多模函数优化的改进花朵授粉算法

Improved flower pollination algorithm for multimodal function optimization

为了探讨花朵授粉算法(FPA)在解算多模函数优化问题中存在的不足,通过定义种群多样性及差异性指标,定性分析了FPA在多模复杂函数优化中的寻优缺点。基于模拟退火思想优化全局授粉过程,并利用Nelder-Mead单纯形搜索技术对花朵局部授粉进行重构,提出一种新的花朵授粉寻优架构。仿真结果表明,相对于基本的FPA、布谷鸟算法、萤火虫算法,改进花朵授粉算法能够有效避免陷入局部最优,具备优异的全局勘探和局部开采能力,对多模优化问题具有一定优势。

In order to discuss the defects of flower pollination algorithm（FPA）in solving multimodal optimization problems,the optimal disadvantages of flower pollination algorithm in multimodal function optimization were qualitatively analyzed by defining population diversity and difference index.And then a new framework of FPA was constructed by optimizing the global pollination process based on the simulated annealing idea and using Nelder-Mead simplex search method to reconstruct the local pollination process.The simulation results show that the improved flower pollination algorithm can effectively avoid falling into local optimum and has better global exploration and local exploitation abilities,which has advantages to solve multimodal function optimization,compared with primary flower pollination algorithm,cuckoo search algorithm and firefly algorithm.