A Smith predictive based MPC in a solar air conditioning plant Amparo N u~ nez-Reyes a, * , Julio E. Normey-Rico b , Carlos Bordons a , Eduardo F. Camacho a a Escuela Superior de Ingenieros, Dpto. Ingenier  ıa de Sistemas y Autom atica, Univ. de Sevilla, Camino de los Descubrimientos s/n, Sevilla 41092, Spain b Dpto. Automac ß~ ao e Sistemas, Univ. Fed. de Santa Catarina, Florian opolis/SC 88.04-900, Brazil Received 8 September 2003; received in revised form 4 May 2004; accepted 6 May 2004 Abstract This paper presents the application of a Model Predictive Controller to the temperature control in a solar air conditioning plant. The controller uses a Smith Predictor and includes a feed-forward control action to reject disturbances caused by solar radiation and the auxiliary gas heater. The tuning procedure is simple and allows a good compromise between robustness and performance. The behaviour of the controller is illustrated by experimental results. Ó 2004 Elsevier Ltd. All rights reserved. Keywords: Model predictive control; Robustness; Time-delay system; Solar energy; Thermodynamics; Smith Predictor 1. Introduction The use of solar energy for air conditioning systems is one of the most evident though not sufficiently exploited applications of this source of renewable energy. The use of solar radiation for cooling allows for time synchro- nization between solar supply and refrigeration demand since cold air is, in general, most in demand when solar radiation is high, thus reducing the need for storage systems which are one of the drawbacks of the use of solar energy for heating. Different procedures exist for refrigeration using solar radiation as primary source (see [2]). One of the most successful methods is by means of an absorption ma- chine which produces cold water when hot water is in- jected into its generator. As the water is heated by the sun, this type of air conditioning system reduces con- ventional energy consumption and therefore contributes to the conservation of the environment. However, operating this type of system presents cer- tain particular issues that must be addressed by the control strategy. Firstly, the primary energy source (the sun) cannot be manipulated, as in the case of any other conventional thermal process. Secondly, great distur- bances exist in the process, mainly due to changes in environmental conditions. There are also dead-times related to fluid transportation which are variable depending on the operating conditions. Finally, the cooling demand is variable since it depends on the occupancy rate and the kind of activity that is being carried out in the cooled space (laboratory in this case). Several control strategies have been tested at solar power plants to address these problems (see, for in- stance, [3,7,8,13]). Model Predictive Controllers (MPC) have received a lot of attention in the last few decades, both within the research control community and in industry [6]. The basic idea is to calculate a sequence of future control signals in such a way that it minimizes a multistage cost function defined over a control horizon. This paper deals with one of the most used algorithms: Generalized Predictive Control (GPC) [5]. The success of GPC de- pends, to a great extent, on the process model chosen. Modelling errors or uncertainties affect the behaviour of GPC. In order to improve the robustness of the closed- loop system when model uncertainties are considered, different strategies have been proposed in the literature. The effects of a prefilter (normally called T-polynomial) have been analysed, first in [14] and later in [1,15]. A more complete analysis was presented in [12] where a * Corresponding author. Tel.: +34-954487360; fax: +34-954487340. E-mail address: amparo@cartuja.us.es (A. N u~ nez-Reyes). 0959-1524/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.jprocont.2004.05.001 Journal of Process Control 15 (2005) 1–10 www.elsevier.com/locate/jprocont