Decoupling power system state estimation by means of stochastic collocation

This paper presents a new approach to the problem of system decomposition in a power system state estimation problem. The complexity of power systems is growing thus challenging the way measurements for state estimation are traditionally managed. Following a previous experience in defining a decentralized solution for state estimation, the authors here propose a procedure to automatically identify how and what state information to exchange for reconstructing the state starting from partial knowledge. In particular the problem of selecting the variable that each observer has to estimate is partially solved within the framework of stochastic systems. An optimization algorithm based on dynamic programming (DP) is developed to determine the optimal set of strongly coupled variables necessary for a sufficiently accurate estimation. The developed procedure is evaluated in simulation. Preliminary results relevant to a small network are presented to show the validity of the proposed approach.