data includes: Dynamically allocate the rainfall intensity values in the meteorological forecast rainfall data according to the spatial distribution law of the terrain slope change data to form the rainfall spatial coverage density with the slope change rate as the weight factor; Based on the guiding effect of the slope change rate on the surface flood flow direction, direct the rainfall intensity values in the high slope change rate area in the rainfall spatial coverage density to the low-lying confluence area, and combine the distribution density of the tributary confluence nodes in the surface flood evolution path to generate the distribution relationship between rainfall and flood runoff based on the terrain slope change; Calculate the difference between the water level mutation amplitude in the abnormal water level change characteristics and the flood runoff intensity in the corresponding geographical area in the distribution relationship, mark the abnormal area where the slope change rate does not match the flood runoff intensity in the distribution relationship, and based on the product coefficient of the difference and the slope change rate, perform gradient compensation on the flood runoff intensity value in the abnormal area according to the slope change rate; Overlay the compensated flood runoff intensity value with the spatial distribution of the abnormal fluctuation points in the water level monitoring data, and screen out the overlapping areas that simultaneously meet the slope change rate threshold, the flood runoff intensity compensation value, and the water level anomaly amplitude as the high-risk areas of flood evolution.

[0009] Optionally, inputting the meteorological forecast rainfall data into the adjusted preliminary framework, and combining the flood interaction parameters and the watershed catchment boundary conditions in the terrain slope change data to construct a watershed digital twin model including surface flood evolution direction abnormal change data and river flood inundation warning signals includes: Based on the watershed catchment boundary conditions in the terrain slope change data, perform spatial cutting on the adjusted preliminary framework to divide the surface flood evolution direction calculation unit based on the terrain slope change rate; Within the surface flood evolution direction calculation unit, determine the initial offset angle of the surface flood evolution direction according to the distribution characteristics of the slope turning points in the terrain slope change data, and correct the initial offset angle in combination with the flood interaction parameters to generate the corrected path of the abnormal change of the flood evolution direction; Couple the river channel water level critical threshold in the flood interaction parameters with the distribution density of river channel monitoring sections in the terrain slope change data. When the distribution density of the monitoring sections exceeds the critical threshold corresponding to the slope change rate, generate a flood inundation warning level based on the coupling result; Perform spatial overlay of the corrected path, the flood inundation warning level, and the meteorological forecast rainfall data, and conduct three-dimensional flood evolution topology reconstruction on the spatial overlay result according to the boundary conditions of the basin catchment area to generate a basin digital twin model that simultaneously includes the sudden change trajectory of the surface flood evolution direction and the flood inundation warning hot area of the river channel.

[0010] Optionally, trace back the continuous change process of the flood evolution path data in reverse chronological order, and simultaneously match the spatial distribution state of the water level monitoring data at the river channel monitoring sections. According to the reverse tracing result and the matching result, construct the spatio-temporal correlation characteristics between the flood flow velocity mutation points in the flood evolution path data and the abnormal fluctuation points in the water level monitoring data, including: Traverse the continuous records of the flood flow velocity in the flood evolution path data in reverse chronological order, identify the mutation points where the change amplitude of the flood flow velocity exceeds the preset mutation threshold within adjacent timestamps, and record the occurrence time of each mutation point and the corresponding river channel position coordinates as the reverse tracing result; Screen the monitoring sections in the water level monitoring data that are closest to the river channel position coordinates in the reverse tracing result, extract the cross-section water level change data within a preset time range before and after the occurrence time, and mark the abnormal fluctuation points where the water level change amplitude exceeds the historical average as the matching result; Align and match the occurrence time in the reverse tracing result with the maximum fluctuation time of the abnormal fluctuation points in the matching result to construct a set of spatio-temporal correlation pairs between the mutation points and the abnormal fluctuation points; Combine the velocity change amount of the mutation points, the water level fluctuation amplitude of the abnormal fluctuation points, and the time interval between the two in each correlation pair in the set of spatio-temporal correlation pairs into spatio-temporal correlation characteristics.

[0011] Optionally, the coupling of the river channel water level critical threshold in the flood interaction parameters with the distribution density of river channel monitoring sections in the terrain slope change data, and when the distribution density of the monitoring sections exceeds the critical threshold corresponding to the slope change rate, generating a flood inundation warning level based on the coupling result includes: Divide the terrain slope change data into regional units according to the slope change rate, count the total number of monitoring sections in each regional unit, and calculate the cross-section distribution density per unit area; Set the cross-section density critical values corresponding to different slope change rate intervals according to the correlation law between the slope change rate and the cross-section distribution density; Within the area unit, couple the critical threshold of the river water level in the flood interaction parameters with the cross-section distribution density according to a preset ratio. If the cross-section distribution density exceeds the critical value of the cross-section density, calculate the product of the water level difference and the slope change rate between this area and the adjacent area based on the coupling result; Divide three levels of flood inundation warning levels according to the product value, where the first-level warning corresponds to the highest interval of the product value, the second-level warning corresponds to the medium interval, and the third-level warning corresponds to the lowest interval.

[0012] Optionally, the generation of a flood emergency regulation plan based on the spatio-temporal distribution characteristics of the flood inundation warning signals in the watershed digital twin model and in combination with river channel terrain elevation data includes: In the watershed digital twin model, screen continuous areas where the signal density of the flood inundation warning signals exceeds a preset threshold, mark them as high-risk areas of flood inundation, and extract the time duration length of the high-risk areas; Match the spatial range of the high-risk areas of flood inundation with the river channel terrain elevation data, identify weak sections within the high-risk areas where the elevation is lower than the flood warning water level, and mark them as key sections for flood diversion and discharge; Based on the elevation gradient direction of the key sections for flood diversion and discharge, generate a multi-level flood diversion channel plan covering the upstream and downstream, where the capacity of the flood diversion channel is positively correlated with the elevation difference; Adjust the flood diversion gate control instructions for the key river channel cross-sections within the high-risk areas through the time duration length and the capacity change trend of the multi-level flood diversion channel plan, and generate a flood emergency regulation plan including the flood diversion channel activation rules and the gate regulation time sequence according to the adjusted control instructions.

[0013] In a second aspect, the present application provides a watershed digital twin construction system based on multi-source data fusion, including: An acquisition module, configured to acquire water level monitoring data of the river channel, surface flood evolution path data, meteorological forecast rainfall data, and terrain slope change data based on the interaction relationship between the river channel structure and surface runoff in the river hydrological interaction scenario; A fusion module, configured to fuse the meteorological forecast rainfall data and the terrain slope change data to generate the distribution relationship between rainfall and flood runoff based on the terrain slope change, and determine the high-risk areas of flood evolution in the distribution relationship according to the abnormal water level change characteristics of the water level monitoring data, where the abnormal water level change characteristics include the water level mutation amplitude at the key monitoring cross-sections in the river channel; An adjustment module, configured to establish a preliminary framework of a basin digital twin model corresponding to the high-risk area of flood evolution, generate associated feature parameters through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and adjust the flood interaction parameters between the river channel structure and the surface runoff in the preliminary framework based on the associated feature parameters; A construction module, configured to input the meteorological forecast rainfall data into the adjusted preliminary framework, and construct a basin digital twin model including abnormal change data of surface flood evolution direction and river channel flood inundation warning signals in combination with the flood interaction parameters and the basin catchment boundary conditions in the terrain slope change data; A generation module, configured to generate a flood emergency control plan according to the spatio-temporal distribution characteristics of the flood inundation warning signals in the basin digital twin model and in combination with the river channel terrain elevation data.

[0014] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for constructing a basin digital twin based on multi-source data fusion as described in the first aspect above.

[0015] In a fourth aspect, an embodiment of the present application provides a computer storage medium, storing a computer program, and when the computer program is executed by a computer, it implements a method for constructing a basin digital twin based on multi-source data fusion as described in the first aspect.

[0016] In the embodiments of the present application, based on the interaction relationship between the river channel structure and surface runoff in the river hydrological interaction scenario, water level monitoring data of the river channel, surface flood evolution path data, meteorological forecast rainfall data, and terrain slope change data are obtained; the meteorological forecast rainfall data and the terrain slope change data are fused to generate the distribution relationship between rainfall and flood runoff based on the terrain slope change, and according to the abnormal water level change characteristics of the water level monitoring data, the high-risk flood evolution area in the distribution relationship is determined, and the abnormal water level change characteristics include the water level mutation amplitude at the key monitoring section in the river channel; a preliminary framework of the watershed digital twin model corresponding to the high-risk flood evolution area is established, and correlation feature parameters are generated through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and the flood interaction parameters between the river channel structure and surface runoff in the preliminary framework are adjusted based on the correlation feature parameters; the meteorological forecast rainfall data is input into the adjusted preliminary framework, and combined with the flood interaction parameters and the watershed catchment boundary conditions in the terrain slope change data, a watershed digital twin model including abnormal change data of the surface flood evolution direction and flood warning signals for river channel inundation is constructed; according to the spatio-temporal distribution characteristics of the flood inundation warning signals in the watershed digital twin model, a flood emergency regulation plan is generated in combination with the river channel terrain elevation data.

