With the increasing power demand and rapid depletion of conventional fossil fuel resources, hydroelectric power resource has caused great attention of the public. A multi reservoir system with multiple objectives including ecological water demand, hydropower generation, and water diversion in Lushui River basin in China is under consideration in this context. Aiming to improve the efficiency of the water resources utilization, a novel method called multi-objective Moth-flame optimization algorithm (MOMFA) has been applied into this problem. The proposed algorithm involves the effective properties of the original Moth-flame optimization algorithm and two efficient mechanisms named opposition-based learning and indicator-based selection have also been integrated into the algorithm with the purpose of assisting the algorithm to accelerate the convergence and maintain the diversity simultaneously. The performance of the proposed MOMFA tested on a series of benchmarks and the Lushui River Basin. The result indicated that the proposed algorithm is not only capable of obtaining the well pareto solutions on standard problem but also can find the best tradeoff of the components and simultaneously achieve a set of well distributed non-dominated solutions for the multi-objective water resources utilization problem. Compare with the results obtained by other algorithms, the superiority of the proposed MOMFA has also been verified.