Estimation of Model order for Electromechanical Modes using Sequential K-Mean Algorithm

One of the most pressing priorities of smart grid adoption is monitoring and consistent recognition of power quality issues. Small signal oscillations are one of the major concerns in the present-day world due to the increase in the interconnection of multiple power sources. This paper proposes a clustering-based sequential K-Mean algorithm (S-K-Mean) to detect the number of low-frequency oscillatory modes in the given power system signal. It is shown that the suggested algorithm can reliably estimate the number of low-frequency modes present in a signal as it considers all the significant eigenvalues which represents the energy levels of dominant modes. The suggested algorithm works even when the signal is subjected to high noise levels. The robustness of the proposed algorithm over other techniques is validated through synthetic signals at various noise level. The efficacy of the proposed scheme is further shown using a two-area network simulated through MATLAB and real-time digital simulator (RTDS) and it is also tested with a real-time signal obtained from the Western Electricity Coordination Council (WECC).