[0017] The technical solution of the present application has the following beneficial effects: Through the collaborative collection of water level sensors, satellite remote sensing, and meteorological models, a three-dimensional monitoring network with high-frequency updates is constructed to realize the real-time perception of the flood discharge capacity of the river channel and the flood diffusion trend, ensuring the high-precision alignment of data in the spatio-temporal dimension; dynamically allocate rainfall intensity based on the terrain slope, and combine the abnormal mutation characteristics of the water level to identify high-risk areas, significantly improving the spatial resolution ability of runoff simulation and the positioning accuracy of risk areas; through the reverse feedback mechanism of flood path and water level data, dynamically optimize key parameters such as river channel roughness and flood storage capacity on the floodplain, enhancing the dynamic adaptation ability of the model to complex hydrological interaction processes; fuse catchment boundaries and meteorological data to accurately capture abnormal changes in flood flow directions, construct a three-dimensional visual warning system, and significantly shorten the response time and decision-making delay of flood inundation range prediction; based on spatio-temporal clustering and multi-objective optimization algorithms, generate a hierarchical regulation strategy for the linkage of flood diversion gates and flood storage and detention areas, effectively improving the water level regulation efficiency after the execution of emergency instructions.

[0018] Furthermore, a digital twin framework is constructed based on the spatial distribution of flood high-risk areas. By reverse-tracking the spatio-temporal correlation between the flood evolution path and water level fluctuations, multi-scale characteristic parameters are extracted to dynamically adjust the model interaction parameters, and the global consistency of the framework is optimized in combination with the spatial compensation mechanism. Through dynamic parameter optimization and spatial compensation, the accuracy of floodplain flood storage capacity prediction and river roughness calibration is significantly improved, the simulation time of the flood evolution path in high-risk areas is greatly shortened, a high-timeliness and highly adaptable decision-making basis is provided for flood diversion scheduling, and the flood control response efficiency of the basin is comprehensively enhanced.

[0019] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0021] Figure 1 Shows a flowchart of a method for constructing a basin digital twin based on multi-source data fusion provided by the present application; Figure 2 Shows a schematic structural diagram of a system for constructing a basin digital twin based on multi-source data fusion provided by the present application; Figure 3 Shows a schematic structural diagram of a computing device provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0023] In some processes described in the specification and claims of the present application and the above drawings, there are multiple operations that appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear herein or may be executed in parallel. The operation numbers such as 101 and 102 are only used to distinguish different operations, and the numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0024] Researchers have found that current flood simulation technologies in river basins generally face bottleneck problems such as insufficient collaboration of multi-source data, static and rigid model parameters, and lagging emergency decision-making, resulting in low accuracy of flood evolution prediction, large deviation in the identification of high-risk areas, and weak adaptability of regulation schemes. Based on this, a method for constructing a digital twin of a river basin based on multi-source data fusion is provided. Specifically, a dynamic runoff distribution relationship is generated by fusing meteorological forecast rainfall and terrain slope data, and high-risk areas are locked in combination with the characteristics of water level mutations; a digital twin framework is constructed and the model parameters are dynamically optimized based on the reverse feedback of flood path and water level data; finally, a hierarchical emergency plan is generated by combining three-dimensional warning signals and terrain elevation. This method can achieve centimeter-level water level perception (sensor accuracy ±1 cm) in the flood evolution process, hour-level model dynamic calibration (iterating once every 2 hours), and hour-level warning response (delay from data collection to warning release ≤ 60 minutes), meeting the actual technical conditions of wide-area monitoring scenarios, and significantly improving the reliability of flood inundation prediction and the timeliness of emergency regulation under complex terrain.

[0025] The technical solution of this application can be applied to the scenario of constructing a digital twin model of a river basin.

[0026] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0027] Figure 1 The following is a flowchart of a method for constructing a digital twin of a river basin based on multi-source data fusion provided for the embodiments of the present application, as Figure 1 shown, the method includes: 101. Based on the interaction relationship between the river channel structure and surface runoff in the river hydrological interaction scenario, obtain the water level monitoring data of the river channel, the surface flood evolution path data, the meteorological forecast rainfall data, and the terrain slope change data; In this step, the water level monitoring data is the water level elevation data (updated once per second, for example) collected in real time by water level sensors deployed at river channel sections, reflecting the real-time flood carrying capacity of the river channel. The surface flood evolution path data is the flood inundation range and flow direction data (updated once per hour, for example) obtained by satellite remote sensing or UAV aerial photography, characterizing the flood diffusion trend. The meteorological forecast rainfall data is the prediction data of rainfall intensity, duration, and spatial distribution within the next 6 hours in the river basin, generated by a numerical weather prediction model. The terrain slope change data is a raster map of the slope change rate calculated based on the basin surface elevation data (resolution 1 m × 1 m) obtained by airborne LiDAR.

[0028] In the embodiments of the present application, first, water level sensors (sampling frequency 1Hz) are deployed at key river cross-sections (such as tributary confluences, downstream of reservoirs), and the flood evolution path is monitored collaboratively by a synthetic aperture radar satellite (SAR) and a multi-rotor unmanned aerial vehicle (spatial resolution 0.5m). The rainfall forecast data output by the WRF meteorological model (temporal resolution 15 minutes) is accessed, and the terrain slope change rate is calculated by invoking the watershed digital elevation model (DEM). Secondly, the water level data eliminates wave interference through Kalman filtering, the flood path data extracts the inundation boundary using an image segmentation algorithm, and the rainfall data and terrain data are aligned to a unified grid (100m×100m) through spatial interpolation. Finally, the above data is associated and stored in the watershed database according to the spatio-temporal tags (UTC time + geographical coordinates) to support dynamic query and analysis.

[0029] A rainstorm flood event occurred in a certain river basin. Water level sensors were deployed at the main stream cross-section S1 (coordinates X: 120, Y: 350), and the water level was recorded in real time from 5.2m to 7.8m (warning water level 7.5m). The Sentinel-1 satellite monitoring showed that the flood evolved from northwest to southeast, and the inundated area reached 12 km². The meteorological model predicted that the basin-wide rainfall would reach 50mm in the next 3 hours, and the maximum point rainfall would be 80mm / h. A slope change map was generated based on 1m resolution LiDAR data, showing a sudden drop in the slope in the downstream area (8%→2%). All data was associated through a spatio-temporal index and marked as "2023-07-15T14:00Z / Basin A".

[0030] 102. Integrate the meteorological forecast rainfall data with the terrain slope change data to generate the distribution relationship between rainfall and flood runoff based on the terrain slope change, and determine the high-risk areas of flood evolution in the distribution relationship according to the abnormal water level change characteristics of the water level monitoring data. The abnormal water level change characteristics include the water level mutation amplitude at key monitoring cross-sections in the river channel.

[0031] In this step, the distribution relationship between rainfall and flood runoff is the spatial distribution law of rainfall converted into surface runoff based on the terrain slope change (for example, the runoff convergence speed is fast in areas with a slope above 8%, and temporary flood storage areas are easily formed in depressions with a slope below 2%). The abnormal water level change characteristics are that the water level at the key monitoring cross-section rises by more than 2 times the standard deviation of the historical average in a short time (such as within 1 hour) (for example, the water level suddenly rises by 1.5 meters), reflecting the abnormal flood carrying capacity of the river channel. The high-risk areas of flood evolution are the overlapping areas with high runoff intensity and frequent water level mutations (such as river sections with a water level mutation > 2 times per hour).

[0032] In the embodiments of the present application, first, the meteorological forecast rainfall data and the terrain slope change data are fused through the watershed hydrological model (SWAT) to generate a slope-oriented runoff distribution raster map (for example, when the slope increases by 5%, the runoff velocity increases by 0.3 m / s); secondly, the sliding window Z-Score detection method (window length 1 hour) is applied to the water level monitoring data to mark the cross-sections where the water level mutation amplitude exceeds the threshold (such as Z>3); finally, the river reaches with runoff intensity >50 mm / h and water level mutation frequency >2 times / hour are screened through spatial hotspot analysis (Getis-Ord Gi*), and are determined as high-risk areas for flood evolution (such as the main stream S1-S2 river reach).

[0033] For example, continuing with the above example, first, the SWAT model fuses rainfall and terrain data to predict that the runoff intensity in the downstream depression area (slope 1.5%-2%) will reach 95 mm / h, forming a temporary flood storage area; secondly, the Z-Score detection finds that the water level at cross-section S3 suddenly rises by 2.6 m within 1 hour (Z = 4.2), far exceeding the historical average; thirdly, the hotspot analysis locks the downstream area X:200-500 (covering cross-sections S1-S3) as a high-risk area, whose characteristics are: runoff intensity >90 mm / h, water level mutation frequency 3 times / hour, and satellite shows that the flood is rapidly evolving towards this area; finally, the river reach X:200-500 is marked as a red high-risk area and pushed to the emergency command system, triggering the pre-discharge of the downstream reservoir and the evacuation order for residents.

[0034] 103. Establish a preliminary framework of the watershed digital twin model corresponding to the high-risk area for flood evolution, generate associated characteristic parameters through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and adjust the flood interaction parameters between the river channel structure and the surface runoff in the preliminary framework based on the associated characteristic parameters; In this step, the reverse feedback relationship is to establish a dynamic influence link between the surface runoff and the flood carrying capacity of the river channel by analyzing the temporal sequence correlation between the changes in the flood evolution path and the water level mutation. The associated characteristic parameters are to quantify the spatio-temporal correlation intensity between the changes in the flood evolution speed and the amplitude of the water level fluctuation. The flood interaction parameters are dynamic parameters such as the roughness coefficient of the river channel and the flood storage capacity of the floodplain defined in the model, which are used to adjust the interaction intensity between the surface runoff and the river channel.

