Adaptive tracking of system oscillatory modes using an extended RLS algorithm

• A nonstationary RLS algorithm is combined with a Kalman filter to deal with measured ambient power system data.
• By tracking the evolving dynamics of system oscillations, the system instability conditions can be determined.
• Extensions and generalizations to current adaptive filtering algorithms to account for nonstationarity are tested.
• The correspondence between the Kalman and RLS variables is examined.
• Early simulation studies conducted on time-synchronized data show that this method can be used for real-time applications.

The study of low-frequency electromechanical modes in power systems has experienced much progress in the past few years. In this research, a nonstationary recursive least-squares (RLS) algorithm with variable forgetting factor is combined with a Kalman filter to simultaneously estimate low-frequency electromechanical modes from measured ambient power system data. Extensions and generalizations to current adaptive filtering algorithms to account for nonstationarity are implemented and tested and the correspondence between the Kalman and RLS variables is examined.Applications of the proposed nonstationary RLS algorithm to track the evolving dynamics of critical power system electromechanical modes in both, simulated and measured data, are presented. Comparison with other RLS and least-mean squares algorithms demonstrate the accuracy of the proposed framework in tracking changes in modal parameters over time. The issues of computational efficiency and memory requirements are discussed in detail.