Recursive Frequency Domain Decomposition for Multidimensional Ambient Modal Estimation

Abstract

This paper proposes a multidimensional recursive ambient modal analysis algorithm called recursive frequency domain decomposition (recursive FDD or RFDD). The method enables simultaneous processing of a large number of synchrophasor measurements for real-time ambient modal estimation. The method combines a previously proposed multidimensional block processing algorithm FDD with a single input recursive least square (RLS) algorithm into developing a new frequency domain multidimensional recursive algorithm. First, an auto-regressive model is fitted onto the sampled data of each signal using the time-domain RLS approach. Subsequent modal analysis is carried out in frequency domain in the spirit of FDD. The conventional FDD method uses non-parametric methods for power spectrum density (PSD) estimation. The proposed method in this paper by estimating PSD with a parametric method provides smoother PSD estimation which results in less standard deviation in RFDD estimates compared to FDD. The algorithm is tested on archived synchrophasor data from a real power system.