A Generalized Subspace Least Mean Square Method for High-resolution Accurate Estimation of Power System Oscillation Modes

A generalized subspace least mean square method is presented for accurate and robust estimation of oscillation modes from exponentially damped power system signals. The method is based on the orthogonality of signal and noise eigenvectors of the signal autocorrelation matrix. Performance of the proposed method is evaluated using Monte Carlo simulation and compared with the Prony method. Test results show that the generalized subspace least mean square method is highly resilient to noise and significantly dominates the Prony method in tracking power system modes under noisy environments.