Nonlinear Modal Identification of Power System Response Signals Using Higher Order Statistics

In this paper a novel technique for the detection and estimation of nonlinear mode coupling in power system time series is presented. This technique allows for the simultaneous determination of both, frequency and damping of the nonlinearly coupled modes and can be used to detect and analyze nonlinearities in stressed power systems. Using higher order statistics, techniques are devised for the detection and frequency estimation of coupling frequencies. First, a parametric method for bispectrum estimation based on a non-Gaussian autoregressive (AR) driven model is developed and used to characterize the nonlinear dynamics of oscillatory processes following large perturbations. Analysis methods are then investigated for extracting the relevant amplitude and phase of the coupled modes in both, the time and frequency domains. The method extends current linear identification techniques based on Prony analysis to the case of nonlinear systems with frequency modulation. A simplified 4-machine, 6-bus test power system is used to illustrate the applicability of the technique.