A Nonparametric Technique for High Precision Estimation of Power System Time-Varying Phasors

Abstract

In this paper we present a non-parametric technique for the estimation of time-varying phasors and frequency in power systems. In this approach, instantaneous RMS value and phase angle of either voltage or current phasor are obtained using Park transform followed by demodulation. A linear function is used to approximate these signals locally on the sliding window of variable length. Optimal window length is selected for each signal sample: large window if the signal is slow time-varying to improve filtering of noise and harmonics, and small window for sudden changes to make signal tracking more accurate. In order to make high precision estimates an approach that combines left-sided, central and right-sided windows is proposed in the paper. The paper concludes with the presentation of examples based on simulated signals as well as real-life signals recorded during power swing condition in ESKOM transmission system.