Classification of Voltage Sags based on Multiway Principal Component Analysis and Case Based Reasoning

Impact of a bad power quality on customers has motivated the development of a classification strategies to identify sags (short duration voltage falls in the power system) and assist the location of its origin. The paper proposes a new method to classify voltage sags registered in distribution substations based on the combination of statistical and reasoning methods. The goal is to associate a sag waveform with its origin (Medium Voltage -MV- or High Voltage -HV-) in the network. Multiway Principal Component Analysis (MPCA) is used as dimension reduction strategy whereas Case Based Reasoning (CBR) is applied in this projection space to retrieve events previously registered and diagnosed. Capability of the proposed method has been demonstrated with data gathered in 25kV power substations. Copyright © 2007 International Federation of Automatic Control All Rights Reserved.