基于需求驱动的CSPS系统多目标优化控制  

Multi-objective optimization control of demand-driven CSPS system

作　　者：谭琦[1] 胡知强 唐昊[1] 戴飞 TAN Qi;HU Zhiqiang;TANG Hao;DAI Fei(School of Electrical Engineering and Automation,Hefei University of Technology,Hefei 230009,China)

机构地区：[1]合肥工业大学电气与自动化工程学院,合肥230009

出　　处：《系统工程理论与实践》2020年第4期1069-1079,共11页Systems Engineering-Theory & Practice

基　　金：国家重点研发计划(2017YFE0129700);国家自然科学基金(61573126);中央高校基本科研业务费专项资金(JZ2017YYPY0261)。

摘　　要：本文以最小化顾客流失率和最小化成品库平均在库数量为优化目标,研究基于需求驱动的传送带给料加工站(CSPS)系统的多目标优化问题.给出了前视距离控制和库存控制两种控制方式,并建立基于需求驱动的CSPS系统多目标优化模型.通过半马尔可夫决策过程长期平均性能公式计算目标值,然后使用NSGAⅡ和SPEA2两种多目标优化算法求解问题的最优策略集合.实验结果表明,加入库存控制方式可以在保证顾客需求的同时有效降低成品库平均在库数量.此外,针对模型特点,给出二进制-实数编码和连续化实数编码两种编码方式.通过验证,连续化实数编码可以有效提高算法的收敛性能及获得的非支配解集的质量.This paper considers the multi-objective optimization problem of demand-driven conveyorserviced production station(CSPS),the objectives are to minimize the customer loss rate and the average number of products in the product bank.We first establish a multi-objective optimization model of the demand-driven CSPS system including look-ahead control and inventory control.By computing the value of two objectives with long-term average performance formula of the semi-Markov decision process,we use two multi-objective optimisation methods,namely,non-dominated sorting genetic algorithm(NSGAII)and strength pareto evolutionary algorithm(SPEA2) to find the pareto-optimal front of the problem.The experimental results show that the addition of inventory control method effectively reduces the average number of products in the product bank while ensuring customer demand.In addition,this paper gives two encoding methods,binary-real coding and continuous real coding according to the characteristics of the model.Through verification,continuous real number coding can effectively improve the convergence performance of the algorithm and the quality of the obtained non-dominated solution set.

关 键 词：需求驱动 传送带给料加工站 多目标优化 半马尔可夫决策过程 

分 类 号：TP278[自动化与计算机技术—检测技术与自动化装置]