Disturbance classification based on the geometrical properties of signal phase-space representation

This paper deals with the use of the geometrical properties of phase-space representation in the problem of disturbance detection and classification. Time-domain periodical signals can be represented using phase-space diagrams which plot trajectories defined by (x(t),x/spl dot/(t)). The trajectories traced show different geometrical characteristics which can be used to distinguish between different types of disturbances. The paper outlines a set of procedures developed for extracting and classifying geometrical parameters. This kind of technique can be used when disturbances have to be classified according to previously registered patterns. The method is developed for voltage or current waveforms of arbitrary sampling rates and numbers of cycles. It has been tested using a power quality instrumentation designed by the authors' research group. The ultimate goal is to obtain a general disturbance dictionary and a classification method which can be used to classify disturbances automatically.