This work describes the implementation of a Predictive Functional Control (PFC) algorithm on a ternary batch distillation column to control temperature by manipulating the reflux ratio. The PFC is tuned off-line by using simplified models obtained by applying identification techniques. The temperature set point is described by a polynomial related with an optimal dynamic behaviour determined to achieve the composition specifications. The separation performance is closely checked with the help of a soft sensor, based on a non linear Hammerstein model described in a previous work, which input is top temperature and the outputs are the estimated compositions. Additionally the model composition estimations are employed for calculating the accumulated compositions for each tank. Therefore a type of split range control for the corresponding valves of each tank is programmed in order to storage each component under quality requirement. Experiments were performed on a rigorous model developed in HYSYS. Plant^® with data of a real pilot column. All concerning to control policy was implemented in MATLAB. Results comparing optimal PID and PFC are presented.