In this paper, we consider the scenario of using un-manned aerial vehicles base stations (UAV-BSs) to serve cellular users. In particular, we focus on finding the minimum number of UAV-BSs as well as their deployment. We propose an optimiza-tion model which minimizes the number of UAV-BSs and optim-ize their positions such that the user equipment (UE) covered ra-tio is no less than the expectation of network suppliers, the UEs receive acceptable downlink rates, and the UAV-BSs can work in a sustainable manner. We show the NP-hardness of this problem and then propose a method to address it. The method first estim-ates the range of the number of UAV-BSs and then converts the original problem to one which maximizes the UE served ratio, given the number of UAV-BSs within that range. We present a maximizing algorithm to solve it with the proof of convergence. Extensive simulations based on a realistic dataset have been con-ducted to demonstrate the effectiveness of the proposed method.