Application and effect evaluation of different optimization algorithms in distributed hydrological model

作     者：Wu, Ming Gao, Yuqin Liu, Yunping Xu, Longsheng Gao, Li 

作者机构：Hohai Univ Coll Water Conservancy & Hydropower Engn Nanjing 210024 Peoples R China 

出 版 物：《JOURNAL OF HYDROLOGY-REGIONAL STUDIES》 (J. Hydrol. Reg. Stud.)

年 卷 期：2025年第58卷

核心收录：

基　　金：National Key Research and Development Program of China [2021YFC3000104] National Natural Science Foundation of China 

主　　题：Optimization algorithm Chaotic particle swarm genetic algorithm Distributed hydrological model Parameter calibration Performance evaluation 

摘      要：Study Region: The upper reaches of the Shaying River Basin (the USR Basin) in the Huai River Basin, China Study Focus: The calibration of model parameters is one of the key challenges in advancing the application of hydrological models. The study proposes a novel metaheuristic optimization algorithm, the chaotic particle swarm genetic algorithm (CPSGA), and compares its performance with four well-known optimization algorithms in the field of hydrological model calibration: GA, PSO, DE, and SCE-UA. The comparison focuses on effectiveness, stability, time consumption, and convergence characteristics. New Hydrological Insights for the Region: During the parameter calibration process of runoff simulation in the USR Basin, CPSGA demonstrates strong effectiveness and convergence characteristics. It enhances the GA framework by integrating initial population chaotization, perturbation evolution, and sub-adaptation strategies, which improve individual diversity and facilitate targeted evolution, thereby increasing effectiveness and convergence. However, these enhancements compromise stability and increase time consumption compared to other algorithms. While PSO shows the best convergence characteristics, it suffers from reduced swarm diversity in later iterations, leading to local optima and poor effectiveness. The complex concept and competitive complex evolution (CCE) strategy of SCE-UA make it less effective for optimization problems with a considerable number of variables, limiting its suitability for calibrating fully distributed hydrological models. These results can provide reference for parameter calibration and uncertainty analysis in distributed hydrological models.