| Low frequency oscillations(LFOs) occur frequently, which threatens the safe and stable operation of power systems. On-line parameters identification for LFOs is the premise to suppress LFOs. Existed identification algorithms based on data would become invalid when identifying signals with noise or contain close oscillation modes. In this view, this paper focus on Prony, Esprit, Matrix Pencil three typical parametric identification algorithms, carried out the research on performance evaluation of typical parametric identification algorithms for LFOs, expecting provide basis for improving the identification performance and applicability of identification algorithms. Main works and innovation points of this paper are:Firstly, focus on the problem that Prony algorithm is very sensitive to noise, based on the sensitivity method, this paper derivate the sensitivity formulation of traditional Prony and extended Prony algorithm with respect to a tiny deviation in measured signal, furthermore, on the basis of incremental sensitivity formulation and probability statistics methods, evaluated the anti-noise ability of the two Prony algorithms.Secondly, proposed a series evaluation indexes consists of location measurements indexes, divergence measurements indexes and normality measurements indexes based on the theory of Probability. Evaluated the anti-noise ability of the three typical algorithms based on applying the proposed indexes on the three algorithms when they identified analog signals and 4M2A(4 Machines and 2 areas simulation system) simulation signals.Finally, into account the problem that close oscillation modes in LFOs would impact the identification performance of algorithms, this paper proposed indexes which could evaluate the frequency resolution of identification algorithms. Applied the indexes on the three typical algorithms to evaluate their frequency distinguish ability based on identifying analog signals and 4M2A simulation signals. |
PDF Full Text Request