一种带有递减扰动项的粒子群优化算法

A particle swarm optimization algorithm with successively decreasing disturbance term

针对粒子群优化算法在进化后期存在收敛速度慢、容易陷入局部极值等问题,提出一种带有递减扰动项的改进粒子群优化算法.当进化中后期粒子位置更新过慢或保持相对不变时,通过在粒子速度更新公式中加入递减扰动项,有效地提高微粒进行全局和局部搜索的能力,减小粒子陷入局部最优的可能.基于随机过程理论分析证明了粒子的运动规律是一种马尔科夫过程,且该方法以均方收敛到全局最优解.典型测试函数的仿真结果表明,该算法的收敛性与已有方法相比有较大提高,且算法能够有效避免粒子陷入局部极值.

Aimed at the problems in particle swarm optimization （PSO） algorithm at the end of its evolution such as the slow convergence and the tendency to be trapped into local extremum, an improved algorithm was presented for particle swarm optimization with a successively decreasing disturbance term in the algorithm. In this algorithm, a successively decreasing disturbance term was added into the velocity upda- ting formula when the particle position updating was too slow or kept relatively unchanged in the middle and final evolution periods. Therefore the ability of global and local particle searching was effectively im- proved, and the probability of being trapped into local optimum was reduced. It was verified by theoretical analysis based on stochastic processes that the particle motion was of Markov process and, by using this algorithm, a global optimal solution could be achieved with mean square convergence. Experimental simulation showed that the improved algorithm could not only improve the convergence of the algorithm significantly compared with the algorithms available but also avoid trapping into local optimization solution.