Non-stationary autoregressive model for on-line detection of inter-area oscillations in power systems

This paper addresses early on-line detection of inter-area electro-mechanical oscillations in power systems using dynamic data such as currents, voltages and angle differences measured across transmission lines in real time. The main objective is to give the transmission operator qualitative information regarding stability margins. In our approach, the observed signal is modeled with the non-stationary second order autoregressive model. Bayesian estimation of the system is based on the forgetting approach. The stability margins are obtained as posterior probabilities that the poles of the estimated system are unstable. The approach is demonstrated on real retrospective data recorded in a 500 kV power grid and voltage data obtained by numerical simulations.