Supervisory framework for event detection and classification using wavelet transform

This paper presents a supervisory framework for enhanced situational awareness in power systems. The discrete wavelet transform (DWT) is used for time-frequency analysis of ambient voltage and frequency signals. A new approach based on wavelet coefficient energy of these measurements is presented for detection of events and localizing them within narrow geographical boundaries. Distinctive features are extracted from voltage and frequency waveforms of five classes of events and a multi-class Support Vector Machine classifier is trained. Next, the trained classifier is used to assort identified events into separate classes. The proposed work is tested on the standard New England-39 bus test system.