Abstract
Low frequency oscillation is a power system dynamic stability problem which describes the response feature of power system under sustained minor disturbances. The low frequency oscillation phenomenon was first observed in the Taipower system in year 1984, which affected power supply quality and system operation safety. This type of self-excited low frequency oscillation has then disappeared since Taipower installed power system stabilizers. Presently the low frequency oscillations observed are all those excited by system’s large disturbances. Because the oscillations excited by large disturbances are contaminated with dc offsets, the conventional low frequency oscillations monitored or oscillation parameter evaluation method, referring to the windows-Furier transform which has well suited the self-excited low frequency oscillation, is however not adapted to the oscillation contaminated with dc.It is thus highly necessitated to develop monitoring or evaluation method for the dc contaminated oscillation. This thesis presents two approaches to evaluate the damping constant and oscillation frequency of the low frequency oscillation waves which are contaminated with either constant or decaying dc offsets the first approach uses discrete wavelet transform to filt out dc offset, and then calculate the damping constant by the windows-Furier transform so to extend the application of conventional windows-Furier transform. However the repetitive filters by the wavelet lengthens the computational time and increases the calculator complexity. To overcome these difficulties, the research develops the second approach called the four quadrants method, which makes use of the feature at four quadrantal points of oscillation wave to simplify the evaluation. Without the filtering process for dc offsets, the method is simple and straightforward and thus can calculate the damping constant and nature frequency in one oscillation cycle, and thus save considerable amount of computing time. Due to its simplicity and high speed, the method has great potential for future application to the on-line monitoring of low frequency oscillation parameters.