基于种群多样性的改进差分进化算法应用于换热网络优化

An Improved Differential Evolution Algorithm based on Population Diversity and its Application in Heat Exchanger Network Synthesis

针对差分进化算法运用于换热网络优化时后期搜索效率低,容易出现早熟收敛而陷入局部最优的不足,以变量方差为种群多样性定量评价指标。分析了差分进化算法的控制参数对种群多样性的影响及种群多样性与算法寻优能力之间的关系,在此基础上建立缩放因子自适应调整和种群动态更新策略的改进差分进化算法并应用于两个换热网络实例,优化参数以年综合费用为目标,结果分别为5 606 985和2 928 032$/a,较改进前分别减少19 089和18 042$/a。改进后的差分进化算法能够在进化过程中提升种群多样性,增强算法搜索能力,抑制算法早熟收敛。

An evaluation criterion of population diversity based on individual variance is developed to improve the searching efficiency and quality of differential evolution algorithm used for heat exchanger network synthesis. Based on the effects of control parameters on population diversity and the relations of the population diversity with the searching ability of differential evolution algorithm,an improved algorithm for population diversity and population dynamic update strategy is established and applied to two examples. The optimization results are 5 606 985 $/year and2 928 032 $/year,respectively,that is,a decrease of 19 089 $/year and 18 042 $/year compared to before optimization. The results show the algorithm has improved the population diversity and searching ability,and suppressed algorithm premature convergence.