融合多重算子的自适应入侵杂草优化算法

A Self-adaptive Invasive Weed Optimization Algorithm with Multi-operators

针对入侵杂草优化算法易早熟、收敛精度低等不足,在分析原算法优化机理、局限性和收敛性的基础上,提出了融合多重算子的自适应入侵杂草优化算法.首先,通过标准差时变正态分布算子和Levy分布算子来模拟杂草的蔓延扩张行为进行优化,利用Lévy分布可产生较大跳跃的特性来避免局部极值的吸引;其次,利用混沌算子的遍历性和随机性来加强算法的局部搜索能力;进一步根据精英个体的进化状况自适应采用再生策略提高算法的鲁棒性.通过仿真测试和对比,表明所设计算法有效克服了原算法的缺陷,在寻优精度、全局收敛性和稳定性方面优于原杂草优化算法,是解决复杂函数优化问题的一种有效工具.

To overcome the weaknesses of invasive weed optimization algorithm(IWO) in easy precocity and low convergence accuracy,a self-adaptive invasive weed optimization algorithm(MOIWO) with multi-operators was developed based on analyzing optimization mechanism,convergence and limitations of the original IWO.Firstly,MOIWO mimics and optimizes the reproductive behavior of colonizing weeds with normal distribution operator and Lévy distribution operator,while generating large jump characteristics by Lévy distribution to avoid local extreme.Secondly,the ergodicity and randomness of chaotic operators are adopted to enhance local search ability of MOIWO.The robustness of the algorithm is further improved by tracking the evolutionary status of current weeds elites with self-adaptive strategy.The simulation results for 11 benchmark functions show that the proposed algorithm improves the global optimization ability remarkably and outperforms IWO and CPSO in terms of optimization accuracy,global convergence property and stability.