Multiscale PMU data compression based on wide-area event detection

In this paper, a multiscale compression process for phasor a measurement unit (PMU) is proposed using a wide-area event detection method. For the first step, the data compression intervals are adaptively selected by monitoring the average of modified wavelet energy (AMWE) in order to reflect two different operating conditions of power system; i.e., ambient and event. In the next step, the interval-selected dataset is compressed by a multiscale dimensionality reduction process. The dimensionality reduction step uses wavelet decomposition to reflect non-stationary characteristics and extract time-varying features from the PMU signals. The principal component analysis is then applied to the wavelet-decomposed matrices for data compression. The effectiveness of the proposed method was confirmed by application to the real-world PMU voltage and frequency data, and comparisons are made with the conventional wavelet compression technique.