TRANSIENT CLUSTERING APROACH FOR PQ MONITORING

Abstract

Power quality monitoring systems play an increasingly important role in the future operation of distribution grids. A major challenge for operating a monitoring system is the processing of measurement data in an efficient and meaningful manner. The generated data sizes are considerable and require experts in the field for a detailed site evaluation. This is both time and cost intensive. One particular problem is the analysis of power quality events, such as transients, which usually occur in high numbers. The identification of the transient’s sources require a detailed analysis of the distorted waveforms and corresponding rms-measurements. The approach in this paper categorizes transients based on the similarity of features in order to link the events to a source. The transient waveforms and rms-measurements are both considered in the segmentation, feature extraction and data reduction process. A data set consisting of 3500 transients, measured at an electric vehicle-charging infrastructure, was utilized to validate the method. Results show that the method was able to categorize the samples into different types of transients, allowing the identification of the sources of the majority of disturbances.