基于多轻量线程的电力系统潮流计算加速方法  

A Method for Accelerating Power Flow Calculation Based on Multiple Light Threads

作　　者：杨罡[1] 喻乐[2] 刘明光[1] 

机构地区：[1]北京交通大学电气工程学院,北京市海淀区100044 [2]国家电网公司华北分部调控分中心,北京市西城区100053

出　　处：《电网技术》2013年第6期1666-1671,共6页Power System Technology

摘　　要：在牛顿-拉夫逊法的基础上,充分利用图形处理器(graphic processing unit,GPU)中数量庞大的轻量线程,实现了电力系统潮流计算的一种并行加速方法。在统一计算设备架构(compute unified device architecture,CUDA)环境下,考虑不同存储器的访问时延,实现了雅可比矩阵的并行更新方法和修正方程的并行迭代解法。最后对该方法进行了测试,相比串行方式可获得最高为10.8的加速比,为大规模电力系统潮流计算提供了一种有效加速方法。On the basis of Newton-Raphson method, a parallel acceleration method for power system power flow calculation is implemented, in which plenty of light threads in Graphic Processing Unit （GPU） are utilized. In the environment of compute unified device architecture （CUDA）, different access latency of memories are considered and parallel updating of Jacobian matrix and paratlel iteration solution of correction equation are realized. Finally the proposed method is tested. Contrasting to serial mode, using the proposed method an acceleration ratio up to 10.8 can be attained, so it is an effective way to accelerating the calculation of power flow.

关 键 词：潮流计算 CUDA 并行加速 图形处理器 

分 类 号：TM744[电气工程—电力系统及自动化]