Recently, with increasing attention paid to energy production and ecological protection, the hydropower reservoirs operation balancing generation benefit and ecological requirement is playing an important role in water resource and power systems. Thus, the parallel multi-objective genetic algorithm is introduced to effectively resolve this multi-objective constrained optimization problem with two competing objectives and numerous physical constraints. In the proposed method, the original large sized swarm is decomposed into several smaller subpopulations that will be simultaneously evolved on several computing units, effectively enhancing the execution efficiency and population diversity. During the evolutionary process, the chaotic initialization method is used to enhance the quality of initial population, while the feasible space identification method and the modified domination strategy are designed to improve the feasibility of solution and convergence rate of individuals. The results from the Wu hydropower system of China show that the presented method can make full use of computationally expensive resources to improve the performance of population. For instance, compared with the traditional method, the presented method can make 69.23% and 27.44% improvements in the standard deviation of power generation and water deficit in normal year, respectively. Thus, this paper provides an effective tool to support the multi-objective operation optimization of hydropower system. (C) 2018 Elsevier Ltd. All rights reserved.