Aiming at the problem that traditional clustering algorithms consider node location apart form energy distribution, this paper proposes a clustering algorithm based on fuzzy boundary detection, which took into account topological boundary information and node energy distribution, and adjusted the probability that boundary nodes participate in cluster head selection according to the network energy distribution. Simulation results indicate that this algorithm can detect boundary nodes of certain boundary bandwidth, make more sensible cluster head distribution than LEACH, and obviously extend the network lifetime.