计算机科学
水准点(测量)
资源配置
数学优化
差异进化
排名(信息检索)
人口
资源管理(计算)
最优化问题
计算复杂性理论
突变
分布式计算
算法
人工智能
数学
计算机网络
生物化学
化学
人口学
大地测量学
社会学
基因
地理
作者
Yaxin Li,Jiye Liang,Caitong Yue,Kunjie Yu,Xuanxuan Ban,Peng Chen
标识
DOI:10.1016/j.ins.2024.120104
摘要
Differential evolution (DE) is an efficient and powerful population-based search algorithm for solving numerical optimization problems in continuous spaces. It has been proven that multi-strategy DE algorithms are more effective than single-strategy DE algorithms in addressing benchmark and real-world problems. However, most multi-strategy DE variants focus on maintaining population diversity and balancing exploitation and exploration, ignoring the dynamic allocation of computational resources. Moreover, the success of these algorithms often depends on additional designed techniques, leading to increased computational complexity. In this paper, the Collaborative Resource Allocation-based Differential Evolution (CRADE) is introduced. It involves a collaborative resource allocation mechanism that utilizes the historical performance ranking of three mutation strategies to automatically allocate computational resources to various subpopulations during the search process. The parameter adaptation technique is used to adjust the associated control parameters of different mutation strategies. As a result, the most efficient mutation strategy consumes the majority of computational resources at different search stages to mitigate inefficient search under constrained resources. The performance of CRADE is evaluated on the well-known CEC2013 benchmark function set. The paper also investigates its application in the parameter identification of photovoltaic solar cells and modules. The overall results show that CRADE exhibits superior and competitive performance compared to other state-of-the-art algorithms. Consequently, CRADE has emerged as a novel and effective approach for addressing numerical optimization problems, distinguished by its excellence, practicality, and unwavering reliability.
科研通智能强力驱动
Strongly Powered by AbleSci AI