摘要
目前对混合整数非线性规划问题做的工作甚少 ,这里在研究了混合整数非线性规划和遗传算法各自特点的基础上 ,提出初始群体飘移思想和变尺度整型细密网格技术 ,构造了一种求解混合整数非线性规划的新的遗传算法～初始群体飘移遗传算法。经理论分析和数值试验表明 :该算法对大范围、多峰、非光滑非线性规划问题有较好的全局求解能力 ,在解的精度、稳定性和收敛速度方面均优于一般的算法。
Until now the research for mixed integer nonlinear programming has hardly made a ny progress. In the paper, genetic algorithms and mixed integer nonlinear progra mming are first discussed. Then based upon the definitions of initial group floa ting and techniques of scaling integer dense grid, a new algorithm with initial group floating is presented for mixed integer nonlinear programming. Theoretical analysis and numerical tests show that the genetic method can perform a good gl obal optimum solution for large scaling multi-apex and non smooth mixed integer nonlinear programming (MINLP) with better results than other algorithms used to resolve MINLP in the feasibility, stabilization and convergent speed.
出处
《物探化探计算技术》
CAS
CSCD
2003年第3期253-258,共6页
Computing Techniques For Geophysical and Geochemical Exploration
关键词
遗传算法
非线性整数规划
初始群体飘移
变尺度整型细密网格
genetic algorithms
mixed integer nonlinear program ming
initial group floating
scaling integer dense grid
