In the open, dynamic cloud computing environment, more and more functional-equivalent services are provided with different QoS levels by competing service providers, which makes service selection problem becoming more and more important. Considering main drawbacks of traditional approaches for service selection, a systemic method for this multi-objective optimization problem has been presented in this paper. More specifically, at first a novel concept, i.e. partial dominance score, has been proposed. By combining two kinds of ranking method, i.e. the partial dominance score as well as the skyline computation, the optimization objective for service selection has been redefined. Then, a BNL_based service selection algorithm has been presented to find top-k skyline solutions that have higher partial dominance score. Finally the efficiency and effectiveness of our proposed algorithm was evaluated through a set of experimental studies.