多目标系统最优控制方法研究  

Research on Optimum Control of Multi-Objective System

作　　者：杜振华[1,2] 谌海云[1] 曾欢[1] 石明江[1] 

机构地区：[1]西南石油大学电气信息学院,成都610500 [2]中海油节能环保服务有限公司,天津300457

出　　处：《航天控制》2014年第5期3-10,15,共9页Aerospace Control

基　　金：四川省教育厅重点科研项目(12ZA193)

摘　　要：用多个性能指标评价系统更有实际意义,但在现有的研究中,针对此类问题所采取的方法通常是线性加权,权值是人为选取以达到"最优",这就使得"最优"控制成为主观最优而非数值实际最优。本文针对这一问题,结合决策者先验知识的多少,给出2种新的解决方法:层次分析和Pareto交互式决策。首次采用免疫克隆选择多目标优化算法来设计多指标情况下的LQR控制器。仿真结果表明:该算法在时间复杂度、分布性等方面均优于现在常用的NSGA-II,且本文给出的解决方法更适合于复杂情况下的最优控制器设计。It has more practical significance to evaluate system with multiple performance indexes. In the existing researches, the approach for this type of problem is usually the linear weighted, and the weights are artificially selected to be ＂optimal＂. It makes ＂optimum＂ control subjective rather than the actual value op- timal. According to this problem, by combining with the quantity of the prior knowledge of decision-makers, two new solutions are proposed in this paper： known as analytic hierarchy process and Pareto interactive de- cision. It is the first attempt to apply immune clonal selection MO optimization algorithm to a multi-index case LQR controller design. The simulation results show that the algorithm is superior to NSGA-II in the as- pects of time complexity, distribution and the others. The solution proposed in this paper is more suitable for the optimal controller design in complex condition.

关 键 词：多目标系统 最优控制 层次分析法 交互式决策 免疫克隆选择多目标优化算法 

分 类 号：TP273.3[自动化与计算机技术—检测技术与自动化装置]