A modified noise assisted EMD to extract low-frequency modes present in a WAMS data of dynamic power system

Empirical Mode Decomposition(EMD)is incapable of separating mode frequencies present in a octave because it suffers from mode mixing problem. This paper proposes modification to noise-assisted Empirical Mode Decomposition technique which can effectively improve the mode mixing problem. The exsiting and proposed improvements to these methodologies in this paper are first applied to an artificial test signal to verify the ability in separating mixing modes. Thereafter, the real-time data of Eastern Interconnect Phasor Project (EIPP), U.S.A are analyzed. Further different modal frequency components are extracted by EMD, Ensemble Empirical Mode Decomposition(EEMD), Complete Ensemble Empirical Decomposition with Adaptive Noise (CEEMDAN), and modified CEEMDAN. Hilbert spectrum analysis is carried out to compare instantineous frequency variation of various extracted modes. From the simulation results, it is concluded that EEMD technique works well in fixing mode mixing problem than previously used EMD based techniques but the problem of noise in the extracted modes of EEMD still remains which is overcome by CEEMDAN technique. CEEMDAN suffers from the problem of presences of noise in the extracted modes and existence of spurious mode, which are then overcome by modified CEEMDAN.