Control of a high-order power system by neural adaptive feedback linearization

Abstract

Discusses the design of an adaptive feedback linearizing excitation controller/power system stabilizer using neural networks, for a high order single machine/infinite bus power system model. The neural networks are used to identify a discrete-time nonlinear dynamical model of the system. Although the controller itself is not implemented using a neural network, it adaptively uses the parameter estimates given by the neural system model to determine an appropriate feedback linearizing control law at each discrete time step. This approach avoids the requirement for exact knowledge of the plant and other difficulties associated with implementing analytical input-output feedback linearizing controllers for complex power systems. Simulation results demonstrate voltage tracking, damping of power angle oscillations and the possibility of user-specified dynamics.