An approach to feedforward distillation control based on inverse computation of nonlinear stage models is presented. The feedforward model calculates dynamic trajectories of manipulated variables from measured disturbances and product purity set points independently of the control configuration. Because the model includes the dominant dynamics and nonlinearities of the column, dynamic decoupling of the control loops is achieved. A superimposed linear controller only has to compensate model uncertainties and disturbances that cannot be measured. The proposed approach improves the control performance. Simulation studies show the applicability of the method to multicomponent distillation as well as to distillation trains. Experiments on a pilot plant scale binary distillation column verify the simulation results.