一种求解非线性规划问题的粒子群算法

Particle Swarm Algorithm of the Nonlinear Programming Problem

为了求解非线性混合整数规划问题,提出了一种基于随机游走的粒子群优化算法(RWPSO).构造出自适应的惯性权重,平衡了算法的全局和局部搜索能力;提出了一种"随机游走"行为,增强粒子的局部寻优能力;为了防止算法出现早熟收敛现象,提出了"优胜劣汰"更新机制.最后,为了验证算法在求解非线性混合整数规划问题方面的可行性和有效性,将提出的算法用16个常用的测试函数进行了测试并与其他3种算法比较.实验结果表明,RWPSO算法在精确度和成功率方面得到了很大的提高.

A particle swarm optimization based on random walk（RWPSO）is proposed to solve nonlinear mixed integer programming problem.To balance local search capability and global search capability,selfadaptive inertia weight is constructed.A ＂random walk＂behavior is proposed to enhance the local search ability of particles.In order to prevent premature convergence,the ＂survival of the fittest＂update mechanism is introduced.Finally,in order to validate the algorithm is feasible and effective for solving nonlinear mixed integer programming problem,RWPSO is tested and compared with the other three algorithms in 16 test functions.The experimental study shows that RWPSO has been greatly improved in terms of accuracy and success rate.