A Systematic Review of Hyper-Heuristics in Power System Optimization: Theory, Applications, and Future Directions

Abstract

Modern power system optimization problems have become increasingly large-scale, non-convex, and highly constrained. Conventional optimization methods and metaheuristics struggle with these complexities, primarily due to premature convergence, extensive parameter-tuning requirements, and the limitations of the No Free Lunch theorem. To address these challenges, hyper-heuristics have emerged as high-level approaches that search the space of heuristics rather than the solution space directly, providing greater generality across different domains. Following the PRISMA 2020 guidelines, this systematic review searched Scopus, Web of Science, ScienceDirect, IEEE Xplore, Springer Nature Link, and Taylor & Francis Online for peer-reviewed studies published up to June 2026 on hyper-heuristics applied to power system optimization. Study eligibility was assessed based on topical relevance, and the findings were synthesized narratively. Out of 162 unique records screened, 21 studies were included in the final review without formal risk-of-bias restrictions. The findings indicate that while hyper-heuristics offer excellent efficiency, robustness, and solution quality, their overall application in power systems remains limited. Current research relies heavily on selection-based architectures, leaving significant gaps in computational time reporting and real-world grid testing. Future research should extend selection-based methods, explore wider problem variants, address underexplored power system optimization domains, and investigate the development of generation hyper-heuristics.