Adaptive Filtering Techniques for Estimating Electromechanical Modes in Power Systems

Previous adaptive filtering algorithms using the least mean square (LMS) technique for estimating electromechanical modes in power systems imposed constraints arising from the variability and time of convergence of the filter estimates. Also these techniques assumed the power system data to be wide-sense stationary. This work presents a combination of adaptive filtering and block processing algorithms to overcome the constraints of variability and time of convergence of the mode estimates. This work also introduces an adaptive step size least mean squares (ASLMS) algorithm assuming the non-stationary nature of power system data. Finally, this paper investigates the use of an error tracking (ET) algorithm, a combination of LMS and ASLMS algorithms based on the estimation error of the adaptive filters. These techniques are applied to two sets of actual power system data recorded from the western North American power grid. The first data set recorded in June 2000 is 18.85 minutes long, and the second data set recorded in May 2005 is 180 minutes long. The June 2000 data has a lower damping ratio and is more non- stationary than the May 2005 data set in a signal processing sense. The results show that using the LMS in combination with the ASLMS algorithm designed to work in non-stationary environments reduces the variability of the mode estimates and the time of convergence.