This paper presents a novel formulation of the Distribution System State Estimation (DSSE) optimization model. For a given electric three-phase circuit feeder, network models are built using a quasi-symmetric impedance matrix TRX representing the entire structure and topology of the radial network. As a key contribution, the state variables of demands and generators connected to large-scale distribution grids are obtained by using a convenient matrix reduction technique. As a result, the size of the optimization problem is considerably reduced with respect to the jacobian formulation by considering radial and weakly meshed exploitation and elimination of interconnecting nodes. Results and comparative analysis are presented using the IEEE 4-, 13-, 37-, 123-, and 8500-node test systems.