Long-term optimal reservoir operation with tuning on large-scale multi-objective optimization: Case study of cascade reservoirs in the Upper Yellow River Basin
Hongyi Yao, Zengchuan Dong, Dayong Li, Xiaokuan Ni, Tian Chen, Mufeng Chen, Wenhao Jia, Xin Huang
Abstract
Reservoir system on the Upper Yellow River Basin (UYRB), China. A multipurpose reservoir system with multi-year regulation capacity calls for new optimization with high efficiency owing to the curse of dimensionality. This paper presents a state-of-the-art large-scale multi-objective evolutionary algorithm (LSMOEA), called the weight optimization framework (WOF) with Non-dominated Sorting Genetic Algorithm II (NSGAII) optimizer, to alleviate the problem, and improve its performance by determining applicable grouping mechanism based on inflow features. A novel constrains handle method named dual progressive repair is used to ensure search progress in feasible decision space. Compared to classic NSGA2, WOF with NSGA2 optimizer (WOF-NSGA2 herein) shows better performance on diversity, convergence, and convergence rate. The tuning method, along with the repair method, makes WOF-NSGA2 outperform in all parameter combinations, and produces satisfying operation schedule in the case of multi-objective reservoir operation in the UYRB. The tuning and repair method could be used widely for the large-scale multi-objective reservoir system operation.