基于罚因子协进化微分算法优化换热网络  

Optimization of Heat Exchanger Networks with Cooperation Differential Evolution Algorithm Based on Penalty Factors

作　　者：方大俊[1] 崔国民[1] 许海珠[1] 万义群[1] 彭富裕[1] 

机构地区：[1]上海理工大学新能源研究所,上海200093

出　　处：《高校化学工程学报》2015年第2期407-412,共6页Journal of Chemical Engineering of Chinese Universities

基　　金：国家自然科学基金资助项目(51176125);沪江基地研究基地专项(D14001);上海市人才发展基金(2009022)

摘　　要：惩罚函数法优化有约束换热网络时,罚因子的取值过大降低计算效率,过小降低优化精度,设置合适的罚因子值却不容易。基于微分进化算法优化连续变量的启发,采用罚因子协同进化机制,同步优化罚因子与解变量,应用于换热网络优化问题。协进化机制优化主要思想如下:建立两类种群,一类为罚因子种群,一类为换热面积种群。优化过程中,每一组换热面积种群用来评价每一个罚因子个体,逐个优化所有解种群,得到所有罚因子个体的评价值,再根据微分进化的思想,对罚因子执行变异、交叉、选择过程,进化罚因子。两类种群如此反复交替优化,直到满足终止条件。算例证明,协进化机制应用于换热网络优化能进一步提高优化效率与精度,也为混合算法选取合适的罚因子优化换热网络提供了一种新思路。, Penalty function methods are usually employed to convert constraint problems to unconstraint problems in heat exchanger networks（HENs） optimization. However, it is difficult to set proper penalty factors, since large penalty factors reduce compute efficiency and small ones decrease optimization accuracy. A cooperative mechanism, i.e. a differential evolution algorithm was developed to optimize penalty factor and HENs area variables simultaneously. Two populations are established： penalty factor population and HENs area variable population. Each penalty factor was evaluated by optimizing the corresponding HENs area variable population. The penalty factor was optimized through mutation, crossover, and selection process based on the differential evolution method. These two steps were repeated until the terminal conditions were satisfied. The results demonstrate that the cooperative differential evolution can further improve the optimization efficiency and accuracy, and also provides a new method to select appropriate penalty factors.

关 键 词：换热网络 约束条件 惩罚函数 惩罚因子 微分进化 

分 类 号：TK124[动力工程及工程热物理—工程热物理]