A dual approach to start-up of an adaptive predictive controller

This paper is concerned with the smoothing of start-up adaptation transients when controlling a priori unknown plants with an adaptive predictive control algorithm. The MUSMAR algorithm embodying an adaptive predictive controller relying on multiple models is considered. The approach followed consists of a dual modification of MUSMAR, including a bicriterial optimization which takes into account both the control objective and the parameter estimation criteria. The advantages of the proposed algorithm in what concerns reduction of the start-up adaptation transient are illustrated by experimental results obtained for the temperature control of a distributed solar collector energy plant.