In this contribution we analyze an industrial multi-component depropanizer
model and show how to rewrite the original model to index-one formulation t
hat is well adapted to dynamic optimization. The optimal control problems a
re then studied with the aim of determination of the optimal output traject
ories for minimum changeover time requirements and disturbance rejection. T
he problems studied are typical situations that frequently occur in the ind
ustrial plant. The optimization method used is the control vector parametri
zation (CVP) which consists in converting the original problem into a non-l
inear programming (NLP) problem which was solved by a successive quadratic
programming (SQP) method. The resulting profiles can be utilized as setpoin
ts for the existing real plant control. (C) 2000 Elsevier Science Ltd. All
rights reserved.