Abstract-In this paper, we focus on the problem of the siting of distribution stations, which is of key importance during designing a logistics distribution network. This problem includes dealing with physical obstacles in real world, making the orders as close as possible to their respective stations and making the workload of each station as balanced as possible. To solve this problem, we use a dataset of the geographical coordinates of all orders of a certain express company per day in Shanghai, and propose an algorithm called Balanced COD-CLARANS, which is a constrained clustering algorithm capable of handling physical obstacles and balance factor and outputting a set of clusters for decision-making. Besides, we design the experiment to prove that Balanced COD-CLARANS works well.