Classification of transients in power systems using multifractal analysis

This paper presents a study of a method for feature extraction and classification of voltage disturbances based on multifractal analysis of a voltage waveform with a transient. In this work, the variance fractal dimension trajectory is used to characterize a transient and to extract its features. The features extracted are a trajectory of fractal dimensions that are calculated over a number of overlapping windows along the transient signal. Based on their extracted features, classification of the transients is carried out using a statistical maximum likelihood classifier, which discriminates between three classes of voltage disturbances such as faults, breaker operations and capacitor switchings. The performance of the classifier is considered with both raw signals and also signals contaminated by noise.