基于一种自适应选择机制的混合优化算法

摘要

Focusing on the disadvantages of differential evolution(DE)and particle swarm optimization(PSO),such as low convergence precision,slow convergence,this paper presented a hybrid optimization algorithm DEPSO based on DE and PSO. DEPSO involved two new variables,which one represented the change of the fitness of the individuals in the population and the other denoted the change of the fitness of the global best individual.It formed a two-dimensional reasonable choice mechanism through introducing two new variables in order to decide which algorithm would operate in the next iteration.DEPSO involved less parameter and it could achieve it simply,meantime,it could avoid the premature problem through using the second new variable.Massive experiments of several standard benchmark functions in different dimension show that compared with the tra-ditional algorithms,the hybrid algorithm can not only own high convergence precision but also speed up the convergence.More importantly,it also has good results in high dimensions.%针对于微分进化（DE）和粒子群优化（PSO）算法收敛精度较低和收敛速度慢的缺点，提出了基于这两种算法的混合优化算法DEPSO。该算法引入了两个新的变量指标，即在迭代过程中种群个体适应值有所优化的概率及种群的全局最优值的变化情况，通过采用这两个变量所形成的一个二维合理的选择机制，实现下一个迭代过程中关于算法的选择迭代问题。该算法一方面参数较少，实现简单；另一方面，利用新引入的第二个变量指标避免种群陷入早熟。对几种典型的测试函数进行数值模拟实验，结果表明与传统的算法比较，新的算法具有收敛精度高和收敛速度快的特点，同时对于高维的问题依然表现出较好的效果。