基于一维搜索和动态调节的非线性规划PSO算法

对非线性规划问题的处理通常采用罚函数法,使用罚函数法的困难在于参数的选取。本文提出了一种解非线性规划问题的新PSO算法(NSDPSO),该方法融入了一维搜索和动态调节技术,使NSDPSO很好地克服了标准PSO算法在前期收敛较快而在后期易陷入局部最优的缺陷。另外,文中还给出了一种新的适应度函数及选择算子,使算法在选择下一代时保持群体中不可行解的一定比例,这样不但能有效地增加群体的多样性,而且可以避免传统的过度惩罚,使群体向最优解逼近。最后的数据实验表明该算法对非线性规划问题求解是非常有效的。

Nonlinear Programming PSO Algorithm Based on a Line Search and Dynamic Adjustment

Penalty function is often used to deal with the constrained optimization problems,but it is difficult to choose parameter property.In this paper,a new PSO algorithm(NSDPSO) solving the nonlinear programming problems is presented.First,in order to make NSDPSO overcome the defaults in which the simple PSO is fast convergence in the early phase and plunges into the local optimal,a line search and dynamic adjustment techniques are introduced in NSDPSO.Second,a new selection operator based on the new fitness function is also given.Using the new selection operator,NSDPSO can keep a ratio of infeasible solutions in the swarm when selecting the next generation swarm.As a result,it can not only increase the diversity of swarm but also avoid the defects of over-penalization and make the swarm approach the optimal solutions.The numerical experiments show that NSDPSO is effective in dealing with the nonlinear programming problems.