两段式差分进化算法及函数优化

Two-stage differential evolution algorithm and function optimization

针对标准差分进化算法解决不同问题时需要对控制参数进行不同的设置,提出了两段式差分进化算法.该算法利用正态分布随机数生成变异率的算子,并把进化过程分为2个阶段,不同阶段分别采用不同的交叉因子,根据不同的配置利用生成变异率来改善算法性能.同时为了加快局部寻优,利用拥有优势解的随机向量指引寻优方向.对一系列Benchmark和Dixon-Szego¨函数进行测试,并与DE以及其他自适应DE算法加以比较,结果显示本算法的收敛速度与优化质量均有显著提高.

As control parameters needs to be set appropriately in differential evolution（DE） algorithm when solving a specific problem,thus,a novel two-stage differential evolution algorithm was proposed.Mutation rates were respectively self-adapted by an operator based on normal distribution random number,and the evolution process was divided into two stages with different crossover probability.At the same time,to enhance the convergence speed,the randomly selected vectors with optimal fitness value were introduced to guide searching direction.Benchmark and Dixon-Szeg problems were used to verify this algorithm.Compared to other classic or adaptive DE algorithms,the simulation results indicate that this algorithm performs better than several other algorithms in terms of solution accuracy and convergence speed.