Simulation And Early Warning Of Urban Rainstorm Waterlogging Based On Hydrodynamics

With the acceleration of urbanization,the frequency and intensity of extreme weather have increased significantly,leading to the prominent problem of urban rainstorm waterlogging,which seriously threatens the safety of people’s lives and property.Rainstorm simulation and early warning is an important means to reduce the risk of urban flood disasters.In urban rainstorm waterlogging simulation,the rainstorm model based on two-dimensional hydrodynamic model is an important tool to realize the refinement simulation of urban flood process.However,this model method has the characteristics of large amount of calculation and high requirement for calculation stability,which restricts its application in practical engineering.This study focuses on the simulation accuracy,speed and early warning platform of urban rainstorm model.Based on high-precision topographic data,an urban rainstorm refinement model is constructed.Parallel acceleration technology is used to improve the computational speed of the model.On the basis of the model,a B/S urban rainstorm simulation and early warning platform is built using Cesium framework.The main research conclusions are as follows:(1)Based on a two-dimensional hydrodynamic model,integrating runoff and dual-layer drainage modules,a refined simulation model for urban rainstorm was established to achieve minute-level time resolution and meter-level spatial resolution in the simulation of the rainstorm process.Through testing with multiple classic cases,the model showed good accuracy and stability.In the case of Nangang District in Harbin,the simulated inundation area was consistent with the measured data in terms of spatial distribution under the condition of a time step of 0.5 s and a grid resolution of 2 m.with an average error of 6.9 cm.(2)Four parallel acceleration technologies,OpenMP,MPI,OpenACC,and CUD A,were used to improve the computational speed of the urban rainstorm refinement model.Among them,the CUDA-based parallel model has the best acceleration effect.It achieved an acceleration ratio of 120 times under 2 million grids and completed a 4hour rainstorm process simulation within 15 minutes,significantly increasing the urban rainstorm warning period.When the number of grids is small(within 300,000 grids),the OpenMP parallel technology has the advantages of simple implementation and high acceleration efficiency compared to other acceleration technologies.(3)A fast urban rainstorm prediction model based on graph convolutional neural network was developed.After training with a large number of learning samples,this model can obtain urban rainstorm inundation results within a few seconds.For urban street cases with simple topological structures,the model’s accuracy in judging inundation(non-inundation)reached over 99%.For the rainstorm case in Nangang District of Harbin,the model’s inundation accuracy was over 85%,which can be used as an effective tool for rapid simulation and early warning of rainstorms.(4)A B/S three-dimensional urban rainstorm simulation and early warning platform was developed,integrating the urban rainstorm refinement model and fast prediction model to achieve scientific visualization of water depth and flow velocity.In the case application in Nangang District of Harbin,B/S visualization of scalar fields such as inundation water depth and risk distribution and IBFV vector field visualization of rainstorm flow velocity were realized.It can display smoothly at a frame rate of 60 frames under 2 million grids on the Web side,providing technical support for urban rainstorm early warning.