Adaptive Constrained Differential Evolution Algorithm by Using Generalized Opposition-Based Learning

基于广义反向学习的自适应约束差分进化算法

¹ Department of Aviation Control and Command, Qingdao Campus, Naval Aviation University, Qingdao 266041, China
² College of Command and Control Engineering, Army Engineering University, Nanjing 210002, China

Received: 8 October 2018

Abstract

Differential evolution is a global optimization algorithm based on greedy competition mechanism, which has the advantages of simple structure, less control parameters, higher reliability and convergence. Combining with the constraint-handling techniques, the constraint optimization problem can be efficiently solved. An adaptive differential evolution algorithm is proposed by using generalized opposition-based learning (GOBL-ACDE), in which the generalized opposition-based learning is used to generate initial population and executes the generation jumping. And the adaptive trade-off model is utilized to handle the constraints as the improved adaptive ranking mutation operator is adopted to generate new population. The experimental results show that the algorithm has better performance in accuracy and convergence speed comparing with CDE, DDE, A-DDE and. And the effect of the generalized opposition-based learning and improved adaptive ranking mutation operator of the GOBL-ACDE have been analyzed and evaluated as well.

摘要

差分进化算法是一种基于"贪婪竞争"机制的全局寻优算法，其控制参量少、结构简单，具有较高的可靠性和收敛性，将约束处理机制引入到差分进化算法可以高效解决约束优化问题。提出一种基于广义反向学习的自适应约束差分进化算法，利用广义反向学习机制生成初始种群并执行种群"代跳"操作，采用自适应权衡模型将约束区分状态处理以及改进自适应变异操作对个体进行排序变异。通过与CDE、DDE、A-DDE、εDE以及DPDE算法进行试验比较以及对广义反向学习和改进自适应排序操作性能分析证明该算法具有较好的寻优精度及收敛速度。