Adaptive control of turbo generator excitation

Concerns the implementation of the GPC self-tuning control on the excitation system of a turbo generator, taking into account several potential problems. The plant signals used for feedback are far from perfect. Although, robustness has received some consideration in self-tuning field, techniques most suitable for the control of excitation problem had to be identified and chosen for implementation. Problems such as unmodelled dynamics, disturbances, steady state offsets and nonlinearities have a considerable effect on the robustness and hence need attention during design. The popular assumption of a persistently exciting plant is generally not true. Hence ways to make the parameter estimator of the self-tuning controller handle those periods during which very little dynamic information about the plant is available, should be carefully considered. Many researchers have attempted to increase the range of applicability of linear self-tuner which include: use of variable forgetting factors for improved adaptivity while also reducing the possibilities of numerical instability; covariance resetting techniques for improved alertness, providing rapid adaptation. Other issues such as cautious least squares estimators, facilitating the input of an element of engineering knowledge, state variable techniques for smoother control via noise rejection and long prediction horizon of the system output have also received attention.