APP下载

一种基于状态空间模型的进化算法

2014-08-08李茂军贾玲

计算技术与自动化 2014年2期

李茂军+贾玲

收稿日期:2013-09-18

作者简介:李茂军(1964—),男,湖南宁乡人,教授,博士,研究方向:智能控制与智能计算。

文章编号:1003-6199(2014)02-0085-04

摘 要:传统进化算法主要通过选择、重组和变异这三种遗传操作实现种群的进化。在进化过程中通常需要设定群体规模、交叉概率和变异概率等参数,而且它们的值会直接影响计算结果及精度。为了简化操作过程,设计一种基于离散系统状态空间模型的进化算法,这种算法采用实数编码方式,构造一个状态进化矩阵来实现重组和变异的功能,提高算法的可操作性和可靠性。并将该算法应用于求解无约束全局优化问题,对几种典型的测试函数进行仿真,结果表明:这种新的进化算法具有搜索能力强、收敛速度快、计算精度高、操作简单等优点,对相关研究有参考作用。

关键词:进化算法;状态空间模型;实数编码;状态进化矩阵

中图分类号:TP301.6文献标识码:A



An Evolutionary Algorithm Based on StatespaceModel



LI Maojun,JIA Ling

(College of Electrical and Information Engineering, Changsha University of Science & Technology, Changsha,Hunan 410114,China)

Abstract:The traditional evolutionary algorithm primarily through three genetic operators: selection, recombination and mutation operations, to achieve the evolution of the population. In the process of evolution, it usually needs to set the crossover probability and mutation probability, which will directly affect the results and precision. In order to simplify the procedure, we design a new evolutionary algorithm, which based on discrete state-space model system and using real-encoding method. The algorithm constructs a state evolution matrix to achieve the function of recombination and mutation, and improve the operability and reliability of the algorithm. We do some simulation based on several typical test functions, the results shows that: this new evolutionary algorithm has many advantages, such as strong search capability, rapid convergence, high precision, simple operation, etc. It has useful reference for relevant studies.

Key words:evolutionary algorithm; state-space model ; real-encoding; state evolution matrix

1 引 言

进化算法(EA)是一类模拟生物进化机制的智能优化方法,如遗传算法(GA)[1]、蚁群算法(ACO)[2]、模拟退火算法(SA)[3]等。同传统的梯度法、牛顿法、穷举法等优化算法相比,进化计算具有自组织、自适应、自学习、不受问题性质限制的优点,因此进化算法常用来解决复杂的工程优化问题[4]。随着科学的发展和应用需求的增加,传统进化算法已不能满足工程应用需要。几十年来,许多学者尝试了很多方法来更好地解决优化问题,如对传统进化算法进行改进、引入新的理论、结合两种或两种以上进化算法等来处理优化问题,取得了一定的效果[5-7]。……

登录APP查看全文