[0035] In the embodiments of the present application, first, based on the spatial scope of the high-risk area of flood evolution (such as the X:200 - 500 river section), a preliminary framework including river cross-sections, levee structures, and surface runoff paths is constructed using a hydrodynamic modeling engine (such as HEC-RAS); second, the flood evolution path data is parsed in reverse chronological order, and a sliding window alignment mechanism (window length 15 minutes) is used to perform time interpolation matching on satellite remote sensing flood evolution data (hourly level) and water level monitoring data (second level). Combining with a lightweight temporal convolutional network (TCN) to extract spatio-temporal correlation features, reducing the algorithm complexity and improving the real-time performance; third, the matched spatio-temporal correlation features are input into a random forest regression model, and an online learning mechanism is introduced. Combining the real-time flood evolution path data and the hydrodynamic equations (such as the Saint-Venant equations) in the hydrodynamic modeling engine for hybrid calibration, and generating characteristic parameters reflecting the correlation strength between flow velocity and water level according to the calibration results (such as when the flow velocity drops by 1 m / s, the water level rises by 0.8 m); finally, based on the gradient descent optimization algorithm, the characteristic parameters are mapped to the flood interaction parameters in the preliminary framework (such as adjusting the river roughness coefficient from 0.025 to 0.035, and the flood storage capacity on the floodplain increases by 15%).

[0036] For example, continuing the above example, first, based on the LiDAR terrain and river cross-section data of the X:200 - 500 river section, a digital twin framework including the main river channel (width 80 m) and the floodplains on both sides is constructed; second, the flood evolution data shows that at 15:00, the flow velocity suddenly drops from 3.2 m / s to 1.0 m / s at X:250, and the water level at cross-section S2 suddenly rises by 1.8 m (Z = 4.0) at 15:05. The DTW algorithm confirms that the time difference between the two is 5 minutes (less than the hydraulic conduction time of 6 minutes); third, the random forest model calculates that the correlation strength between flow velocity and water level is 0.75 (when the flow velocity drops by 1 m / s, the water level rises by 0.75 m), adjusts the flood storage capacity on the floodplain from 500,000 m³ to 575,000 m³ (+15%), and adjusts the river roughness from 0.028 to 0.032; finally, the updated digital twin framework is pushed to the flood forecasting system for subsequent emergency dispatching.

[0037] 104. Input the meteorological forecast rainfall data into the adjusted preliminary framework, and combine the flood interaction parameters and the basin catchment boundary conditions in the terrain slope change data to construct a basin digital twin model including surface flood evolution direction abnormal change data and river flood inundation warning signals; In this step, the basin catchment boundary conditions are the sub-basin watershed boundaries divided based on the digital elevation model (DEM), which are used to constrain the flood simulation range. The flood inundation warning signal is a risk alert triggered when the model predicts that the water level exceeds the levee design standard or the flow velocity exceeds the safety threshold.

[0038] In the embodiments of the present application, first, the real-time meteorological rainfall data is input into the adjusted digital twin framework, and the surface flood evolution direction is simulated based on the two-dimensional shallow water equation (SWE) to detect abnormal areas where the deviation from the terrain slope direction is greater than 15°; second, according to the catchment boundary conditions, the adaptive mesh refinement technique (AMR) is used to divide the computational units to ensure that the mesh resolution at the boundary reaches 10 m × 10 m; third, combined with the floodplain flood storage capacity in the flood interaction parameters, the water level and inundation range of the river channel are calculated through a hydrodynamic coupling model (such as MIKE FLOOD), and a flood inundation warning is triggered when the predicted water level exceeds the warning line or the flow velocity exceeds the threshold; finally, the flood direction abnormal trajectory (such as a blue vector arrow) and the warning hot zone (such as a red inundated block) are fused through a GPU-accelerated rendering engine to generate an interactive three-dimensional twin model.

[0039] For example, continuing the above example, first, the meteorological forecast data (the peak rainfall in the next 2 hours is 60 mm / h) is input. The simulation of the SWE model shows that the flood direction in the river section X: 300 - 400 deviates from the main river channel axis by 25° (abnormal change); second, based on the catchment boundary (sub-watershed W-05), grids are encrypted and generated, and it is detected that the floodplain flood storage capacity in this area has reached 575,000 m³ (the critical value is 600,000 m³) after adjustment; third, the MIKE FLOOD model predicts that the water level at section S4 will exceed the warning line by 1.2 m (design standard + 1.0 m) in 1 hour, triggering a red flood inundation warning; finally, the three-dimensional model marks the river section X: 300 - 400 as an abnormal evolution area (blue arrow) and a high flood inundation risk area (red block), and pushes it to the flood control command platform to initiate the flood diversion sluice scheduling.

[0040] 105. Generate a flood emergency regulation plan based on the spatio-temporal distribution characteristics of the flood inundation warning signals in the digital twin model of the basin.

[0041] In this step, the spatio-temporal distribution characteristics are the aggregation density of the flood inundation warning signals in space and the persistence in time. The river channel terrain elevation data is the river channel cross-section and surrounding terrain elevation data obtained through airborne LiDAR or underwater sounding, which is used to analyze the flood evolution trend and the flood diversion path planning. The flood emergency regulation plan is an integrated plan containing instructions such as flood diversion sluice scheduling, activation of flood storage and detention areas, and allocation of emergency rescue materials, which is used to reduce the flood risk.

[0042] In the embodiments of the present application, first, the spatio-temporal clustering algorithm (ST-DBSCAN) is used to analyze early warning signals, and high-risk river reaches with the density of early warning points exceeding the threshold (such as 8 per 5 square kilometers) and lasting for more than 2 hours are screened; secondly, based on the river channel terrain elevation data, a flood diversion path is generated through a hydrodynamic optimization model (such as NSGA-II) (such as opening the flood diversion sluice at X:300 and diverting 30% of the flow to the flood storage and detention area); thirdly, combined with the early warning level (red / yellow / blue), the optimal regulation combination is dynamically selected through a multi-objective decision-making system (such as activating 2 flood storage and detention areas + deploying 3 emergency rescue teams for garrison when a red alert is issued); finally, the flood diversion path, the flood storage and detention area dispatching and the emergency rescue instructions are integrated to generate an emergency regulation plan with time and space coordination.

[0043] For example, continuing with the above example, first, ST-DBSCAN clustering shows that the early warning density in the river reach of X:300-500 reaches 12 per 5 km² and lasts for 3 hours, which is marked as a red risk area; secondly, based on the LiDAR terrain data (the main river channel elevation is 5.2 m and the flood diversion channel elevation is 4.8 m), the NSGA-II model generates the optimal flood diversion plan: open the flood diversion sluice at X:350 and divert 25% of the flow (about 500 m³ / s) to the flood storage and detention area A; thirdly, according to the red early warning level, the multi-objective decision-making system synchronously issues the following instructions: open the floodgate of the flood storage and detention area A for flood storage (60% of the remaining capacity); deploy the emergency rescue team to the weak section of the dike at X:400; issue a secondary evacuation alert to the downstream residential area; finally, after the generated emergency regulation plan is executed for 2 hours, the water level in the river reach of X:300-500 drops by 0.8 m, and the early warning is downgraded to yellow.

[0044] In steps 101-105, by constructing an integrated sky-earth-space hydrological monitoring network, integrating multi-source information such as high-frequency water level data, satellite remote sensing images, weather forecasts, and terrain slopes, a dynamically coupled digital twin model is established. Machine learning algorithms are used to optimize river channel parameters in real time, accurately identify high-risk areas of flood evolution, and based on three-dimensional visualization technology, intuitively present abnormal flow directions and inundation ranges. By generating a flood diversion scheduling plan through intelligent optimization algorithms, the timeliness of flood early warning and the accuracy of emergency regulation are significantly improved, and finally a closed-loop management from real-time monitoring, risk prediction to intelligent decision-making is formed.

[0045] In some embodiments, to establish a preliminary framework of the basin digital twin model corresponding to the high-risk area of flood evolution, correlation characteristic parameters are generated through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and based on the correlation characteristic parameters, the flood interaction parameters between the river channel structure and the surface runoff in the preliminary framework are adjusted, including: 201. Based on the spatial distribution range of the high-risk area of flood evolution, establish a preliminary framework of the basin digital twin model including the river channel topological structure and the surface flood evolution path; In step 201, the river channel topology refers to the spatial connection relationship of the main river channel, tributaries, and river network nodes, while the surface flood evolution path characterizes the dynamic diffusion trajectory of floods within the basin.

[0046] In the embodiments of the present application, first, based on LiDAR terrain data and river channel cross-section measurement results, the spatial boundary of the high-risk area is delimited through a GIS platform; second, the Delaunay triangulation algorithm is used to construct the topological connection relationship of the river network; third, the flood evolution path data obtained from satellite remote sensing is converted into a vector trajectory; finally, the river channel topology and the surface path are spatially superimposed and fused to form a preliminary digital twin framework including three-dimensional terrain, river channel structure, and flood diffusion direction.

[0047] 202. Reverse-track the continuous change process of the flood evolution path data in chronological order, and at the same time match the spatial distribution state of the water level monitoring data at the river channel monitoring cross-section. According to the reverse-tracking result and the matching result, construct the spatio-temporal correlation characteristics between the flood velocity mutation points in the flood evolution path data and the abnormal fluctuation points in the water level monitoring data; In step 202, the spatio-temporal correlation characteristics refer to the quantitative relationship between flood velocity mutation and water level abnormal fluctuation in terms of time synchronization and spatial proximity.

[0048] In the embodiments of the present application, first, starting from the latest inundation boundary point, extract the velocity mutation points in the SAR satellite image in reverse chronological order (such as the velocity drops from 3.0 m / s to 1.5 m / s at X:200); second, synchronously extract the water level mutation data at cross-sections S1 - S3 during the corresponding period (such as the water level at S2 suddenly rises by 1.8 m at 15:05); third, match the time difference (<5 minutes) and spatial distance (<100 m) between the two through the dynamic time warping (DTW) algorithm; finally, generate spatio-temporal correlation characteristics and mark the strong correlation area between velocity and water level (such as the correlation strength R²>0.8 in the X:200 - 250 river section).

