基于GPU的电磁暂态仿真可行性研究  

Feasibility study of GPU based electromagnetic transient simulation

作　　者：陈来军[1] 陈颖[1] 许寅[1] 梅生伟[1] 

机构地区：[1]清华大学电机系电力系统国家重点实验室,北京100084

出　　处：《电力系统保护与控制》2013年第2期107-112,共6页Power System Protection and Control

基　　金：国家自然科学基金(51277104;51207076)~~

摘　　要：由于CPU体系结构的制约,基于CPU的电磁暂态仿真正面临性能提升的瓶颈。近年来,GPU因其巨大的计算潜力在通用计算领域正受到越来越多的关注。从硬件结构和软件编程两个方面分析了GPU计算的特点,设计了基于CPU-GPU混合编程模型的电磁暂态仿真程序并构建了原型仿真系统。测试结果表明,当系统规模不断扩大时,GPU在电磁暂态仿真中的效率优势将越来越明显。在此基础上,结合仿真中关键环节的耗时情况,分析了影响仿真性能的关键因素,提出了改进GPU仿真效率的若干对策。Electromagnetic transient simulation based on CPU is now facing the difficulty of performance improvement due to the constraints of CPU architecture. In recent years, GPU based simulation has attracted more and more attention in the field of general purpose computing for its great potential in solving large scale problem. In this paper, the characteristics of GPU computing are analyzed from the view of hardware structure and software programming. An electromagnetic transient simulation test bed is constructed in a GPU computing environment. Power systems with various scales are used to test the performance. Test results show that the efficiency advantage of GPU-based electromagnetic transient simulation becomes more and more obvious as the scale of the power system increases. Furthermore, the key factors which affect the performance of GPU-based electromagnetic transient simulation are investigated, which provides a reference for relative researchers to speed up their simulations.

关 键 词：GPU 电磁暂态仿真 并行计算 混合编程 效率分析 

分 类 号：TM743[电气工程—电力系统及自动化]