A Data Driven Denoising Framework for Error Corrupted Synchrophasor Data

Recently, Phasor Measurement Units (PMUs) have been utilized in several Wide Area Measurement System (WAMS) based applications and hence the quality of estimates is of utmost concern. In this paper, the non - Gaussian and coloured nature of PMU error is asserted using various statistical (null hypotheses) tests, which forms the basis for a purely data driven denoising framework for PMU data. The proposed framework detects the outliers from measurement data thereby serving as a basis to distinguish between data induced by power system events versus measurement anomalies. A formulation of Empirical Mode Decomposition (EMD) technique with Detrended Fluctuation Analysis (DFA) based thresholding, tailor made for colored noise is used to denoise PMU measurement signals. The technique may be used towards creating a PMU error emulator for testing and research of algorithms focused on mining power system stability and other crucial WAMS based applications.