基于私有云和改进粒子群算法的约束优化求解  

Constrained optimization problems solving based on private cloud and improved particle swarm optimization

作　　者：张永强[1,2] 徐宗昌[1] 呼凯凯[1] 胡春阳[1] 

机构地区：[1]装甲兵工程学院技术保障工程系,北京100072 [2]海军航空兵学院,辽宁葫芦岛125000

出　　处：《系统工程与电子技术》2016年第5期1086-1092,共7页Systems Engineering and Electronics

摘　　要：为提高约束优化模型的求解准确度和运算速度,针对粒子群算法及其计算方法进行了改进。引入多样化机制避免算法陷入局部最优的危险:创建多个子群将决策空间划分为多个搜索子空间,多子群独立搜索以保证群间解的多样化;用量子粒子代替普通粒子,为其添加服从球状分布的伴随粒子来提高群内解的多样化。多样化的引入增加了计算量和计算复杂度,利用并行计算提高算法运行速度:分析了改进粒子群算法并行计算的方法,在私有云计算平台上编写了基于MapReduce的并行求解流程。实验结果表明,本文方法具有较高准确度,算法的稳定性也较好,运算速度可成倍提高。In order to solve constrained optimization problems with higher accuracy and faster computing speed, several improvements are raised on particle swarm optimization（PSO） and its computing method. Solu- tions＇ diversification mechanism is applied in PSO to improve its global optimization ability., decision space is di- vided into multiple searching subspaces, while multi-subswarms are created according to those searching sub- spaces, and multi-subswarms are searched independently to get solutions＇ diversification among subswarms; or- dinary particles is replaced by quantum particles in PSO, while associated particles that follow globular distribu- tion is vested in each quantum particle, which could improve solutions＇ diversification in subswarms. Running speed of the improved PSO is increased via parallel computing： Parallel computing flow of the improved PSO is analyzed based on the private cloud platform and the algorithm for the flow is programmed based on MapRe- duce. The experimental results show that the proposed method has higher accuracy solutions and stability, and the performance and computing speed is exponentially improved.

关 键 词：约束优化 粒子群算法 私有云计算平台 并行求解 多样化 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] E911[自动化与计算机技术—计算机科学与技术]