量子粒子群优化算法的控制参数分析

Control parameter analysis of quantum-behaved particle swarm optimization algorithm

针对现有的量子粒子群优化算法(QPSO)中收缩-扩张系数α取固定值或线性变化时,不能很好地适应复杂的多维非线性优化搜索问题,提出了两种参数α控制策略:基于Logistic函数的动态非线性递减策略和自适应参数调整策略。在第一种策略中引入S型函数来描述α值在进化过程中的动态变化特性,第二种策略中引入反馈调节方式来控制α值的变化。几个典型函数的实验测试结果表明,两种改进后的参数调整策略对于复杂优化问题在收敛速度和平均最优值上都有所改善,明显优于取固定值或线性变化策略。

In this paper, two parameter-control methods were proposed to remedy deficiencies of existed Quantumbehaved Particle Swarm Optimization（ QPSO） algorithm based on fixed contraction-expansion coefficient α or linear variation not being able to well address problems in complicated nonlinear optimization search. The first was dynamic nonlinear regressive strategy based on logistic function, in which S-type function was introduced to describe the dynamic nature of the value in its evolvement. The second was the adaptive parameter adjustment strategy, in which feedback regulation was introduced to control change in the value. In the case of complicated optimization, experimental results on several typical functions show that the proposed strategy significantly outperforms the existed ones（ those based on fixed value or linear variation） on both average optimal value and convergence rate.