Comprehensive performance assessment of state-of-the-art metaheuristic algorithms for multi-scenario reservoir flood control optimization

Flooding is one of the most destructive natural disasters in the world, posing a serious threat to socio-economic and livelihood security. With the intensification of climate change, the frequent occurrence of extreme flood events not only highlights the challenges of reservoir flood control scheduling in terms of accuracy, timeliness, and multi-scenario adaptability but also exacerbates the urgent need for effective flood control solutions. Although traditional optimization methods, such as dynamic programming and linear programming, are widely used in reservoir scheduling, they often face the problems of dimensionality disaster and insufficient processing capacity constraints when dealing with complex constraints and diverse scenarios, which make it difficult to meet the actual needs. In recent years, metaheuristic algorithms have gradually become an important tool for solving such problems due to their excellent global search capability, high robustness, and wide adaptability. However, there are still obvious gaps in current research in terms of the complexity of algorithmic improvement, the singularity of evaluation scenarios, and the diversity of algorithmic performance comparison. Aiming to fill these gaps, this study explores in-depth algorithm selection and evaluation frameworks through systematic innovations. We construct a comprehensive evaluation system containing nine meta-heuristic algorithms, covering both the classical Differential Evolution (DE), Particle Swarm Optimization (PSO), Whale Optimization Algorithm (WOA), and Teaching-Learning-Based Optimization (TLBO), as well as more recently introduced algorithms such as the Art of War Optimizer (AOW), Bald Eagle Search Algorithm (BES), Gold Rush Optimization (GRO), Marine Predators Algorithm (MPA), and Red Kite Optimization Algorithm (ROA). These novel algorithms show strong competitiveness in performance, with the advantages of fast convergence, efficient solutions to large-scale problems, and low time cost, but their application potential in the field of reservoir flood control and scheduling has not been fully explored. In addition, we designed a multi-dimensional evaluation scenario covering short-term single-reservoir flood control scheduling, long-term single-reservoir flood control scheduling, and complex reservoir group joint scheduling to comprehensively examine the adaptive ability of the algorithm in different flood scenarios. We established a comprehensive evaluation system, which not only focuses on the traditional scheduling results and computational efficiency but also introduces in-depth evaluation indexes such as objective function value, convergence ability, and population diversity, and applies three statistical methods, namely, Wilcoxon signed-rank test, Friedman's test, and Nemenyi post hoc test, to ensure that the evaluation results are scientific and reliable. Finally, this study pays special attention to the uncertainty factors in the scheduling process and compares them with previous studies to provide a reasonable basis for algorithm selection in the field of reservoir flood control and scheduling. This systematic research framework not only fills the research gap of the application of new algorithms in the field of flood control and scheduling but also provides an important theoretical and methodological reference for the optimal scheduling of complex water resource systems.