摘要
针对层流冷却系统粗调区多目标优化问题,对Pareto前沿面进行了理论分析和证明,提出了基于模式提取优化的多目标遗传算法,用来搜索粗调区集管的最佳开闭模式集。该算法挖掘Pareto前沿面交集中的较优模式并建立模式库,将其中符合模式定理的模式进行耦合传承,以消除种群进化的随机漫游性;模式库的淘汰机制和基于集管开闭模式的拥挤距离策略,维持了种群解的多样性,有利于在更广的空间去搜索更优的集管开闭模式;模式的随机抽取为Pareto前沿面在空间的均匀分布奠定了基础;模式库的记忆、固化功能是驱动算法收敛于真实的Pareto最优解集的强力引擎。最后使用微软基类库编写了仿真程序,给出了该多目标优化控制策略明显优于常规的层流冷却喷水模式的实验结果。
Modular In view of multi-objective optimization problem of coarse control zone in laminar cooling system, this paper investigates the Pareto frontier and proposes optimized multi-objective genetic algorithm based on transgenic, which is used to search the best opening and closing pattern set of headers in coarse control zone. The algorithm digs into the intersection of ancient populations of the Pareto frontier to extract optimum mode and set up gene banks, and couples and inherits pattern theorem mode in order to eliminate the population evolution of random roaming property. Elimination mechanism of pattern bank and crowded distance strategy based on manifolds open and close mode maintain the population diversity of solution, which is conducive to a broader space to search better manifolds open and close mode. Random extraction for the Pareto frontier in the space of the model of uniform distribution laid a foundation; Memory, curing function of pattern bank is the powerful engine of driven algorithm converges to the true Pareto optimal solution set. The simulation program is written by using Microsoft Foundation Classes. Simulation results show that the multi-objective optimal control strategy was superior to the conventional laminar cooling spray pattern.
出处
《控制工程》
CSCD
北大核心
2016年第1期117-123,共7页
Control Engineering of China
基金
北京市重点学科建设项目(XK100080537)