[0049] 203. Combine the distribution density of the flood velocity mutation points and the water level fluctuation amplitude in the spatio-temporal correlation characteristics at multiple scales, establish a correlation characteristic parameter reflecting the correlation between the flood evolution path change pattern and the water level response intensity, and map the correlation characteristic parameter to the flood interaction parameter between the river channel structure and the surface runoff in the preliminary framework; In step 203, the correlation characteristic parameter is a coupling index quantifying the velocity mutation frequency and the water level fluctuation amplitude, and the flood interaction parameter includes key hydraulic factors such as river channel roughness and flood storage capacity on the floodplain.

[0050] In the embodiment of the present application, first, the spatio-temporal correlation features are divided into grids of 500m×500m, and the mutation density of the flow velocity (such as 5 times / h) and the mean value of the water level fluctuation (such as ±1.2m) in each grid are statistically calculated; secondly, the mutation density and the mean value of the water level fluctuation are trained through a random forest regression model to generate the response coefficient between the runoff and the water level (such as when the flow velocity drops by 1m / s, the water level rises by 0.75m); thirdly, based on the watershed hydrological model (HEC-RAS), the response coefficient is mapped to the channel roughness (0.025→0.032) and the floodplain capacity (50→575,000 m³); finally, the updated parameters are mapped into the flood interaction parameters between the channel structure and the surface runoff in the preliminary framework through the API interface.

[0051] 204. According to the mapping result, in combination with the change rate of the flood evolution path data at the boundary of the high flood risk area and the continuous fluctuation amplitude of the water level monitoring data, the flood interaction parameters are dynamically updated, and spatial compensation is performed on the updated flood interaction parameters through the spatial distribution range to adjust the preliminary framework.

[0052] In step 204, dynamic update means iteratively optimizing the model parameters based on real-time data, and spatial compensation eliminates local distortion caused by parameter mutation through an interpolation algorithm.

[0053] In the embodiment of the present application, first, the change gradient of the flow velocity at the boundary of the high-risk area (such as a decrease of 1.2m / s / h at X:300) and the continuous fluctuation of the water level (±1.5m at section S3 for 1h) are detected; secondly, after removing the noise by using the Kalman filter, the floodplain capacity is adjusted by the gradient descent method (575,000→600,000 m³); thirdly, Kriging interpolation is used to smooth the global parameters to ensure continuous transition of the channel roughness between 0.030 and 0.035; finally, the optimized preliminary framework is output.

[0054] The following is a specific example: Suppose heavy rain and flood occur in a river basin. In step 201, a digital twin framework is constructed based on LiDAR terrain and SAR images (0.5m resolution), including the main stream cross-sections S1 - S3 (width 80m) and the floodplain areas on both sides (slope 1.5% - 8%), and the flood evolution path monitored by Sentinel-1 (12km² inundation range) is overlaid. In step 202, reverse tracking shows that the flow velocity suddenly drops from 3.2m / s to 1.0m / s at X:250 at 15:00, and it is matched that the water level at cross-section S2 suddenly rises by 1.8m (Z = 4.0) at 15:05, generating spatio-temporal correlation features (time difference 5 minutes, distance difference 50m, water level rise of 0.75m corresponding to a flow velocity drop of 1m / s). In step 203, the flow velocity mutation density (4 per km²) and water level fluctuation amplitude (1.2m) in the section of X:300 - 350 are statistically analyzed within a 100m grid, and correlation parameters (floodplain flood storage coefficient 0.73) are generated through random forest regression, and the roughness of this section is updated from 0.028 to 0.032. In step 204, an abnormal flow velocity gradient (1.8m / s / 100m) is detected in the section of X:400 - 500. Combining with the continuous fluctuation of cross-section S3 (±0.6m / 30 minutes), the floodplain coefficient is dynamically adjusted to 0.68, and parameter smooth transition is achieved through Kriging interpolation. Finally, the error of the inundation range predicted by the model is reduced from 15% to 3.5%, providing accurate decision-making support for flood diversion sluice scheduling.

[0055] Steps 201 - 204 construct a dynamically feedback digital twin model by integrating river channel topology, flood path, and water level monitoring data. Based on spatio-temporal correlation features, hydraulic parameters are inversely optimized to achieve accurate mapping of the flow velocity - water level coupling relationship; combining multi-scale analysis and spatial compensation mechanism significantly improves the simulation reliability of the model in high-risk areas.

[0056] In some embodiments, the meteorological forecast rainfall data and the terrain slope change data are fused to generate the distribution relationship between rainfall and flood runoff based on the terrain slope change, and according to the abnormal water level change characteristics of the water level monitoring data, the high-risk areas of flood evolution in the distribution relationship are determined. The abnormal water level change characteristics include the water level mutation amplitude at key monitoring cross-sections in the river channel, including: 301. Dynamically allocate the rainfall intensity values in the meteorological forecast rainfall data according to the spatial distribution law of the terrain slope change data to form the rainfall spatial coverage density with the slope change rate as the weight factor; In step 301, the rainfall spatial coverage density is the redistribution result of the rainfall intensity under the terrain slope difference. Steep slope areas are given higher rainfall weights due to fast runoff velocity, while low-lying areas need to receive migrated rainfall due to weak water storage capacity.

[0057] In the embodiment of the present application, first, the rainfall forecast data (at 15-minute intervals) output by the WRF model is spatially registered with the LiDAR slope raster at a resolution of 1 m. Secondly, the inverse distance weighting (IDW) interpolation algorithm is used to establish the mapping relationship between the slope change rate (such as 8% / 100 m) and the rainfall weight (1.0 - 1.5 times). Thirdly, a rainfall weight of 1.4 times is assigned to the steep slope area in the upper reaches of the main stream (slope change rate > 10%). Finally, the rainfall spatial coverage density with a grid accuracy of 100 m is generated (for example, the rainfall intensity in the area of X: 150 - 200 is adjusted from 50 mm / h to 70 mm / h).

[0058] Optionally, when performing step 301, for different scenarios, different data can be dynamically allocated. For example, if the high-slope area is covered by forests, its runoff velocity will be significantly lower than that of the bare ground surface. At this time, the rainfall intensity value in the meteorological forecast rainfall data can also be dynamically allocated according to the terrain slope change data, soil permeability (based on the soil type database), vegetation coverage index (based on multi-spectral remote sensing data), and land use type (such as the proportion of urban impervious surfaces), and the SCS-CN comprehensive weight model is used to generate the rainfall spatial coverage density, where the slope change rate is the core weight factor, and the soil permeability and vegetation coverage coefficient are used as correction factors.

[0059] It should be noted that the dynamic allocation of the rainfall intensity value in the meteorological forecast rainfall data according to the spatial distribution law of the terrain slope change data in step 301 to form the rainfall spatial coverage density with the slope change rate as the weight factor is only an example. The formation of the rainfall spatial coverage density with different weight factors by dynamically allocating different data under different scenarios is within the protection scope of the present application.

[0060] 302. Based on the guiding effect of the slope change rate on the surface flood flow direction, the rainfall intensity value in the high-slope change rate area in the rainfall spatial coverage density is directionally migrated to the low-lying confluence area, and combined with the distribution density of the tributary confluence nodes in the surface flood evolution path, the distribution relationship between rainfall and flood runoff based on the terrain slope change is generated. In step 302, the guiding effect refers to the control effect of the slope gradient on the main flood flow direction. The distribution density of the tributary confluence nodes is identified by aerial drone photography.

[0061] In the embodiments of the present application, first, the dominant runoff direction is determined based on the D8 algorithm (e.g., the flow direction of section S1 is 145°). Secondly, the slope-directed migration (SDM) algorithm is used to migrate 30% of the rainfall intensity in the upstream steep slope area (X: 100 - 150) to the downstream depression (X: 300 - 350). Thirdly, the tributary confluence hotspots are calculated through kernel density estimation (KDE) (e.g., the density reaches 5 / km² at X: 250). Finally, the migrated rainfall intensity (85 mm / h) and the confluence point density are fused according to a 7:3 weight to generate the distribution relationship between rainfall and flood runoff (e.g., the comprehensive runoff intensity at X: 250 is marked as 92 mm / h).

[0062] 303. Calculate the difference between the water level mutation amplitude in the abnormal water level change characteristics and the flood runoff intensity in the corresponding geographical area in the distribution relationship. Mark the abnormal area where the slope change rate does not match the flood runoff intensity in the distribution relationship, and based on the product coefficient of the difference and the slope change rate, perform gradient compensation on the flood runoff intensity value in the abnormal area according to the slope change rate; Among them, in combination with the SCS-CN curve number correction, based on the product coefficient of the difference and the slope change rate, perform gradient compensation on the flood runoff intensity value in the abnormal area, and introduce an urbanization correction factor (the compensation amount increases by 20% when the impervious surface ratio ≥ 30%).

[0063] In step 303, the water level mutation amplitude refers to the value by which the water level at the river channel monitoring section rises within a short period of time (usually 1 hour) exceeding the historical normal fluctuation range, reflecting the abnormal change in the flood discharge capacity of the river channel.

[0064] In the embodiments of the present application, first, locate the runoff intensity (80 mm / h) corresponding to the water level mutation point (rising 1.5 m) at X: 200. Secondly, calculate the theoretical runoff intensity difference (80 - 65 = 15). Thirdly, based on the product coefficient (15 × 6% = 0.9), compensate the runoff intensity to 80 + (0.9 × 10) = 89 mm / h. Finally, mark X: 200 - 220 as the abnormal area where runoff and slope do not match (red warning area) in the model.

