Abstract: To address the weak local search ability and low optimization accuracy of the original dynamic Self-Adaptive Differential Evolution(SADE) algorithm,this paper proposes a Neighborhood Search Differential Evolution(NSDE) algorithm for solving Dynamic Optimization Problem(DOP).By introducing the neighborhood search mechanism,the neighborhood space of the best individual of the population is divided properly,and a set of candidate solutions are generated within the divided space.The optimal solution in this set is selected to iterate the best individual of the population,which enhances the local search ability of the algorithm.At the same time,the hill-valley function is introduced into the traditional distance-based exclusion scheme to track the adjacent peaks to improve the optimization accuracy of the algorithm.Experimental results show that in terms of the average error, NSDE algorithm outperforms the original dynamic SADE on 28 test problems(49 test problems in total), Dynamic Optimization of Artificial Immune Network(dopt-aiNet) on 38 problems, Differential Evolution with Competitive Strategy Based on Multi-Population(DECS) on 29 problems, and Modified Differential Evolution(MDE) on 38 problems, which demonstrates the NSDE algorithm has better overall performance.

Key words: Self-Adaptive Differential Evolution(SADE), Dynamic Optimization Problem(DOP), neighborhood search, exclusion scheme, average error

摘要： 针对原始动态自适应差分进化（SADE）算法局部搜索能力弱和寻优精度低的问题，提出一种求解动态优化问题的邻域搜索差分进化（NSDE）算法。通过引入邻域搜索机制，在划分种群最优个体的邻域空间范围内产生候选解，选取候选解集合中的最优解并对种群最优个体进行迭代，增强算法局部搜索能力。在传统基于距离的排斥方案中，引入hill-valley函数追踪邻近峰，提高算法寻优精度。实验结果表明，与SADE、人工免疫网络动态优化、多种群竞争差分进化和改进差分进化算法相比，NSDE算法在49个测试问题中分别有28、38、29和38个测试问题的平均误差更小，综合性能表现更好。

关键词: 自适应差分进化, 动态优化问题, 邻域搜索, 排斥方案, 平均误差

CLC Number: 

• TP301.6