| In this thesis, online monitoring of the Sub-Synchronous Oscillation (SSO) using ringdown and also ambient data has been studied. IEEE second benchmark model for Sub-Synchronous Resonance (SSR) is used as the case study system. Nonlinear and linearized equations of this system have been extracted to test the performance of the measurement based methods and to compare the results with the real modes of the system. Also, in addition to frequency and damping ratio, torsional mode shapes can be identified by implementing the identification methods on the ambient data. The accuracy of the results shows the possibility of the mechanical parameters online identification using ambient data.;Several ringdown algorithms, namely, Prony, Matrix Pencil, HTLS and ERA methods have been tested for ringdown analysis of SSR related modes. The observability of the sub-synchronous oscillation modes from mechanical and also electrical signals has been examined and it is shown that both electrical and mechanical signals can be used for the SSR monitoring purpose while these modes are more observable in the mechanical signals. Using mechanical signals, such as generator speed or difference between speeds of two turbine shaft masses, shorter analysis window can be used for SSR monitoring which helps earlier SSR detection. It is shown that the analysis can provide early alarm or generator trip signal in about 350 milliseconds.;Ambient data as the system response to the noise input is used for the frequency, damping ratio and also mode shape identification of the torsional modes. Two different type ambient analysis algorithms, namely, Frequency Domain Decomposition (FDD), and Recursive Least Square (RLS) are tested as non-parametric and parametric methods respectively for assessing their suitability for online monitoring of SSR related oscillatory modes. The test results show excellent performance of both methods on tracking the changes in damping levels of the torsional modes and also for providing early alarms.;As part of this thesis and motivated by FDD algorithm, two new algorithms named FFDD and WSD developed to address limitations and improving FDD algorithm. |
