Adaptive tracking of ambient system oscillations by nonstationary RLS techniques

Measured ambient data in power system are known to exhibit noisy, nonstationarity fluctuations resulting primarily from small magnitude, random changes in load. Accounting for stochastic and time-varying features can provide a better description of the data and result in improved estimation algorithms. In this paper, a new hybrid algorithm combining a recursive least-square (RLS) algorithm and a Kalman filter described by a random walk correlation model is proposed to characterize the time evolution of ambient system oscillations. Extensions and generalizations to current RLS algorithms to deal with nonstationarity are discussed and the relationship between Kalman filter parameters and RLS algorithms is analyzed. Examples of the developed procedures to track the evolving dynamics of critical system modes in both simulated and measured data are presented. Comparisons with well-established approaches such as the exponentially-weighted RLS algorithm, RLS algorithms with adaptive memory, least-mean squares (LMS) algorithms and normalized LMS algorithms demonstrate the accuracy of the proposed procedure.