A segmented differential evolution with enhanced diversity and semi-adaptive parameter control

Differential evolution (DE) is widely recognized as one of the most potent optimization algorithms, capable of effectively addressing a broad spectrum of optimization challenges. Nevertheless, even the most advanced variants of DE share some common challenges. This paper introduces a novel multi-stage semi-adaptive DE algorithm with enhanced diversity (MSA-DE), offering several key contributions: first, the algorithm is structured into three distinct stages, each employing a unique new mutation strategy and designed a new evolutionary scheme based on this segmentation, to better balance exploration and development at all stages of the process. Secondly, building on the idea of parameter restriction in the evolutionary stage of delineation, a semi-adaptive parameter control method based on the fitness of the irrelevant function is proposed which effectively solves the instability problem of excessive fluctuations in the convergence of adaptive parameters. Thirdly, new diversity maintenance mechanisms are proposed, including population initialization, shrinkage, and updating, which better ameliorated the conflicting issues of search range and search rate that existed at all stages of the DE variant. Finally, comprehensive experiments were conducted on the CEC2013, CEC2014, and CEC2017 benchmark test suites to rigorously assess the accuracy, convergence rate, and overall effectiveness of each module. The results show that MSA-DE exhibits strong competitiveness in single-objective optimisation problems. In addition, the experimental results demonstrate the superiority of the algorithm for real-world engineering problems.