水文模拟中并行参数优化算法  

Parallel algorithm for parameter optimization of hydrologic simulation

作　　者：申蒙蒙 陆忠华[1] 王彦棡[1] 

机构地区：[1]中国科学院计算机网络信息中心超级计算中心,北京100190 [2]中国科学院大学,北京100049

出　　处：《计算机工程与设计》2017年第4期1002-1007,共6页Computer Engineering and Design

摘　　要：针对水文模拟中参数优化问题,提出一整套合理的并行参数优化算法。改进优化算法中的相关算子,提高算法的搜索效率;结合遗传算法、模拟退火算法、粒子群算法和差分进化算法4种算法,进一步提高算法的搜索效率;使用基于MPI(message passing interface)通信子空间的双层主-从并行框架完成实现,减少算法的运行时间。通过调用SWAT(soil and water assessment tool)、HIMS(hydro informatic modeling system)两大模型的参数优化过程进行实验,实验结果表明,该方法能够有效提高Nash-Suttcliffe系数和运行速度,提升了水文模拟中参数优化过程的效率。A set of reasonable parallel optimization algorithm was proposed to optimize the parameters of hydrological simulation. The search efficiency of the algorithm was increased by improving some operators, and higher efficiency was achieved by combi- ning various optimization algorithms. The running time of the algorithm was reduced using the double layer master-slave parallel framework based on communicator in MPI （message passing interface）. Extensive simulations were conducted by invoking SWAT （soil and water assessment tool） and HIMS （hydro informatic modeling system）. Results show that the proposed method can improve Nash-Suttcliffe coefficient and running speed. The efficiency of optimizing parameters in hydrological simulation is increased highly leveraging the proposed method.

关 键 词：水文模拟 水土评价工具模型 水循环综合模拟系统 参数优化算法 主-从并行框架 

分 类 号：TP301.5[自动化与计算机技术—计算机系统结构]