[0065] 304. Superimpose the compensated flood runoff intensity value on the spatial distribution of the abnormal fluctuation points in the water level monitoring data, and screen out the overlapping area that simultaneously meets the slope change rate threshold, the flood runoff intensity compensation value, and the water level abnormal amplitude as the high-risk area for flood evolution.

[0066] In step 304, the high-risk area for flood evolution: the spatial overlapping area that simultaneously meets the steep slope change, the runoff intensity exceeding the threshold, and the abnormal water level fluctuation, usually located at the tributary confluence or the river channel turning section, has the dual risks of rapid flood accumulation and diffusion.

[0067] In the embodiments of the present application, first, the flood runoff intensity value after superposition compensation (89 mm / h at X:200) is superposed with the water level anomaly fluctuation point (section S2). Secondly, grids that simultaneously meet the conditions of slope > 5%, runoff > 85 mm / h, and water level rise > 1 m in the superposition result are screened (such as X:180 - 250). Thirdly, the flood evolution high-risk area (with an area of 3.2 km²) is generated for the grids through spatial clustering (DBSCAN algorithm). Finally, the accuracy of the model is verified by comparing this area with the actually measured inundation range of Sentinel-1 (coincidence rate 89%).

[0068] The following is a specific example: Suppose during a rainstorm in a certain basin, in step 301, the predicted rainfall peak (80 mm / h) is fused with the LiDAR slope data, and a 1.5-fold weight (actually 120 mm / h) is assigned to the upstream steep slope area (X:100 - 200, slope change rate 12%). In step 302, 30% of the rainfall (36 mm / h) in this area is migrated to the downstream depression (X:300 - 350) through the SDM algorithm, and a runoff distribution map (comprehensive value 105 mm / h at X:300) is generated in combination with the tributary confluence density (4 per km²). In step 303, it is detected that the water level at section S3 suddenly rises by 2.1 m (Z = 4.5). After calculating the runoff difference (105 - 82 = 23), compensation is carried out based on the slope change rate of 8% (23×8%×8 = 14.7), and the runoff at X:300 is updated to 119.7 mm / h. In step 304, the area of X:280 - 350 (slope 7%, runoff > 110 mm / h, water level rise > 1.5 m) is screened out. After morphological processing, a 2.8 km² high-risk area is determined. Comparing with satellite images shows that 85% of the area has been covered by the flood front, verifying the effectiveness of the model warning.

[0069] Steps 301 - 304 construct a high-precision flood runoff distribution model by dynamically allocating rainfall intensity to different slope areas and combining a terrain-guided runoff migration algorithm. Based on the correlation analysis of water level mutations and slope changes, areas where runoff and terrain do not match are automatically identified and compensated. Finally, the high-risk areas of flood evolution are accurately locked through multi-threshold spatial superposition, realizing a full-chain closed loop from rainfall prediction to risk warning.

[0070] In some embodiments, inputting the meteorological forecast rainfall data into the adjusted preliminary framework, and combining the flood interaction parameters with the basin catchment boundary conditions in the terrain slope change data to construct a basin digital twin model including abnormal change data of surface flood evolution direction and river channel flood inundation warning signals, includes: 401. Based on the basin catchment boundary conditions in the terrain slope change data, perform spatial cutting on the adjusted preliminary framework to divide surface flood evolution direction calculation units based on the terrain slope change rate; In step 401, the boundary condition of the watershed catchment refers to the closed boundary of the watershed extracted based on the Digital Elevation Model (DEM), which is used to define independent hydrological response units.

[0071] In the embodiments of the present application, first, the D8 flow direction algorithm is used to extract the watershed boundary from the 1m resolution LiDAR data. Secondly, the digital twin framework is cut into several sub-watersheds (such as the upper reaches of the main stream, the confluence area of tributaries, etc.) through the constrained Voronoi diagram. Thirdly, each sub-watershed is subdivided into 100m×100m surface flood evolution direction calculation units according to the slope change rate raster (graded from 0 to 15%). Finally, the adaptive grid encryption technology is applied to ensure that the coefficient of variation of the slope within the unit is less than 5%.

[0072] 402. Within the surface flood evolution direction calculation unit, according to the distribution characteristics of the slope turning points in the terrain slope change data, determine the initial offset angle of the surface flood evolution direction, and correct the initial offset angle in combination with the flood interaction parameters to generate a corrected path for the abnormal change of the flood evolution direction; In step 402, the distribution characteristics of the slope turning points refer to the slope mutation positions of the longitudinal profile of the river channel (such as steep banks, waterfalls, etc.). The initial offset angle is the angle between the theoretical water flow direction calculated by the D8 algorithm and the axis of the main river channel.

[0073] In the embodiments of the present application, first, curvature analysis is used to identify the slope turning points (such as the slope suddenly drops from 8% to 3% at X:120). Secondly, the least squares method is used to fit the initial offset angle of the water flow direction within the unit (such as the S1 section deflects 18°). Thirdly, based on flood interaction parameters such as the roughness coefficient of the river channel (0.025 - 0.035) and the flood storage capacity of the floodplain (500,000 - 600,000 m³), the offset angle is corrected through the Bayesian optimization algorithm (such as adjusted to 22°). Finally, a corrected path for bypassing the river bend is generated (the radius of curvature of the path ≥ 3 times the river width).

[0074] 403. Couple the critical threshold of the river water level in the flood interaction parameters with the distribution density of the river channel monitoring sections in the terrain slope change data. When the distribution density of the monitoring sections exceeds the critical threshold corresponding to the slope change rate, generate a flood inundation warning level based on the coupling result; In step 403, the critical threshold of the water level refers to the warning water level corresponding to the levee design standard. The distribution density of the monitoring sections reflects the number of monitoring points per unit river length (such as 2 sections per kilometer).

[0075] In the embodiment of the present application, first, an association matrix between the water level threshold and the slope change rate is established (for example, a slope of 8% corresponds to a water level threshold of 7.2 m). Second, the cross-section density index is calculated (for example, in the river section of X: 200 - 300, it reaches 3 per km). Third, when the density index exceeds the limit (> 2.5 per km) and the real-time water level exceeds the associated threshold. Finally, the flood inundation warning level is output through a logistic regression model (for example, red warning: water level exceeding the threshold + density exceeding the limit + slope > 10%).

[0076] 404. Perform a spatial overlay of the corrected path, the flood inundation warning level, and the meteorological forecast rainfall data, and conduct a three-dimensional flood evolution topology reconstruction on the spatial overlay result according to the boundary conditions of the watershed catchment area, so as to generate a watershed digital twin model that simultaneously includes the sudden change trajectory of the surface flood evolution direction and the flood inundation warning hot area of the river channel.

[0077] In step 404, the three-dimensional topology reconstruction refers to fusing multi-source data to construct a three-dimensional model of the watershed with hydraulic connectivity.

[0078] In the embodiment of the present application, first, the corrected path (in GeoJSON format), the warning level (raster data), and the WRF rainfall forecast (in NetCDF format) are uniformly projected onto the UTM coordinate system. Second, a Delaunay triangulation skeleton (node spacing ≤ 50 m) is generated based on the sub-watershed boundary. Third, the sudden change trajectory of the evolution direction is simulated through a GPU-accelerated water particle system (the particle size changes with the flow velocity). Finally, the warning hot area is rendered using the ray casting algorithm (the transparency is positively correlated with the warning level), forming an interactive watershed digital twin model.

[0079] The following is a specific example: Suppose during a rainstorm in a certain river basin, in step 401, 8 sub-watersheds are divided based on the catchment area boundary of 30 km² (elevation difference ≥ 50 m), and 572 calculation units are generated (slope variation coefficient ≤ 4.8%). In step 402, a slope turning point (5% → 1%) is identified in the tributary confluence area (X: 180), and the initial offset angle of 23° is corrected to 27° through the roughness coefficient (0.032), forming a corrected path (curvature radius 280 m) that bypasses the alluvial fan. In step 403, it is detected that the cross-section density in the middle reaches of the main stream is 3.2 per km (the threshold for a slope of 12% is 2.8 per km), and the water level of 7.9 m exceeds the associated threshold (7.6 m), triggering a red warning. In step 404, by fusing the corrected path (deflected by 27°), the red warning area (X: 150 - 250), and the forecast rainfall (60 mm in the next 3 h), the constructed watershed digital twin model accurately predicts the inundation range of this area after 6 hours (error rate < 5%), supporting the reservoir operation decision-making.

[0080] Steps 401-404 achieve high-precision digital twin modeling through the coupling of watershed terrain dynamic zoning and flood evolution parameters. Based on the intelligent correction of the water flow direction using slope turning characteristics, combined with water level thresholds and cross-section density to generate hierarchical warnings, and finally integrating multi-source data to construct a three-dimensional visualization model, which can accurately predict the flood evolution path and inundation risk, providing full-process intelligent support for watershed flood control decision-making.

[0081] In some embodiments, the continuous change process of the flood evolution path data is traced backward in chronological order, and at the same time, the spatial distribution state of the water level monitoring data at the river channel monitoring cross-section is matched. According to the backward tracing result and the matching result, the spatio-temporal correlation characteristics between the sudden change points of the flood flow velocity in the flood evolution path data and the abnormal fluctuation points in the water level monitoring data are constructed, including: 501. Traverse the continuous records of the flood flow velocity in the flood evolution path data in reverse chronological order, identify the sudden change points where the change amplitude of the flood flow velocity exceeds the preset sudden change threshold within adjacent timestamps, and record the occurrence time of each sudden change point and the corresponding river channel position coordinates as the backward tracing result; In step 501, the backward tracing result refers to the result of reverse analyzing the historical change process of the flow velocity from the current flood front position. The sudden change points reflect the abnormal changes in the flood carrying capacity of the river channel (such as levee breaches or tributary backwater).

