Classification of sags gathered in distribution substations based on multiway principal component analysis

Abstract

Voltage sags, whether they occur in transmission or distribution systems, may severely damage the loads connected to the power system. As these problems could cost a great deal financially, electric utilities are very interested in finding the origins of sags, that is, whether they have been originated in the transmission network (high voltage (HV)) or in the distribution system (medium voltage (MV)). In addition to the needs of utilities and regulators, many researchers have been prompted to develop reliable methods to properly classify sags. Several of these methods, based on classifying meaningful features extracted from data and waveforms, have been proposed in the literature. Unlike those methods, though, we propose a systematic transformation of data, based on multiway principal component analysis (MPCA), to develop a new voltage sag classification procedure. Sampled voltage and current waveforms of previously registered sags are used together with the MPCA technique to obtain a lower dimensional model. This model is then used to project new sags and classify them according to their origin in the power system. Different classification criteria and parameters are examined to maximize the classification rates of not yet seen sags. Applying the proposed method to real sags recorded in substations demonstrates its applicability and power.