摘要:为解决粒子群优化算法易陷入局部最优值的问题，提出一种引入多级扰动的混合型粒子群优化算法.该算法结合两种经典改进粒子群优化算法的优点，即带惯性参数的标准粒子群优化算法和带收缩因子的粒子群优化算法，在此基础上，引入多级扰动机制：在更新粒子位置时，引入一级扰动，使粒子对解空间的遍历能力得到加强；若优化过程陷入“局部最优”的情况，则引入二级扰动，使得优化过程继续，从而摆脱局部最优值.使用了6个测试函数——Sphere函数、Ackley函数、Rastrigin函数、Styblinski-Tang函数、Duadric函数及Rosenbrock函数来对所提出的混合型粒子群优化算法进行仿真运算和对比验证.模拟运算的结果表明：所提出的混合型粒子群优化算法在对测试函数进行仿真时，其收敛精度和收敛速度都优于另外两种经典的改进粒子群优化算法；另外，在处理多峰函数时，本算法不易被局部最优值所限制.



Abstract:To avoid the locally optimum which is frequently be the result of a calculation of particle swarm optimization (PSO) algorithm, it is proposed in this study a new mixed PSO algorithm with multistage disturbance (MPSO). MPSO combined features from two former classic improved PSO algorithms, which are standard particle swarm optimization (SPSO) and standard particle swarm optimization with a constriction factor (PSOCF). Furthermore, a strategy with multistage disturbances was also introduced into the algorithm:The first-level disturbance was used to enhance the ability of the particles to traverse the solution space when renewing the positions, while the second-level disturbance would be introduced when locally optimal solution was received to continue the optimization process. Six test functions, namely the Sphere, Ackley, Rastrigin, Styblinski-Tang, Duadric, and Rosenbrock functions, were used to simulate the optimization calculation, and the results from proposed algorithm MPSO were compared with those from SPSO and PSOCF. The results show that for the test functions, MPSO can get the optimal value much more quickly and easily than the other two algorithms, and the convergence precision of MPSO was significantly higher than the others. It can be concluded that MPSO can get over the problem of locally optimal solution when dealing with multimodal functions.



PDF全文下载地址:

http://jos.org.cn/jos/article/pdf/5600