粒子群算法和灰狼算法的融合

为了解决传统粒子群算法(PSO)容易“早熟”、陷入局部最优以及灰狼算法(GWO)收敛速度慢的问题。首先,采用GWO算法的个体极值更新策略来实现个体包围式向最优值趋近,融入PSO算法的速度更新策略来实现群体向最优值的趋近,并且在原始粒子群算法基础上加入线性惯性权重递减来提高算法的收敛速度,从而提出了一种基于灰狼算法和改进的粒子群算法(IPSO)的融合优化算法(GW-IPSO);其次,通过6个经典算例进行仿真试验,将融合算法与PSO算法、IPSD算法、灰狼和粒子群结合算法(GW-PSO)进行对比;最后,应用融合算法对二级直线倒立摆的控制器设计进行参数寻优。结果表明：针对6个标准测试函数,混合算法的30次试验结果平均值更接近最优值,且标准差几乎都是最小的;应用在倒立摆控制问题上,系统在5 s左右进入稳定状态。融合后的GW-IPSO算法能够在一定程度上避免早熟和陷入局部极值的问题发生,并且能够很好地应用于控制器设计过程中参数寻优问题。

Research on the fusion of PSO and GWO

The purpose of this paper is to solve the problems that traditional particle swarm optimization (PSO) algorithm is prone to DK]"precocious", easy to fall into local optimum, and the grey wolf optimization (GWO) algorithm converges slowly. Firstly, the individual extreme value update strategy of GWO algorithm was used to realize the individual enveloping approach to the optimal value, the speed update strategy of the PSO algorithm was integrated to achieve the approach of the population to the optimal value, and the linear inertia weight reduction was added to improve the convergence speed of the algorithm on the basis of the original particle swarm algorithm, so as to propose an optimization algorithm (GW-IPSO) based on the fusion of gray wolf algorithm and improved particle swarm algorithm. Secondly, through six classical examples, the GW-IPSO algorithm was compared with the PSO algorithm, the improved particle swarm algorithm, the gray wolf and the particle swarm combination algorithm. Finally, the fusion algorithm was applied to optimize the parameters of the controller design of the two-stage linear inverted pendulum. The experimental results show that for the six standard test functions, the average value of the 30 experimental results of the GW-IPSO algorithm is closer to the optimal value, and the standard deviation is almost the smallest. Applied to the inverted pendulum control problem, the system enters a stable state in about 5 seconds. The GW-IPSO algorithm can avoid the problems of precocious maturation and falling into local extremes to a certain extent, and can be well applied to the parameter optimization problem in the controller design process.