摘要
智能电网技术的发展需要快速电磁暂态程序(EMTP),而日益广泛应用的图像处理器(GPU)为电磁暂态仿真提供了高效的仿真环境和平台。文中首先提出了细粒度并行算法的运算级并行策略,即基于单指令多数据流(SIMD)的运算级并行策略和基于共享内存的运算级并行策略。随后,设计了应用这两种并行策略的改进电磁暂态细粒度并行算法。三相脉宽调制(PWM)变流器仿真测试表明,适用于GPU的细粒度并行算法能够在保证仿真正确性的同时,显著提高仿真效率,从而验证了基于GPU的细粒度并行仿真算法适用于带有开关过程和复杂控制的大规模电力系统快速电磁暂态仿真应用的可行性。
Development of the smart grid technology requires fast electromagnetic transient simulation, which is provided with a highly efficient simulation environment and platform by ever-increasing widely-used graphic processing unit (GPU). First, this paper puts forward two fine grained parallel operational level strategies, which are parallelisms based on single instruction multiple data (SIMD) and shared memory. Next, fine-grained parallel electromagnetic transient simulation algorithms for GPU are designed, and simulation procedures are implemented by adopting the two strategies in the improved electromagnetic transient algorithm. Test results of pulse width modulation (PWM) converters show that, the fine-grained parallel simulation algorithm for GPU enhances simulation efficiency enormously while maintaining accuracy. It' s also shown that the algorithm is especially useful when dealing with large system simulation problems.
出处
《电力系统自动化》
EI
CSCD
北大核心
2014年第6期43-48,79,共7页
Automation of Electric Power Systems
基金
国家自然科学基金资助项目(51277104
51207076)
清华大学自主科研计划资助项目(2012Z02140)~~
关键词
脉宽调制变流器
电磁暂态
细粒度并行
图像处理器(GPU)
pulse width modulation (PWM) converters
electromagnetic transient
fine-grained parallel
graphic processingunit (GPU)