粗粒度并行遗传算法在水库调度问题中的应用  

Application of coarse-grained genetic algorithm to reservoir operation

作　　者：李想[1] 魏加华[1] 傅旭东[1] 

机构地区：[1]清华大学水沙科学与水利水电工程国家重点实验室,北京100084

出　　处：《水力发电学报》2012年第4期28-33,共6页Journal of Hydroelectric Engineering

基　　金：国家"十一五"科技支撑计划项目(2009BAC56B03;2008BAB29B09)

摘　　要：巨型水库群实时优化调度涉及大规模、高维及非线性问题,庞大、动态、复杂的搜索空间,采用传统遗传算法(GA)求解该类问题几乎不可行,并行遗传算法(PGA)除具有GA的优势外,还能充分利用并行计算机的计算能力、有效地提高求解质量和求解速度,在解决巨型水库群优化调度问题方面具有广阔的应用前景。本文采用PGA粗粒度模型(CGGA),引入迁移算子,以三峡-葛洲坝梯级水库为例,将基于双向环迁移拓扑的CGGA应用于水库调度模型求解。计算结果表明,CGGA能够有效地提高求解质量和求解速度,从收敛性能看,CGGA由于种群隔离保证了种群间的个性,能够在整个计算过程中不断进化,避免总群体趋于同化;从并行性能看,CGGA加速比远大于线性加速比,说明CGGA能够充分地利用各计算进程,提高并行效率,避免资源浪费。Real-time optimal operation of giant reservoirs is characterized by large-scale, high-dimensional and nonlinear problems. Large, dynamic and complex search space makes it almost impossible to solve such problems with a simple genetic algorithm. In addition to the advantage of simple genetic algorithm, parallel genetic algorithm can make full use of the computing power of parallel computers in improving solution quality and increasing solution speed, showing a promising prospect in resolving these difficult problems. This paper applies a coarse-grained genetic algorithm based on bi-directional ring topology to optimal operation of the Three Gorgest-Gezhouba cascade reservoirs. Results show that this algorithm can effectively improve the solution quality and increase the solution speed. In terms of convergence performance, it ensures the personality of each population through population isolation and allows populations to evolve constantly throughout the calculations, thus avoiding assimilation of the populations. In terms of parallel computing efficiency, the speedup of CGGA is much greater than the linear one, indicating its full use of calculation processes with less resources waste.

关 键 词：并行遗传算法 水库调度 粗粒度模型 三峡-葛洲坝梯级 

分 类 号：TV697.1[水利工程—水利水电工程]