Feedback linearizing generator excitation control designs have demonstrated improved performance over conventional controls, such as power system stabilizers, in simulations. This type of control aims to cancel the nonlinearities in the dynamics of the generator, resulting in a closed-loop system that is linear. However, feedback linearizing control. or FBLC, depends on a measurement of the rotor acceleration, which is subject to considerable noise from shaft vibrations. This thesis examines the impact that these vibrations have on the operation of FBLC. Several possibilities for reducing the effects of torsional shaft dvnamics on control performance are also explored.The torsional dynamics are represented by a linear model. The addition of these dynamics does not affect the linearity of the closed-loop FBLC system, although the closed-loop eigenvalue placement is distorted. Furthermore, the damping of the shaft modes is much larger in the presence of FBLC. In fact, FBLC is capable of damping out shaft oscillations that are otherwise unstable due to subsynchronous resonance. However, the torsional dynamics greatly increase the tendency of the field voltage to saturate at its upper and lower limits, degrading the performance of FBLC.Several options for improving FBLC performance are considered. The acceleration measurement may be low-pass filtered; however, the phase shift from the filter in the torsional range is capable of exciting the shaft modes, leading to instability. Redesigning FBLC to include torsional dynamics produces even larger oscillations in the field voltage and poor performance in practical situations. An alternative control strategy is sliding mode control, which allows for a range of modeling errors. Because torsional oscillations produce large, high frequency uncertainties, sliding mode control does not provide any improvement over FBLC. Without modifications, FBLC is observed to remain stable over large variations in shaft parameters.