基于Oracle罚函数的自适应约束差分进化算法  

ADAPTIVE CONSTRAINED DIFFERENTIAL EVOLUTION ALGORITHMS BASED ON ORACLE PENALTY FUNCTION

作　　者：董明刚[1] 程小辉[1] 牛秦洲[1] 

机构地区：[1]桂林理工大学信息科学与工程学院,广西桂林541004

出　　处：《计算机应用与软件》2014年第1期290-292,322,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(61203109;61262076);广西自然科学基金项目(2012GXNSFBA053172)

摘　　要：为有效求解约束优化问题,减少算法参数,提出基于Oracle罚函数方法的自适应约束差分进化算法。为满足求解优化问题的常用标准,提出一种改进的Oracle罚函数方法。将改进的Oracle罚函数方法与三种自适应差分进化算法相结合,提出三种自适应约束差分进化算法。对11个典型测试函数的优化结果验证了Oracle罚函数方法与自适应差分进化算法结合的有效性。与参考文献中提出的算法的比较结果表明该方法具有良好的寻优性能,因此基于Oracle罚函数方法的自适应约束差分进化算法是一种有效约束优化方法。In order to effectively solve the constrained optimisation problems and reduce algorithm parameters, we propose an Oracle pen- alty method-based adaptive constraint differential evolution algorithm. To meet the common standards for solving the optimisation problems, we present an improved Oracle penalty function. Then, we combine the improved Oracle penalty method with three kinds of adaptive differen- tial evolution algorithms, and present three kinds of adaptive constraint differential evolution algorithms. Results of optimising 11 typical benchmark functions verify the effectiveness of the combination of Oracle penalty function method with adaptive differential evolution algo- rithms. Results of the comparison with the algorithm presented in literature show that the proposed method has excellent optimisation perform- ance. The adaptive constrained differential evolution algorithm based on Oracle penalty function method is an effective constrained optimisation method.

关 键 词：罚函数 自适应 约束优化 差分进化 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]