Advanced Predictive Control Based on Multimodel Learning Techniques
Abstract
In this paper a new algorithm for local estimation for approximating a function by means of local models is introduced. The instant or in-time estimation provides an attractive alternative for nonlinear identification, since it requires less structural decisions to be made by the user. It represents a hybrid between local and global modeling. This estimation procedure is combined with predictive algorithm with a modification cost function for efficient control. Simulation results are provided to show the results of the proposed approach.
