|
| SERGHEI-SWMM：并行计算与可移植性城市洪涝模型及其应用 |
| SERGHEI-SWMM: Parallel computing and portability-based urban flood model and its Application |
| 投稿时间：2025-01-16 修订日期：2025-07-01 |
| DOI： |
| 中文关键词: 城市洪涝 耦合模型 Kokkos 可移植性 并行计算 CPU/GPU异构计算 |
| 英文关键词: urban flood coupled model Kokkos model portability parallel computing CPU/GPU heterogeneous computing |
| 基金项目:国家重点研发计划（2022YFC3803001）；国家自然科学基金青年项目（42207057）；江苏大学高级人才基金资助项目（JDKQ20240405） |
|
| 摘要点击次数: 24 |
| 全文下载次数: 0 |
| 中文摘要: |
| 随着气候变化的不确定性日益加剧，极端天气引发的洪涝灾害已成为制约城市韧性提升的重要因素。城市洪涝模拟模型作为评估洪涝灾害风险、支撑城市防灾减灾系统、提升城市应急管理能力的关键工具，其计算方法已趋于成熟。然而，当前广泛应用于城市洪涝模拟的管道排水与地表径流耦合模型在多样化的高性能计算硬件架构上缺乏良好的可移植性。本研究基于Kokkos异构并行计算框架，构建了一维管道排水与二维地表径流耦合的水动力模型SERGHEI-SWMM。首先，通过基准算例验证，SERGHEI-SWMM模拟结果与InfoWorks ICM等模型的相对差值均低于9%。随后，以同济大学校园为研究区域，建立了其一维管道与二维地表耦合的洪涝模型，并结合洪涝实测数据对模型进行率定及验证。结果表明，该模型能够准确模拟洪涝过程中地表径流与地下管流之间的交换以及积水的时空演变特征，在不同CPU与GPU硬件架构上均表现出良好的并行性能与可移植性。本研究成果可为城市洪涝灾害风险评估提供高效、可靠的技术支撑，也为后续城市洪涝的实时预报预警奠定了基础。 |
| 英文摘要: |
| With the increasing uncertainty of climate change, flood disasters caused by extreme weather events have become one of the key issues affecting the development of urban resilience systems. Urban flood simulation models, as important tools for assessing flood risk, supporting urban disaster prevention and mitigation systems, and enhancing emergency management capabilities, have seen relatively mature development in terms of computational methodologies. Currently, coupled drainage and surface runoff models are extensively used for urban flood simulations; yet these models lack robust portability across diverse high-performance computing architectures. In this study, a coupled hydrodynamic model SERGHEI-SWMM is proposed to integrate one-dimensional pipe flow simulation and two-dimensional surface runoff simulation using the Kokkos heterogeneous parallel computing framework. The accuracy and reliability of SERGHEI-SWMM are first validated through a benchmark test against established models such as InfoWorks ICM, with the relative differences remaining below 9%. Subsequently, we implement a detailed coupled model for an urban flood scenario within Tongji University campus. After calibration and validation, our results demonstrate that SERGHEI-SWMM accurately simulates interactions between surface runoff and subsurface pipe flow, as well as the spatiotemporal evolution of inundation during flood events. Performance analysis further illustrates that SERGHEI-SWMM achieves high computational efficiency on different CPU and GPU platforms. Overall, this study demonstrates that the SERGHEI-SWMM model offers excellent computational accuracy, reliability, and portability. It thus provides a robust technical foundation for efficient urban flood risk assessment and sets the stage for future real-time flood forecasting. |
|
查看/发表评论 下载PDF阅读器 |
| 关闭 |