Feedback linearization (FL) has recently been used to control the temperature in a solar field using the water flow rate as the manipulated variable. Nevertheless, the system shows undesirable control signal oscillation caused by a delay uncertainty. In this paper, the dynamics resulting from applying the FL transformation to the nonlinear process model is approximated by a first-order model plus dead-time, and a filtered Smith predictor (FSP) is proposed as the control strategy. This controller deals with dead-time errors but also includes control signal saturation caused by flow rate constraints. Simulations demonstrate the robustness of the FSP structure in comparison to a simple PID and a Smith predictor (SP) approach when a faster closed-loop is required in a system with delay uncertainties. Real experimental results included show the improved behavior of the proposed control scheme. (C) 2011 Elsevier Ltd. All rights reserved.