基于改进离散粒子群算法的炼钢连铸最优炉次计划  

Optimum charge plan of steelmaking continuous casting based on the modified discrete particle swarm optimization algorithm

作　　者：薛云灿[1,2] 郑东亮[1,2] 杨启文[1,2] 

机构地区：[1]河海大学江苏省输配电装备技术重点实验室,江苏常州213022 [2]河海大学计算机与信息学院,江苏常州213022

出　　处：《计算机集成制造系统》2011年第7期1509-1517,共9页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(61074056);江苏省自然科学基金资助项目(BK2010201)~~

摘　　要：提出了炉次数未知的炼钢连铸一体化生产的组炉模型。对该模型直接求解存在大量不可行解的困难进行分析,提出将该模型转化为伪旅行商问题的方法,并提出采用离散粒子群优化算法求解该问题。针对离散粒子群优化收敛速度和精度低的缺点,提出了一种基于序列倒置的改进离散粒子群优化算法。引入学习选择概率来选择学习粒子,利用运行代数阈值常数确定当前粒子何时向全局最优粒子学习,并通过局部最优子粒子群比决定局部最优子群的规模。讨论了这些参数的选择原则,并给出了相应参考选择范围。实验研究表明,所提模型是合适的,所提改进算法是有效的。An integrated charge plan model of steelmaking continuous casting with unknown charge number was presented.Based on analysis of difficulties in solving the problem with a large number of infeasible solutions,the model was transformed into Pseudo Traveling Salesman Problem（PTSP） model and discrete Particle Swarm Optimization（PSO） algorithm was applied to solve this model.Aiming at low convergence speed and precision of PSO algorithm,a Inver-over Discrete PSO algorithm（IDPSO） was proposed to solve the PTSP model.Learning selection probability was introduced into the IDPSO to select the learning particles.A threshold of computation generation was introduced into the IDPSO to define when present particles learned from the global particles.Local sub-optimum particle swarm was introduced into the IDPSO to define the scale of the sub-optimum particle swarm.Selection principles of these parameters were discussed in detail and corresponding reference scope was also presented.Experimental results demonstrated that the proposed model was appropriate and the algorithm was effective.

关 键 词：炉次计划 离散粒子群优化 序列倒置算子 旅行商问题 钢铁工业 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]