惯性权值线性递减(LDI)的粒子群算法不能很好地反映粒子搜索过程的复杂非线性行为,收敛速度和收敛精度仍不够理想.对此,提出一种动态非线性改变惯性权(DNI)的自适应粒子群算法.在该算法中通过引入非线性指数函数来描述惯性权值在进化过程中的动态变化特性,并通过数值实验确定了非线性函数关键控制参数的合适取值范围.通过典型测试函数验证算法的性能,并与文献报道的已有结果比较.实验表明:对单峰值函数优化问题,DNI自适应粒子群算法收敛速度明显优于LDI算法;对多峰值函数优化问题,DNI算法跳出局部最优的能力及收敛精度也好于LDI算法.%The particle swarm optimization (PSO) is a relatively new generation of intelligent optimization algorithm which is based on a metaphor of social interaction, namely bird flocking or fish schooling. Although preexist PSO algorithms with linearly decreasing the inertia weight (LDI) have shown some important advances in convergence velocity and quality, its linear variation couldn' t effectively simulate complicated nonlinear search behavior of the particle during iterations. There is still some problem in jumping from local optimal solution for these algorithms. A new improvement of PSO is proposed with dynamic nonlinear inertia weight (NDI) variation. A nonlinear exponent function is introduced to describe the dynamic variation character of inertia weight with swarm evolution. A suitability set of control parameters of the nonlinear function through numeric experiments is also presents. The performance of the proposed PSO algorithm is demonstrated by applying it for several benchmark problems and comparing it with results reported in literatures. It is clear that DNI particle swarm algorithm shows faster convergence velocity for uni-modal functions and better ability of jumping from getting into near optimal solution to better solution quality.
展开▼