[0082] In the embodiments of the present application, first, the dynamic time warping (DTW) algorithm is used to match the Sentinel-1 satellite flow velocity sequence (30-minute interval) in reverse chronological order. Secondly, the sudden change points where the flow velocity drops suddenly by more than the threshold (such as from 3.0 m / s to 1.0 m / s) are identified. Thirdly, adjacent sudden change points are merged through spatial clustering (DBSCAN algorithm) (for example, 3 sudden change points at X:250 are merged into a cluster). Finally, a structured backward tracing result (timestamp 15:00, coordinate X:250, flow velocity change amount 2.0 m / s) is generated.

[0083] 502. Screen the monitoring cross-section in the water level monitoring data that is closest to the river channel position coordinates in the backward tracing result, extract the cross-section water level change data within the preset time range before and after the occurrence time, and mark the abnormal fluctuation points where the water level change amplitude exceeds the historical average as the matching result; In step 502, the abnormal fluctuation point refers to the monitoring cross-section where the water level rises more than 2 times the standard deviation of the historical same period within 1 hour (such as the sudden rise of 1.8 m at section S2, Z = 4.0).

[0084] In the embodiments of the present application, first, the monitoring section closest to the mutation point is located based on the Voronoi diagram (e.g., X: 250 → section S2, distance 80 m). Secondly, the water level data for 1 hour before and after the mutation time (14:30 - 15:30) is extracted. Thirdly, the abnormal fluctuation points are marked through the sliding Z-Score detection (window length: data of the same period in 1 year) (the water level rises by 1.8 m, Z = 4.2). Finally, the matching result is recorded (section S2, the water level suddenly rises at 15:05, amplitude 1.8 m).

[0085] 503. Align and match the occurrence time in the reverse tracking result with the maximum fluctuation time of the abnormal fluctuation point in the matching result to construct a set of spatio-temporal correlation pairs between the mutation point and the abnormal fluctuation point; In step 503, the set of spatio-temporal correlation pairs includes the quantitative relationships of flow velocity mutation, water level fluctuation, and their spatio-temporal deviation.

[0086] In the embodiments of the present application, first, the time difference (5 minutes) between the mutation point (15:00) and the water level fluctuation (15:05) is calculated. Secondly, the flood wave propagation time is calculated based on the Saint-Venant equation (distance 80 m / wave velocity 1.6 m / s ≈ 50 s). Thirdly, the correlation strength is evaluated through the spatio-temporal coupling index (STCI) (e.g., if the time difference < conduction time × 1.2, the correlation is effective). Finally, a set of spatio-temporal correlation pairs with high confidence is generated (X: 250 flow velocity mutation ↔ S2 water level sudden rise, R² = 0.89).

[0087] 504. Combine the velocity change amount of the mutation point, the water level fluctuation amplitude of the abnormal fluctuation point, and the time interval between the two in each correlation pair in the set of spatio-temporal correlation pairs into spatio-temporal correlation features.

[0088] In step 504, the spatio-temporal correlation features are multi-dimensional vectors (e.g., [velocity change amount, water level amplitude, time difference]) that quantify the dynamic relationship between "flow velocity and water level" and are used for optimizing the model parameters.

[0089] In the embodiments of the present application, first, the correlation pair features are extracted (the flow velocity drops by 2.0 m / s at X: 250, the water level at S2 rises by 1.8 m, time difference 5 minutes). Secondly, it is converted to [1.0, 0.9, 0.2] through min-max normalization. Thirdly, based on the random forest feature weights (flow velocity 0.6, water level 0.3, time difference 0.1), the spatio-temporal correlation feature vector [0.6, 0.27, 0.02] is generated through weighting. Finally, it is stored in the feature library for real-time correction of the flood storage parameters in the digital twin model (e.g., the weight 0.6 corresponds to the flood storage capacity + 15%).

[0090] The following is a specific example: Suppose during a flood in the main stream of a certain river, at step 501, a sudden drop in the flow velocity from 2.8 m / s to 0.9 m / s (threshold 1.5 m / s) is detected at X:300 at 15:30. After spatial clustering, it is confirmed as a tributary backwater event. At step 502, the nearest cross-section S3 (distance 120 m) is matched, and it is found that the water level suddenly rises by 2.1 m (Z = 4.5) at 15:35, exceeding the historical extreme value. At step 503, the time difference of 5 minutes is calculated, the flood wave propagation time is 4 minutes (120 m / 3.0 m / s), and the STCI index is 0.92, which is determined to be strongly correlated. At step 504, a feature vector [1.9, 1.0, 0.25] is generated. After weight calculation, it triggers the update of the model parameters (the floodplain roughness coefficient is adjusted from 0.028 to 0.034), reducing the inundation prediction error of the downstream section X:400 - 500 from 12% to 3%.

[0091] Steps 501 - 504 construct a high-precision dynamic response model of basin floods by retrospectively tracking the spatio-temporal correlation between sudden changes in flood flow velocity and water level fluctuations. Based on the flood wave propagation theory and random forest feature weighting, the system intelligently identifies the causal relationship between abnormal points in river flood routing and sudden changes in water level, generates quantitative "flow velocity and water level" correlation features, and real-time optimizes the parameters of the digital twin model, significantly improving the accuracy of flood evolution prediction and providing a scientific decision-making basis for basin flood control and scheduling.

[0092] In some embodiments, coupling the channel water level balance value in the hydrological interaction parameters with the pipe network cross-section distribution density in the terrain slope change data, and when the monitoring cross-section distribution density exceeds the critical threshold corresponding to the slope change rate, generating a backwater warning level based on the coupling result includes: 601. Divide the terrain slope change data into regional units according to the slope change rate, count the total number of monitoring cross-sections in each regional unit, and calculate the cross-section distribution density per unit area; In step 601, the cross-section distribution density reflects the spatial coverage intensity of the water level monitoring cross-sections per unit basin area and is used to evaluate the risk of monitoring blind spots in the river flood routing capacity.

[0093] In the embodiments of the present application, first, based on LiDAR slope data (1 m resolution), the slope units are divided using the Jenks natural break method (such as 0 - 5%, 5 - 10%, >10%). Second, the number of cross-sections in each unit is counted through spatial join analysis (for example, the upstream unit of the main stream contains 8 cross-sections). Finally, the cross-section distribution density is calculated (for example, the density corresponding to a unit area of 5 km² is 1.6 / km²), and the low-density units (<1 / km²) are identified as monitoring weak areas.

[0094] 602. Set the cross-section density critical values corresponding to different slope change rate intervals according to the correlation law between the slope change rate and the cross-section distribution density; In step 602, the cross-section density critical value refers to the minimum cross-section configuration requirement to ensure the accuracy of flood monitoring. A higher density is required in the steep slope area to capture rapid water level changes.

[0095] In the embodiments of the present application, first, the correlation between slope and density in historical flood events is analyzed (for example, the ideal density in areas with a slope > 10% is ≥ 2 per km²). Secondly, a quantile regression is used to establish a dynamic threshold model (the density threshold increases by 0.5 per km² for every 5% increase in slope). Finally, a slope classification critical value table is generated to set the cross-section density critical values corresponding to different slope change rate intervals (for example, 0 - 5%: 1 per km², 5 - 10%: 1.5 per km², > 10%: 2 per km²).

[0096] 603. In the regional unit, couple the river water level critical threshold in the flood interaction parameters with the cross-section distribution density according to a preset ratio. If the cross-section distribution density exceeds the cross-section density critical value, calculate the product of the water level difference and the slope change rate between this region and adjacent regions based on the coupling result. In step 603, the product of the water level difference and the slope change rate quantifies the flood evolution prediction error caused by unreasonable cross-section layout. The larger the product value, the higher the risk of monitoring blind spots.

[0097] In the embodiments of the present application, first, extract the river water level critical threshold (such as the warning water level of section S2 is 7.5m) and the cross-section distribution density (such as 1.2 per km²) within the unit. Secondly, when the density is lower than the critical value (for example, the requirement in the area with a slope > 10% is 2 per km²), calculate the product of the water level difference between adjacent units (such as the water level of the upstream unit is 7.8m vs the water level of this unit is 7.2m, the difference is 0.6m) and the slope change rate (8%) (0.6 × 8 = 4.8). Finally, mark the units with a product > 3 as high-risk monitoring blind spots.

[0098] 604. Divide the three-level flood inundation warning levels according to the product value, where the first-level warning corresponds to the highest interval of the product value, the second-level warning corresponds to the medium interval, and the third-level warning corresponds to the lowest interval.

[0099] In step 604, the three-level flood inundation warning levels reflect the impact degree of monitoring blind spots on flood forecasting. For the first-level warning, temporary monitoring facilities need to be immediately supplemented.

[0100] In the embodiments of the present application, based on the historical disaster database (flood events in the past 10 years), a logistic regression model is used to train the correlation between the product value and the actual inundation depth, and the three-level warning thresholds are redefined: first level (product value > 8, corresponding to inundation depth ≥ 2m), second level (5 - 8, 1 - 2m), third level (< 5, < 1m), and the classification accuracy is verified through a confusion matrix (F1-score ≥ 0.85).

