PSO用于优化换热网络时对全局搜索能力的研究  

Study of the Overall Searching Ability of the Particle Swarm Optimization Algorithm When It Is Being Applied in Optimizing a Heat Exchange Network

作　　者：肖媛[1] 崔国民[1] 彭富裕[1] 周静[1] 

机构地区：[1]上海理工大学能源与动力工程学院,上海200093

出　　处：《热能动力工程》2016年第1期20-26,128,共7页Journal of Engineering for Thermal Energy and Power

基　　金：国家自然科学基金资助项目(51176125);沪江基金研究基地专项(D14001)

摘　　要：将PSO(粒子群算法)应用于优化换热网络时,能够快速找到一个全局搜索的最优区域,但同时也会出现局部极值问题。这些问题是由于全局搜索能力的退化和算法的早熟收敛所致。本研究针对该退化现象的机理进行了深入分析,找到了粒子群算法早熟收敛的本质,提出了一种强制跳出的改进策略,通过激活陷入局部极值的粒子,恢复种群多样性并继续搜索全局最优解。算例证明,改进后的粒子群算法的搜索策略适用于换热网络连续变量优化,应用于文献[16]10SP2算例,得到的年综合费用较文献[14]下降了205$/a;应用于文献[18]8SP1算例,得到了目前最小的费用30 793$/a。改进的PSO较标准PSO和文化基因PSO优化后的费用均有所下降。When the particle swarm optimization algorithm is used to optimize a heat exchange network,it can quickly find out the optimum zone in all zones,however,in the meantime,it may produce a local extremum value problem,which arises from its premature convergence and deterioration in its overall searching ability. The mechanism governing such a deterioration phenomenon was analyzed in depth and the essence of the premature convergence happened when using the particle swarm optimization algorithm was found. To solve this problem,a forced jump-out strategy was proposed. Through activating the particles involved in the local extremum value problem,the population diversity was restored and the overall optimum solution was continuously searched. The case calculation results show that the improved searching strategy for the particle swarm optimization algorithm can be suited for a continuous variable optimization of a heat exchange network. When the improved searching strategy for the particle swarm optimization algorithm was applied to the calculation example 10SP2 in the literature No. 16,the annual comprehensive expenses thus obtained decreased by 205 $ / a as compared with that in the literature No. 14. When the improved searching strategy for the particle swarm optimization algorithm was applied to the calculation example 8SP1 in the literature No.18,a currently smallest annual comprehensive expense of 30793 $ / a was obtained. The annual comprehensive expense calculated by using the improved particle swarm optimization algorithm invariably decreased as compared with those calculated by using the standard particle swarm optimization algorithm and the cultural genetic particle swarm optimization algorithm.

关 键 词：换热网络优化 粒子群算法 局部退化 强制跳出 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]