Clustering algorithms have been widely used in many domains so as to partition a set of elements into several subsets, each subset (or ”cluster”) grouping elements which share some similarities. These algorithms are particularly useful in wireless sensor networks (WSNs), where they allow data aggregation and energy cuts. By forming clusters and electing cluster heads responsible for forwarding their packets, the small devices that compose WSNs have not to reach directly the base station (BS) of the network. They spare energy and they can lead further in time their measuring task, so as to detect forest fires or water pollution for example. In this paper, we will apply a new and general clustering algorithm, based on classificability and ultrametric properties, to a WSN. Our goal is to get clusters with a low computational complexity, but with an optimal structure regarding energy consumption.