Abstract
Abstract Low frequency oscillation phenomena first appeared in Taipower system in 1984. The phenomena not only deteriorated power supply quality but also threatened the system security. During 1990’s, Taiwan Power (Taipower) Company had strengthened its transmission system. Thus the spontaneous low frequency oscillation phenomena have not appeared again. However, because of the inherent longitudinal structure of transmission network and the regional power supply and demand imbalance, the low frequency oscillation is still a potential problem in the Taipower system. The large-scale perturbations caused by machine or transmission line trippings in the power system will still lead to the transient oscillations of power angle and power flow between substations. Such oscillations are normally unsteady and thus contain the variational DC offsets. A method integrating the discrete Fourier transform at a variable size of sampling window and the discrete wavelet transform is proposed to analyze the signals of low frequency oscillations, including the calculation for the oscillation parameters such as the main natural frequency and the damping constant, based on them to assess the system dynamic stability. For the spontaneous low frequency oscillations with the constant DC offsets, it is proposed in this thesis to apply the DC offset filtering method by detecting the three waveform peaks, and then apply the discrete Fourier transform at the variable window size to the calculation for the oscillation parameters. As to the transient oscillatory signals having a variable DC offset which are usually excited by the large-scale perturbations, a process is proposed which applies the wavelet decomposition to extract the low frequency oscillation signal components, and then applies the same discrete Fourier transform as for the spontaneous low frequency oscillations. Seven kinds of mother wavelets all integrated with the discrete Fourier transform at a variable window size have been tested on the Taipower system data. Among them, the db10 wavelet and the dmey wavelet functions have been chosen and suggested to use for monitoring the low frequency oscillation parameters of Taipower system.