改进种群多样性的双变异差分进化算法

Differential evolution algorithm with double mutation strategies for improving population diversity

差分进化算法(DE)是一种基于种群的启发式随机搜索技术,对于解决连续性优化问题具有较强的鲁棒性.然而传统差分进化算法存在种群多样性和收敛速度之间的矛盾,一种改进种群多样性的双变异差分进化算法(DADE),通过引入BFS-best机制(基于排序的可行解选取递减策略)改进变异算子"DE/current-to-best",将其与DE/rand/1构成双变异策略来改善DE算法中种群多样性减少的问题.同时,每个个体的控制参数基于排序自适应更新.最后,利用多个CEC2013标准测试函数对改进算法进行测试,实验结果表明,改进后的算法能有效改善种群多样性,较好地提高了算法的全局收敛能力和收敛速度.

Differential Evolution （DE） is an efficient population-based heuristic s- tochastic search technique. It is robust for solving continuous optimization problems. However, the discrepancy of population diversity and convergence rate exists in tradi- tional Differential Evolution. In this paper, differential evolution algorithm based on dou- ble mutation strategies for improving population diversity （DADE） was proposed. This algorithm presents a BFS-best mechanism to improve ＂current-to-best＇, which cooper- ates with DE/rand/1 to ensure population diversity. Meanwhile, the control parameters of individuals are updated automatically based on ranking. Finally, several benchmark functions in CEC2013 are used to test the proposed algorithm. The simulation result- s show that DADE can effectively improve population diversity, achieve better global searching ability and a higher convergence rate.