基于改进果蝇算法求解混合整数非线性规划问题  

SOLVING MIXED INTEGER NONLINEAR PROGRAMMING BASED ON THE IMPROVED FRUIT FLIES ALGORITHM

作　　者：朱志同 赵阳[2] 李炜[1,2] 郭星[1,2] 

机构地区：[1]安徽大学计算智能与信号处理重点实验室,安徽合肥230039 [2]安徽大学计算机科学与技术学院

出　　处：《计算机应用与软件》2017年第6期262-266,308,共6页Computer Applications and Software

基　　金：国家科技支撑计划项目(2015BAK24B00)

摘　　要：在科学及工程系统设计中存在许多混合整数非线性规划MINLP(Mixed-Integer Non Linear Programming)问题,该类问题变量类型丰富且约束条件较多,难以求解,为此提出一种改进果蝇算法。该算法对不同类型变量的更新采取不同的策略,并采用周期性的步长函数指导果蝇的寻优,使其避免陷入局部最优。并通过与另外两种常用的算法在稳定性、收敛速度等方面进行了比较,实验结果表明该改进的果蝇算法效果较优,能有效地解决MINLP问题。There are many MINLP problems in the design o f science and engineering systems, w h ich are r ic h in variables and have many constraints and are d i f f ic u l t to solve. There fore, this paper proposes an improved f ru i t f lies algorithm . The algorithm uses diffe re nt strategies to update dif fe re nt types o f va r ia b le s, and uses the periodic step function to guide the optimization o f FOA so as to avoid fa l l in g in to lo cal o p t im iza tion. Compared w ith the other two commonly used algorithms in terms o f s tabi l i ty , convergence speed and so o n , experimental results show that the improved fruit f l ies algorithm can ef fect ively solve the M IN LP problems.

关 键 词：混合整数非线性规划 智能计算 果蝇算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]