基于ICTS的水文地质参数并行寻优机制研究

摘要

连续禁忌搜索改进算法(ICTS)是一种启发式全局迭代寻优算法,在水文地质参数反演计算中具有收敛速度快、求解精度高、寻优能力强的优点.该文利用MPI函数库对ICTS算法的初始化、参数设置、邻域搜索等环节实行不同的并行化策略,提出三种分别基于解空间划分(PICTS-1)、邻域分解(PICTS-2)和多任务搜索(PICTS-3)的并行连续禁忌搜索算法(PICTS),并以水文地质参数寻优问题为例,在两台计算机并行计算环境下验证并行PICTS算法的有效性.算例研究表明,三种PICTS算法的加速比分别为1.91、1.76和1.26,可根据需要灵活使用,具有较强的实践意义.%Improved continuous tabu search algorithm (ICTS) is a meta-heuristic global iterated optimization procedure, which has great advantages with fast convergence speed, high solving precision and strong optimization applicability in hydrogeologic parameters inversion. Using different parallel strategies on initialization, parameters settings, neighborhood search and other sectors of ICTS algorithm with MPI function library, we propose three parallel continuous tabu search (PICTS) algorithm separately on the solution space (PICTS - 1), neighborhood decomposition (PICTS - 2) and multi-task search (PICTS - 3). Finally we apply PICTS algorithms to hydrogeological parameters identification so as to test the validity with parallel computing of two computers. The results show that the speedups of three PICTS algorithms are 1.91,1.76 and 1.26.Based on different needs, we can flexibly choose specific PICTS algorithm,which has good performance in solving practical problems.