A linear time-invariant model has been fitted to simulated results from the nonlinear air heating plant described in Part I. The model describes the plant well subject to uncertainty in two of its parameters. These uncertainties have been used to derive an uncertainty weight that, together with a performance weight, have led to an augmented plant description. The resulting augmented plant model was used to develop a robust controller that minimizes the H∞-norm of the closed-loop plant based upon the solution of two algebraic Riccati equations. The resulting controller is of the same order as the augmented (fifth-order) plant. Simulation results of the robust controller compare very favourably with a locally-optimized PID controller developed in Part I. The main advantage of the robust controller is that it merely requires a knowledge of the uncertainty limits of a simple linear model of the plant, whereas a PID controller of comparable performance would need extensive parameter tuning in practice.
|Journal||Building Services Engineering Research and Technology|
|Publication status||Published - Feb 2000|