基于双变异算子的差分进化算法

A Differential Evolution Algorithm Based on Double Mutation Operators

针对差分进化算法中全局搜索能力和收敛速度不可兼得的问题,提出了一种具有双变异算子的新型差分算法.将current-to-best/1和rand/1两种变异算子结合,通过控制两个算子的执行比例获得了快速收敛模式和强搜索模式,并进一步通过计算和判断每个个体的进化停滞标识,使算法在运行时自适应的选择两种模式.在CEC2005测试集上的实验证明:与传统算法相比,新算法具有更好的性能和更广的适应性.

The original differential evolution algorithms, which is quickly converged, can＇t have strong global search ability. In this paper, a new double mutation operators based on differential evolution algorithm was proposed to overcome this problem. Firstly, current-to-best/1 and rand/lmutation operators were combined. A fast converge model and a strong search model were born by these combined operators. Secondly, A stop flag, which was added to each individuals in a population, was calculated and judged. Thus, the new algorithm could select the two proposed model automatically. The experiment results which was run by the 25 test functions in CEC2005 benchmark showed that the proposed algorithm had higher precision and high stability than the old algorithms.