Automatic capacitor bank identification in power distribution systems

Tracking the performance and health of capacitor banks in distribution systems is a challenging task due to their high number and the widespread geographical distribution of feeder circuits. In this work we propose a signal processing technique capable of identifying and characterizing the number of capacitor banks connected to a standard North-American feeder circuit. The way the technique is applied allows a real-time remote monitoring of their operation, automatically identifying the switching activity for each capacitor bank connected. The technique is based on an unsupervised clustering of the three phase reactance step magnitudes. We demonstrate that using only passive monitoring of conventional substation bus PTs and feeder CTs, without any communication, nor visual inspection, to individual banks, it is possible to predict the number of capacitor banks on the distribution feeder and track their performance and activity over time.