摘要
提出了一种改进的基于拥挤距离的动态粒子群多目标优化算法。为提高粒子的全局搜索能力,提出了新的动态变化惯性权重和加速因子的方法。引进了拥挤距离排序方法维护外部精英集和更新全局最优值。为保持非劣解的多样性,采用了小概率变异机制,并根据种群的大小选择不同的变异概率。最后,把算法应用到5个典型的多目标测试函数并与其他算法进行比较。实验结果表明,该算法所得的Pareto解集有很好的收敛性和多样性。
An improved dynamic particle swarm algorithm for multi-objective optimization based on crowding distance is proposed.To explore the global space more efficiently,the inertia weight and acceleration coefficients are dynamically changed.Meanwhile,the crowding distance sorting is used to maintain the external elitist archive and select the global social leaders.To keep the diversity of the non-dominated solutions,the mutation operator mechanism is adopted,and the probability of mutation is selected according to the size of the population.At last,the algorithm is applied to five classical test functions and compared to other algorithms.It is shown from the results that the Pareto solution obtained from this strategy has a good convergence and diversity.
出处
《计算机工程与设计》
CSCD
北大核心
2011年第4期1422-1425,1452,共5页
Computer Engineering and Design
基金
国家自然科学基金重点项目(60736024)
中央高校基本科研业务费基金项目(2009ZM192)
关键词
多目标优化
拥挤距离
粒子群
惯性权重
外部精英集
非劣解
multi-objective optimization
crowding distance
particle swarm
inertia weight
external elitist archive
non-dominated solutions
