一种更简化而高效的粒子群优化算法

A Simpler and More Effective Particle Swarm Optimization Algorithm

针对基本粒子群优化(basic particle swarm optimization,简称bPSO)算法容易陷入局部极值、进化后期的收敛速度慢和精度低等缺点,采用简化粒子群优化方程和添加极值扰动算子两种策略加以改进,提出了简化粒子群优化(simple particle swarm optimization,简称sPSO)算法、带极值扰动粒子群优化(extremum disturbed particle swarm optimization,简称tPSO)算法和基于二者的带极值扰动的简化粒子群优化(extremum disturbed and simple particle swarm optimization,简称tsPSO)算法.sPSO去掉了PSO进化方程的粒子速度项而使原来的二阶微分方程简化为一阶微分方程,仅由粒子位置控制进化过程,避免了由粒子速度项引起的粒子发散而导致后期收敛变慢和精度低问题.tPSO增加极值扰动算子可以加快粒子跳出局部极值点而继续优化.对几个经典测试函数进行实验的结果表明,sPSO能够极大地提高收敛速度和精度;tPSO能够有效摆脱局部极值点;以上两种策略相结合,tsPSO以更小的种群数和进化世代数获得了非常好的优化效果,从而使得PSO算法更加实用化.

The basic particle swarm optimization （bPSO） has some demerits, such as relapsing into local extremum, slow convergence velocity and low convergence precision in the late evolutionary. Three algorithms, based on the simple evolutionary equations and the extrenum disturbed arithmetic operators, are proposed to overcome the demerits of the bPSO. The simple PSO （sPSO） discards the particle velocity and reduces the bPSO from the second order to the first order difference equation. The evolutionary process is only controlled by the variables of the particles position. The extremum disturbed PSO （tPSO） accelerates the particles to overstep the local extremum. The experiment results of some classic benchmark functions show that the sPSO improves extraordinarily the convergence velocity and precision in the evolutionary optimization, and the tPSO can effectively break away from the local extremum, tsPSO, combined the sPSO and tPSO, can obtain the splendiferous optimization results with smaller population size and evolution generations. The algorithms improve the practicality of the particle swarm optimization.