To control the sequencing of the batch distillation steps and to guarantee that all products will meet their purity specifications, this work has proposed an inferential control system for high-purity multicomponent batch distillation columns, that is able to cope with difficulties commonly encountered in practice, including lack of on-line composition analyzers, plant/model mismatch, uncertainty in the initial system state, and measurement noise. An extended Kalman filter has been implemented to provide updated composition estimates from temperature measurements, and globally linearizing controllers have been designed to control the distillate composition, The choice of the number and location of temperature sensors, as well as the sensitivity of the controller settings to changes in the operating conditions, has also been discussed. Tight composition control was achieved, and the proposed inferential control scheme was shown to be robust and suitable for practical implementation.