[0101] Here is a specific example: Assume that during a rainstorm in a river basin, step 601 divides the tributary confluence unit (slope 12%, area 3km²) into only 4 sections (density 1.3 / km²), which is lower than the critical value (2 / km²). Step 603 calculates the product of the water level difference (0.8m) between this unit and the main river unit and the slope change rate (7%) as 5.6. Step 604 determines that it is a first-level warning, and the system automatically dispatches 2 mobile monitoring stations to strengthen data collection, reducing the water level prediction error in the area from 15% to 6%.

[0102] Steps 601-604 construct an intelligent monitoring and early warning system through slope and density coupling analysis, dynamically identify flood prediction blind spots caused by insufficient section layout, quantify risk levels based on the product of water level difference and slope change, achieve precise scheduling of monitoring resources and real-time calibration of flood evolution models, improve water level prediction accuracy in key areas, and provide data quality assurance for basin flood control.

[0103] In some embodiments, the flood emergency control plan is generated according to the spatiotemporal distribution characteristics of the flood inundation warning signal in the watershed digital twin model, combined with the river terrain elevation data, and the spatiotemporal distribution characteristics include signal density, including: 701. Filtering, in the watershed digital twin model, continuous areas where the signal density of the flood warning signal exceeds a preset threshold, marking them as high-risk areas for flooding, and extracting the time duration of the high-risk areas; In step 701, the high-risk area for flood inundation refers to the river section where the warning signals are continuously clustered in space and the duration exceeds the limit, reflecting the core area of ​​the continuous threat of floods.

[0104] In the embodiment of the present application, spatiotemporal hot spot analysis (Getis-Ord Gi*) is first used to identify cluster areas with a warning signal density of >15 / 5km², then the discrete warning points are merged through a morphological expansion algorithm (3×3 kernel), and finally the duration of the risk zone from the start to the release is extracted (such as the X:200-300 river section continuously exceeding the warning for 6 hours).

[0105] 702. Match the spatial range of the high-risk flooding area with the river terrain elevation data, identify the weak sections in the high-risk area whose elevation is lower than the flood warning water level, and mark them as key sections for flood diversion and discharge; In step 702, the key flood diversion and discharge section refers to the river section where the embankment elevation is lower than the design flood level or there is a significant terrain depression, which requires priority engineering intervention.

[0106] In the embodiments of the present application, first, the boundary of the risk area is superimposed on the 1m LiDAR elevation data. Secondly, the dike sections below the warning water level (such as 7.5m) are extracted (for example, the elevation at X:250 is 7.2m). Thirdly, terrain depressions are detected through curvature analysis (curvature > 0.05). Finally, X:240 - 260 is marked as the key section for flood diversion and discharge.

[0107] 703. Generate a multi - level flood diversion channel plan covering the upstream and downstream based on the elevation gradient direction of the key section for flood diversion and discharge, where the capacity of the flood diversion channel is positively correlated with the elevation difference. In step 703, the multi - level flood diversion channel refers to a stepped flood discharge path designed according to the elevation difference between the upstream and downstream, and realizes staged flood detention through natural gravity.

[0108] In the embodiments of the present application, first, determine the elevation gradient direction (for example, the elevation rises 3m from X:250 to X:280). Secondly, calculate the designed discharge of the flood diversion channel based on the elevation gradient direction through a hydraulic model (the discharge increases by 50m³ / s for every 1m of elevation difference). Thirdly, generate a three - level flood diversion channel plan (main channel 800m³ / s + auxiliary channel 300m³ / s) in combination with the distribution of flood detention areas. Finally, verify the effectiveness of the channel through HEC - RAS simulation.

[0109] 704. Adjust the flood diversion gate control instructions for the key cross - sections of the river channel in the high - risk area through the time duration and the capacity change trend of the multi - level flood diversion channel plan, and generate a flood emergency regulation plan including the flood diversion channel activation rules and the gate regulation time sequence according to the adjusted control instructions.

[0110] Among them, in combination with a real - time social and economic impact assessment model (such as the vulnerability index of disaster - bearing bodies), dynamically adjust the flood diversion gate control instructions through the time duration and the capacity change trend of the flood diversion channel, and give priority to ensuring the safety of densely populated areas and key infrastructure.

[0111] In step 704, the gate regulation time sequence refers to the strategy of dynamically adjusting the opening of the flood diversion gate according to the flood evolution speed to achieve flood peak staggering regulation.

[0112] In the embodiments of the present application, first, establish an association model between the time duration and the capacity of the multi - level flood diversion channel plan (for example, 60% of the total reservoir capacity needs to be released when the warning level is exceeded for 6 hours). Secondly, use the model predictive control (MPC) algorithm to generate a gate instruction sequence based on the association model (for example, the opening of the gate in front of X:250 is 50% in the first 2 hours and 75% in the next 4 hours). Finally, integrate the flood diversion channel activation rules (main first and then auxiliary) and the gate time sequence to form a complete regulation plan.

[0113] The following is a specific example: Suppose the section of the main stream of a certain river basin from X:200 to 400 is marked as a high-risk area (lasting for 8 hours). Step 702 identifies that the section from X:300 to 320 is the key section for flood diversion and discharge (elevation 7.1m < 7.5m). Step 703 designs a three-level flood diversion channel (main channel 1000m³ / s + two auxiliary channels each 400m³ / s). Step 704 generates an instruction through the MPC algorithm: open the main channel in the first hour (gate opening 30%), and open the auxiliary channels starting from the 3rd hour (opening 50%). The cumulative flood diversion volume reaches 70% of the designed storage capacity within 8 hours, successfully controlling the downstream water level below the warning line.

[0114] Steps 701 - 704 accurately locate the key flood diversion areas through spatio-temporal hot spot identification and terrain matching, design multi-level flood diversion channels based on elevation gradients, and dynamically optimize gate scheduling in combination with model predictive control to achieve scientific management and control of flood risks, improve flood diversion efficiency, and provide intelligent decision-making support for flood control scheduling in the river basin.

[0115] Figure 2 The following is a schematic structural diagram of a system for constructing a digital twin of a river basin based on multi-source data fusion provided by an embodiment of the present application. As Figure 2 shown, the system includes: An acquisition module 21, configured to obtain water level monitoring data of the river channel, surface flood evolution path data, meteorological forecast rainfall data, and terrain slope change data based on the interaction relationship between the river channel structure and surface runoff in the river hydrological interaction scenario; A fusion module 22, configured to fuse the meteorological forecast rainfall data and the terrain slope change data to generate a distribution relationship between rainfall and flood runoff based on the terrain slope change, and determine the high-risk area of flood evolution in the distribution relationship according to the abnormal water level change characteristics of the water level monitoring data, where the abnormal water level change characteristics include the water level mutation amplitude at key monitoring sections in the river channel; An adjustment module 23, configured to establish a preliminary framework of a digital twin model of the river basin corresponding to the high-risk area of flood evolution, generate associated feature parameters through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and adjust the flood interaction parameters between the river channel structure and surface runoff in the preliminary framework based on the associated feature parameters; A construction module 24, configured to input the meteorological forecast rainfall data into the adjusted preliminary framework, and construct a digital twin model of the river basin including abnormal change data of the surface flood evolution direction and flood inundation warning signals of the river channel in combination with the flood interaction parameters and the watershed catchment boundary conditions in the terrain slope change data; A generation module 25, configured to generate a flood emergency control plan according to the spatio-temporal distribution characteristics of the flood inundation warning signals in the digital twin model of the river basin, in combination with the river channel terrain elevation data.

[0116] Figure 2 The described system for constructing a digital twin of a river basin based on multi-source data fusion can execute Figure 1 For the method for constructing a digital twin of a river basin based on multi-source data fusion described in the above embodiments, its implementation principle and technical effects will not be elaborated further. For the system for constructing a digital twin of a river basin based on multi-source data fusion in the above embodiments, the specific ways in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0117] In a possible design, Figure 2 The system for constructing a digital twin of a river basin based on multi-source data fusion in the above embodiments can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, where the one or more computer instructions are called and executed by the processing component 32.

[0118] The processing component 32 is used for the above Figure 1 The method for constructing a digital twin of a river basin based on multi-source data fusion in the above embodiments.

[0119] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0120] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0121] Of course, the computing device may also necessarily include other components, such as input / output interfaces, display components, communication components, etc.

[0122] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0123] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0124] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device can refer to a cloud server, and the above processing components, storage components, etc. can be basic server resources leased or purchased from a cloud computing platform.

[0125] The embodiments of the present application also provide a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above Figure 1 A method for constructing a digital twin of a river basin based on multi-source data fusion shown in the embodiments.

[0126] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0127] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0128] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solutions or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A method for constructing a watershed digital twin based on multi-source data fusion, characterized in that: include: Based on the interaction between river channel structure and surface runoff in river hydrological interaction scenarios, obtain river water level monitoring data, surface flood evolution path data, meteorological forecast rainfall data and terrain slope change data; The meteorological forecast rainfall data is integrated with the terrain slope change data to generate a distribution relationship between rainfall and flood runoff based on terrain slope change, and the high-risk area for flood evolution in the distribution relationship is determined according to the abnormal water level change characteristics of the water level monitoring data, wherein the abnormal water level change characteristics include the amplitude of water level mutation at a key monitoring section in the river channel; Establishing a preliminary framework of a watershed digital twin model corresponding to the high-risk area of ​​flood evolution, generating associated characteristic parameters through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and adjusting the flood interaction parameters between the river channel structure and the surface runoff in the preliminary framework based on the associated characteristic parameters; Inputting the meteorological forecast rainfall data into the adjusted preliminary framework, combining the flood interaction parameters with the watershed catchment boundary conditions in the terrain slope change data, and constructing a watershed digital twin model including abnormal change data of surface flood evolution direction and river flood inundation warning signals; According to the spatiotemporal distribution characteristics of the flood warning signal in the watershed digital twin model, a flood emergency control plan is generated in combination with the river terrain elevation data.

