The extractive distillation of ethanol using glycerol as the entrainer was studied to determine its optimal control profiles when the azeotropic feed was subjected to composition disturbances. The process was modeled by a differential-algebraic equation (DAE) system that represents the dynamics of the equilibrium stages in the extraction column. The model equations were solved by discretizing the time domain using orthogonal collocation on finite elements. Initially, the effects of feed disturbances on the product flow rate and quality were analyzed. Subsequently, a profit objective function was formulated, and the optimal profiles of the manipulated variables (reflux ratio and reboiler duty) were determined, subject to quality constraints. The solution was obtained by solving the nonlinear programming (NLP) problem that resulted from the discretization. The problem was solved in GAMS using IPOPT as the nonlinear solver, testing two different linear solvers, the Harwell subroutines MA57 and MA86. The optimal control strategy was compared to a simple PI control scheme.