Support vector machines to estimate window length in nonparametric time-varying phasor estimation

A new approach to nonparametric modelling techniques for tracking time-varying voltage phasors in power systems is proposed. A first order polynomial is used to approximate these signals locally on a sliding window of variable length. To estimate the length of this variable window the Intersection of Confidence Intervals (ICI) method could be used. This method requires a number of calculations over a range of different window lengths. In this paper we propose to use a machine learning technique, namely a support vector machine (SVM) to estimate the appropriate window length. A SVM is trained to model the ICI method and once the SVM is trained, it is much faster than the ICI method.