刘小娟, 王联国. 一种基于差分进化的正弦余弦算法[J]. 工程科学学报, 2020, 42(12): 1674-1684. DOI: 10.13374/j.issn2095-9389.2020.07.26.002
引用本文: 刘小娟, 王联国. 一种基于差分进化的正弦余弦算法[J]. 工程科学学报, 2020, 42(12): 1674-1684. DOI: 10.13374/j.issn2095-9389.2020.07.26.002
LIU Xiao-juan, WANG Lian-guo. A sine cosine algorithm based on differential evolution[J]. Chinese Journal of Engineering, 2020, 42(12): 1674-1684. DOI: 10.13374/j.issn2095-9389.2020.07.26.002
Citation: LIU Xiao-juan, WANG Lian-guo. A sine cosine algorithm based on differential evolution[J]. Chinese Journal of Engineering, 2020, 42(12): 1674-1684. DOI: 10.13374/j.issn2095-9389.2020.07.26.002

一种基于差分进化的正弦余弦算法

A sine cosine algorithm based on differential evolution

  • 摘要: 正弦余弦算法是一种新型仿自然优化算法,利用正余弦数学模型来求解优化问题。为提高正弦余弦算法的优化精度和收敛速度,提出了一种基于差分进化的正弦余弦算法。该算法通过非线性方式调整参数提高算法的搜索能力、利用差分进化策略平衡算法的全局探索能力及局部开发能力并加快收敛速度、通过侦察蜂策略增加种群多样性以及利用全局最优个体变异策略增强算法的局部开发能力等优化策略来改进算法,最后通过仿真实验和结果分析证明了算法的优异性能。

     

    Abstract: In 2016, a novel naturally simulated optimization algorithm, termed the sine cosine algorithm (SCA), was proposed by Seyedali Mirjalili from Australia. This algorithm uses the sine cosine mathematical model to solve optimization problems and has attracted extensive attention from numerous scholars and researchers at home and abroad over the last few years. However, similar to other swarm intelligence optimization algorithms, SCA has numerous shortcomings in optimizing some complex function problems. To address the defects of basic SCA, such as low optimization precision, easy dropping into the local extremum, and slow convergence rate, a sine cosine algorithm based on differential evolution (SCADE) was proposed. First, the search capabilities of the new algorithm was improved by adjusting parameter r1 in a nonlinear manner and ensuring that each individual adopts the same parameters r1, r2, r3, and r4. Then, differential evolution strategies, including crossover, variation, and selection, were adopted to fully utilize the leading role of the globally optimal individual and information of other individuals in the population. This approach balanced the global exploration and local development abilities and accelerated the convergence rate of the algorithm. Next, using the reconnaissance bees’ strategy, random initialization was performed on individuals whose fitness values showed no improvement in continuous nlim times, which increased the population diversity and improved the global exploration ability of the algorithm. Moreover, the globally optimal individual variation strategy was used to conduct a fine search near the optimal solution, which enhanced the local development ability and optimization accuracy of the algorithm. Based on the above optimization strategies, the algorithm exhibits improvements and its excellent performance is validated by the result analysis of a simulation experiment.

     

/

返回文章
返回