This paper presents a hierarchical structure for decentralized power system state estimation whose lower level is conducted at each power network substation. The local level modeling is based on a nonlinear Coestimation algorithm which provides not only the nodal state variables, but also the substation topology. The higher hierarchical level, performed at the Regional Control Center, coordinates the local estimates by processing the active power flow measurements taken on the transmission lines connecting the substations. In addition to other attractive features, the proposed estimation architecture also allows a substantial reduction on the amount of data transmitted from substations to control centers. The proposed methodology is tested through a number of case studies performed on the IEEE 9-bus network modeled at bus section level.Index Terms-Hierarchical state estimation, power system state estimation, real-time power system modeling, state and topology coestimation.