面向应用的快速多峰寻优算法

Application-oriented fast optimizer for multi-peak searching

工程应用中的多峰寻优问题要求搜索目标函数的多个极值点,现有的多峰优化方法难以直接利用应用问题的先验知识引导算法过程,多峰寻优效率较低。基于粒子群优化算法设计一种面向应用的多峰寻优算法,能有效利用易于获得的先验参数,如峰间分辨率、峰位置精度、峰值个数等实现快速多峰搜索。该算法保持了粒子群算法的简单性并改善了搜索多样性,使其可控地收敛到多个峰值上。将该算法与几种典型的多峰寻优方法进行了对比测试和分析,结果表明,对复杂多峰函数,该算法能以最快的收敛速度实现多峰搜索。

Multi-peak optimizations in engineering applications are needed to search muhiple extrema of objective functions. Previous multi-peak searching methods usually cannot make use of prior parameters to guide the algorithm directly. This paper proposed an application-oriented multi-peak optimizer based on the particle swarm optimization （PSO）. It took advantage of prior parameters such as peak resolution, solution accuracy, and peak number demanded, which could usually be ascertained in real-world problems. The algorithm kept the simplicity of basic PSO and expanded its searching diversity. This paper compared the new algorithm with some typical muhimodal optimization algorithms on the basis of which the tests and analyses of them were conducted. Results show that the new algorithm can successfully locate the multiple extrema that outhors need at highest speed.