Nonparametric frequency estimation for power system applications

In this paper, we present application of the nonparametric signal modeling technique for tracking frequency in power systems. This technique relies on polynomial signal model with variable degree, and left-sided sliding window with variable size. The polynomial is fitted locally using the samples from the sliding window. In general, for each new sample the window size and the polynomial order are automatically selected to smooth the frequency with the highest possible accuracy. Specifically, for frequency tracking in power systems, the first order polynomial model and few carefully selected window sizes is enough to design the algorithm that estimates frequency very accurately in all possible situations. The automatic selection of the window size for each new sample is performed by the Intersection of Confidence Intervals algorithm. The paper concludes with the presentation of the representative testing results.