Clustering is a widely adopted energy-saving technique in wireless sensor networks (WSNs). In this paper, we study algorithms for finding the best locations of cluster heads in WSNs to minimize the overall energy consumption. Specifically, based on the assumption that the global information of all the sensors’ locations or location distribution is available, algorithms are proposed for finding (1) the best location of the cluster head in a single given cluster; (2) the best formation of a given number of clusters where each cluster head has to communicate with base station directly; and (3) the best formation of a given number of clusters where there can be ad-hoc transmission between cluster heads, respectively. For each case, algorithms are designed for free-space and multipath energy consumption models respectively. Theoretical analysis and extensive simulation results show that the proposed algorithms can steadily and quickly achieve satisfactory results. The calculation results of the proposed algorithms provide a useful benchmark for evaluating various local information-based distributed clustering schemes or schemes based on partial or inaccurate global information.