State estimation and learning of unknown branch current flows using decentralized Kalman filter with virtual disturbance model

This paper presents the design of a decentralized Kalman filter (DKF) without communication to be used for state estimation in distributed generation-based power systems. The idea is to reconstruct information about the system states in the power network, avoiding as much as possible the use of communication channels. The DKF is synthesized based on local models of the power network associated with a virtual disturbance model. The synthesized local Kalman filters of the DKF approach are used for local state estimation while the dynamics of the rest of the power network are lumped into the time-varying virtual disturbance model. The proposed solution is applied to an interconnected power network. By choosing appropriate models for the virtual disturbance the DKF can be suited for both DC and AC distribution systems. It is shown for both cases that the DKF can learn (infer) the local states of the network including the aggregated branch currents coming from the other buses. The herein presented approach is well suited for the agent-based distributed control of micro-grids.