A Genetic Algorithm Parallel Strategy for Optimizing the Operation of Reservoir with Multiple Eco-environmental Objectives

Optimizing the operation of reservoir involving ecological and environmental (eco-environmental) objectives is challenging due to the often competing social-economic objectives. Non-dominated Sorting Genetic Algorithm-II is a popular method for solving multiobjective optimization problems. However, within a complex search space, the NSGA-II population (i.e., a group of candidate solutions) may be trapped in local optima as the population diversity is progressively reduced. This study proposes a computational strategy that operates several parallel populations to maintain the diversity of the candidate solutions. An improved version of the NSGA-II, called c-NSGA-II is implemented by incorporating multiple recombination operators. The parallel strategy is then coupled into the routine of the c-NSGA-II and applied to the operation of the Qingshitan reservoir (Southwest of China) which includes three eco-environmental and two social-economic objectives. Three metrics (convergence, diversity, and hyper volume index) are used for evaluating the optimization performances. The results show that the proposed parallel strategy significantly improves the solution quality in both convergence and diversity. Two characteristic schemes are identified for the operation of the Qingshitan reservoir for trade-off between the eco-environmental and social-economic objectives.