基于格子Boltzmann高性能并行计算的城市内涝模型研究

随着城市化的加快和极端气候的增加，我国城市遭受内涝的频率和规模正逐年扩大，这严重制约了城市的正常发展。为了防灾减灾，开发精确、高效的城市内涝模拟模型非常迫切。针对精细模拟城市洪水演进具有边界条件复杂、存在非连续地形、干湿地面相互交替以及现有计算机的计算能力有限的难题，本项目基于格子Boltzmann方法兼具微观本质和介观特点、算法简单、边界条件易于实现和适合并行计算的突出优点，开发出一套耦合地下排水管网的高效、稳定的城市内涝数值模型。同时，进一步扩展该模型，从而在GPU上实现并行计算。在采用相应的实验验证后，并行版本的城市内涝模型用来模拟典型的城市内涝过程，并分析内涝的特点。本项目不仅为更深入探讨城市内涝灾害特性提供了一种新的研究手段，而且可为防止和减少城市内涝灾害提供技术支撑，具有重要的研究价值。

With the increase of urbanization and extreme climate, our city are suffering waterlogging more and more frequently and the scale of waterlogging also increases year by year. Waterlogging disasters have seriously hampered the normal development of the city. In order to disaster prevention and mitigation, it is very urgent to develop accurate and efficient model of simulating city waterlogging. There are problems of complex boundary conditions, existence of discontinuous terrain, alternate dry-wet ground and limited computing power when urban flood is simulated in detail. Based on the outstanding advantages of lattice Boltzmann method with microscopic and mesoscopic characteristics, simple algorithm, easy implementation of boundary conditions and suitable characteristics for parallel computing, this project is to develop an efficient and stable numerical model coupling with underground drainage pipe network for simulating city waterlogging. Furthermore, the model is extended to realize parallel computing on GPU. After validated by the corresponding experiments, the parallel model is used to simulate the process of waterlogging in a typical city and analyze the characteristics of waterlogging. This project not only provides a new method for studying the characteristics of city waterlogging disaster in-depth, but also can provide technical support for preventing and reducing waterlogging disaster in city. Therefore, this project has an important research value.