求解约束优化问题的多成员组合差分进化算法

Diversity composite differential evolution algorithm for constrained optimization problems

针对约束优化问题,提出一种新颖的多成员组合差分进化算法。新算法为了充分利用种群中一个成员的信息,设计了基于组合测试向量产生策略的多成员机制。该机制针对种群中一个成员产生两组候选子成员,每组候选子成员都包括利用三种变异策略分别产生的三个不同的个体。为了利用种群中不可行解的信息,新算法还设计了一种新颖的分散机制。当对最优候选子个体与当前个体进行选择操作时,该机制分别利用概率SR和1-SR选择基于目标函数的比较准则和DEB比较准则,同时概率SR随着进化代数的增加逐渐减小到0。实验结果表明,新算法在三个最难优化的问题g02,g10和g13上具有明显优势。

Aiming at the problem of constrainted optimization,a novel diversity composite differential evolution algorithm was proposed.A multi-member mechanism based on composite trial vector generation strategies was designed by new algorithm to make full use of the information of a member in the population.Two groups of candidate submembers were generated from one member in the population,and each group of candidate submember included three different individual members which generated by three different trial vector generation strategies respectively.In addition,a novel dispersed mechanism was designed to utilize the information in infeasible solutions.When selecting occurred between the current member and optimum candidate submember,the comparison rules based on objective functions and DEB comparison rules were selected by using probability SR and 1-SR respectively.Simultaneously,SR was decreased gradually to 0 with the evolution.The experimental results indicated that the new algorithm exhibited obvious superiority in solving the most difficult optimization problems of g02,g10,and g13.