Power disturbance classifier using a rule-based method and wavelet packet-based hidden Markov model

This paper presents a novel classification method for power distribution line disturbances using a rule-based method and a wavelet packet-based hidden Markov model (HMM). The rule-based method is utilized for the classification of time-characterized-feature disturbances, and the wavelet packet-based HMM is utilized for the frequency-characterized-feature power disturbances. This proposed method classifies six types of actual recorded power distribution disturbances, i.e., sag, interruption, fast capacitor switching, capacitor switching, normal variation, and impulse disturbance, and obtains 98.7 % correct classification rate for 670 actual disturbance events tested. Index Terms--Classifier, hidden Markov model (HMM), power disturbances, wavelet packet.