This paper presents an application of the general identification metho
dology to obtain neural predictors for use in a nonlinear predictive c
ontrol scheme derived from the generalized predictive controller (GPC)
structure. Every step of the design procedure is illustrated using a
real-life problem: the identification of a solar power plant for predi
ctive control. Different solutions are discussed for the problems that
arise in the application of the methodology. Experimental results are
given, showing the performance of the predictor and the controller. (
C) 1998 Elsevier Science Ltd. All rights reserved.