2. The method according to claim 1, characterized in that The preliminary framework of the digital twin model of the watershed corresponding to the high-risk area of ​​flood evolution is established, and the associated characteristic parameters are generated through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and the flood interaction parameters between the river structure and the surface runoff in the preliminary framework are adjusted based on the associated characteristic parameters, including: Based on the spatial distribution of high-risk areas for flood evolution, a preliminary framework of a watershed digital twin model including the river channel topology and surface flood evolution path is established; Reversely track the continuous change process of the flood evolution path data in chronological order, and match the spatial distribution state of the water level monitoring data at the river monitoring section, and construct the spatiotemporal correlation characteristics of the flood flow velocity mutation point in the flood evolution path data and the abnormal fluctuation point in the water level monitoring data according to the reverse tracking result and the matching result; The distribution density of flood velocity mutation points in the spatiotemporal correlation characteristics and the water level fluctuation amplitude are combined at multiple scales to establish correlation characteristic parameters reflecting the correlation between the flood evolution path change pattern and the water level response intensity, and the correlation characteristic parameters are mapped to the flood interaction parameters between the river channel structure and the surface runoff in the preliminary framework; According to the mapping results, combined with the change rate of the flood evolution path data at the boundary of the high-risk flood area and the continuous fluctuation amplitude of the water level monitoring data, the flood interaction parameters are dynamically updated, and the updated flood interaction parameters are spatially compensated through the spatial distribution range to adjust the preliminary framework.

3. The method according to claim 1, characterized in that The step of fusing the meteorological forecast rainfall data with the terrain slope change data to generate a distribution relationship between rainfall and flood runoff based on terrain slope change, and determining a high-risk area for flood evolution in the distribution relationship according to abnormal water level change characteristics of the water level monitoring data includes: Dynamically distribute the rainfall intensity values ​​in the meteorological forecast rainfall data according to the spatial distribution law of the terrain slope change data to form a rainfall spatial coverage density with the slope change rate as a weight factor; Based on the guiding effect of the slope change rate on the flow direction of surface floods, the rainfall intensity values ​​in the high slope change rate areas in the rainfall spatial coverage density are directed to the low-lying confluence areas, and combined with the distribution density of tributary intersection nodes in the surface flood evolution path, the distribution relationship between rainfall and flood runoff based on terrain slope changes is generated; Calculate the difference between the amplitude of the water level mutation in the abnormal water level change characteristic and the flood runoff intensity of the corresponding geographical area in the distribution relationship, mark the abnormal area where the slope change rate does not match the flood runoff intensity in the distribution relationship, and based on the product coefficient of the difference and the slope change rate, perform gradient compensation on the flood runoff intensity value of the abnormal area according to the slope change rate; The compensated flood runoff intensity value is superimposed on the spatial distribution of abnormal fluctuation points in the water level monitoring data, and the overlapping areas that simultaneously meet the slope change rate threshold, flood runoff intensity compensation value and water level abnormal amplitude are screened out as high-risk areas for flood evolution.

4. The method according to claim 1, characterized in that: The meteorological forecast rainfall data is input into the adjusted preliminary framework, and the flood interaction parameters are combined with the boundary conditions of the watershed catchment area in the terrain slope change data to construct a watershed digital twin model including abnormal change data of surface flood evolution direction and river flood inundation warning signals, including: Based on the boundary conditions of the watershed catchment area in the terrain slope change data, the adjusted preliminary framework is spatially cut to divide the surface flood evolution direction calculation units based on the terrain slope change rate; In the surface flood evolution direction calculation unit, the initial offset angle of the surface flood evolution direction is determined according to the distribution characteristics of the slope turning points in the terrain slope change data, and the initial offset angle is corrected in combination with the flood interaction parameter to generate a correction path for abnormal changes in the flood evolution direction; The critical threshold of the river water level in the flood interaction parameter is coupled with the distribution density of the river monitoring section in the terrain slope change data, and when the distribution density of the monitoring section exceeds the critical threshold corresponding to the slope change rate, a flood inundation warning level is generated based on the coupling result; The corrected path, the flood inundation warning level and the meteorological forecast rainfall data are spatially superimposed, and the three-dimensional flood evolution topology is reconstructed for the spatial superposition result according to the boundary conditions of the watershed catchment area to generate a watershed digital twin model that includes both the sudden change trajectory of the surface flood evolution direction and the river flood inundation warning hot zone.

5. The method according to claim 2, characterized in that: The continuous change process of the flood evolution path data is reversely tracked in chronological order, and the spatial distribution state of the water level monitoring data at the river monitoring section is matched. According to the reverse tracking result and the matching result, the spatiotemporal correlation characteristics of the flood velocity mutation point in the flood evolution path data and the abnormal fluctuation point in the water level monitoring data are constructed, including: Traversing the continuous records of flood flow velocity in the flood evolution path data in reverse chronological order, identifying mutation points where the change amplitude of flood flow velocity exceeds a preset mutation threshold within adjacent timestamps, and recording the occurrence time of each mutation point and the corresponding river channel position coordinates as a reverse tracking result; Selecting the monitoring section closest to the river channel location coordinates in the reverse tracking result from the water level monitoring data, extracting the section water level change data within a preset time range before and after the occurrence time, and marking the abnormal fluctuation points where the water level change amplitude exceeds the historical average as matching results; Aligning and matching the occurrence time in the reverse tracking result with the maximum fluctuation time of the abnormal fluctuation point in the matching result, and constructing a set of spatiotemporal association pairs of the mutation point and the abnormal fluctuation point; The velocity change of the mutation point, the water level fluctuation amplitude of the abnormal fluctuation point and the time interval between the two of each association pair in the set of spatiotemporal association pairs are combined into a spatiotemporal association feature.

6. The method according to claim 4, characterized in that The method of coupling the critical threshold of the river water level in the flood interaction parameter with the distribution density of the river monitoring section in the terrain slope change data, and generating a flood inundation warning level based on the coupling result when the distribution density of the monitoring section exceeds the critical threshold corresponding to the slope change rate, includes: Divide the terrain slope change data into regional units according to the slope change rate, count the total number of monitoring sections in each regional unit, and calculate the section distribution density per unit area; According to the correlation law between the slope change rate and the cross-section distribution density, the cross-section density critical values ​​corresponding to different slope change rate intervals are set; In the regional unit, the critical threshold of the river water level in the flood interaction parameter is coupled with the cross-section distribution density according to a preset ratio. If the cross-section distribution density exceeds the cross-section density critical value, the product of the water level difference between the region and the adjacent region and the slope change rate is calculated based on the coupling result; According to the product value, three levels of flood inundation warning are divided, where the first level warning corresponds to the highest range of the product value, the second level warning corresponds to the middle range, and the third level warning corresponds to the lowest range.

7. The method according to claim 1, characterized in that The method of generating a flood emergency control plan based on the spatiotemporal distribution characteristics of the flood warning signal in the watershed digital twin model and combining the river terrain elevation data includes: In the watershed digital twin model, continuous areas where the signal density of the flood warning signal exceeds a preset threshold are screened, marked as high-risk areas for flooding, and the time duration of the high-risk areas is extracted; Matching the spatial scope of the high-risk flooding area with the river terrain elevation data, identifying the weak sections in the high-risk area whose elevation is lower than the flood warning water level, and marking them as key sections for flood diversion and discharge; Based on the elevation gradient direction of the key flood diversion and discharge section, a multi-level flood diversion channel scheme covering the upstream and downstream is generated, wherein the capacity of the flood diversion channel is positively correlated with the elevation difference; By adjusting the duration of the time and the capacity change trend of the multi-level flood diversion channel scheme, the flood diversion gate control instructions of the key sections of the river in the high-risk area are adjusted, and a flood emergency control plan including flood diversion channel activation rules and gate control timing is generated according to the adjusted control instructions.

8. A watershed digital twin construction system based on multi-source data fusion, characterized in that: include: The acquisition module is used to obtain river water level monitoring data, surface flood evolution path data, meteorological forecast rainfall data and terrain slope change data based on the interaction relationship between river channel structure and surface runoff in the river hydrological interaction scenario; A fusion module is used to fuse the meteorological forecast rainfall data with the terrain slope change data to generate a distribution relationship between rainfall and flood runoff based on terrain slope change, and determine the high-risk area for flood evolution in the distribution relationship according to the abnormal water level change characteristics of the water level monitoring data, wherein the abnormal water level change characteristics include the amplitude of water level mutation at a key monitoring section in the river channel; An adjustment module is used to establish a preliminary framework of a watershed digital twin model corresponding to the high-risk area of ​​flood evolution, generate associated characteristic parameters through the reverse feedback relationship between the flood evolution path data and the water level monitoring data, and adjust the flood interaction parameters between the river structure and the surface runoff in the preliminary framework based on the associated characteristic parameters; A construction module is used to input the meteorological forecast rainfall data into the adjusted preliminary framework, combine the flood interaction parameters with the watershed catchment boundary conditions in the terrain slope change data, and construct a watershed digital twin model including abnormal change data of surface flood evolution direction and river flood inundation warning signals; A generation module is used to generate a flood emergency control plan based on the spatiotemporal distribution characteristics of the flood inundation warning signal in the watershed digital twin model and combined with the river terrain elevation data.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a watershed digital twin construction method based on multi-source data fusion as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a method for constructing a watershed digital twin based on multi-source data fusion as described in any one of claims 1 to 7 is implemented.

Images