Present state estimator performance is not 100% reliable. Specifically, there is significant probability that the state estimators may not converge to an acceptable solution. On an industry wide basis, the probability of non convergence is about 5%. The reasons for this performance have been investigated and have been reported in earlier papers. In this paper we report on a new approach that alleviates the sources of state estimator unreliability and at the same time distributes the computational procedure to each substation of the system, assuming there is at least one GPS-synchronized device (relay, PMU, recorder, meter, etc.) at each substation. This results in a true distributed estimator. The results of the distributed state estimator are communicated to the control center where the overall system state is constructed. The approach has been implemented to two subsystems of two substations each. This paper describes the overall approach and provides results from the two pilot implementations.