EFEN-SP_3方法的加速与并行计算研究  

Acceleration and Parallelization Calculation of EFEN-SP_3 Method

作　　者：杨文[1] 郑友琦[1] 吴宏春[1] 曹良志[1] 李云召[1] 

机构地区：[1]西安交通大学核科学与技术学院,陕西西安710049

出　　处：《原子能科学技术》2013年第B12期664-667,共4页Atomic Energy Science and Technology

基　　金：国家自然科学基金资助项目(11105104);863计划资助项目(2013AA051402)

摘　　要：针对EFEN-SP3方法用于全堆芯pin-by-pin计算的计算效率仍有待进一步提高的问题,本文研究了粗网加速方法和空间并行算法。其中,粗网加速方法以空间、角度和能群在粗网内的中子守恒关系为基础,通过求解考虑粗网不连续因子的粗网再平衡方程来实现迭代计算的加速;基于MPI的空间并行算法,通过设计合理的并行方案保证负载平衡和最小化通信量,以充分发挥并行CPU的计算和存储能力,在保证并行效率的前提下进一步提高程序的计算效率。商用核电站堆芯算例的数值结果表明:粗网加速的加速比约为40;空间并行可在40个核的情况下保持60%以上的并行效率;在采用粗网加速的48核(主频为2.40GHz)并行计算条件下,4群、空间网格为289×289×218的压水堆全堆芯pin-by-pin计算用时约为100s。Due to the fact that the exponential function expansion nodal-SP3 （EFEN-SP3） method needs further improvement in computational efficiency to routinely carry out PWR whole core pin-by-pin calculation, the coarse mesh acceleration and spatial parallelization were investigated in this paper. The coarse mesh acceleration was built by considering discontinuity factor on each coarse mesh interface and preserving neutron balance within each coarse mesh in space, angle and energy. The spatial parallelization based on MPI was implemented by guaranteeing load balancing and minimizing communications cost to fully take advantage of the modern computing and storage abilities. Numerical results based on a commercial nuclear power reactor demonstrate an speedup ratio of about 40 for the coarse mesh acceleration and a parallel efficiency of higher than 60% with 40 CPUs for the spatial parallelization. With these two improvements, the EFEN code can complete a PWR whole core pin-by-pin calculation with 289×289×218 meshes and 4 energy groups within 100 s by using 48 CPUs （2.40 GHz frequency）.

关 键 词：EFEN—SP3方法 全堆芯pin-by—pin计算 粗网加速 空间并行 

分 类 号：TL329[核科学技术—核技术及应用]