Proceedings of the 1 st Annual Gas Processing Symposium H. Alfadala, G.V. Rex Reklaitis and M.M. El-Halwagi (Editors) © 2009 Elsevier B.V. All rights reserved. 1 Towards the dynamic initialization of C4 splitter models Ivan Dones a , Heinz A. Preisig a , Stefan de Graaf b aDept. of Chemical Engineering – NTNU, Sem Sælandsvei 4, 7491Trondheim, Norway bCybernetica AS, Leirfossveien 27, 7038 Trondheim, Norway Abstract A simple initialisation of distillation models is presented. It is robust and shows to have a much wider convergence radius for this notorious problem than the traditional approaches. The procedure uses a model that, compared to the traditional models, adds the gas dynamics. The current implementation has also been used to implement a new portable thermodynamic module, which adds significant flexibility and accuracy to the use of thermodynamic information. The new approach is evaluated against the traditional method on an industrial C4 splitter taken from a gas plant. Keywords: add initialisation, dynamic model, distillation. 1. Background With energy being a major issue of today's society, all activities around the production of natural gas are of particular interest. The main operation performed in a gas plant is separation of different components and, if required, liquefaction of the products. Separation is done in distillation towers, some of which are referred to as splitters. Being able to model distillations is essential for many tasks, for instance design, construction, retrofitting and control. Steady-state models are most commonly used in design and retrofitting, but also in some control and planning applications. If dynamic models are being used, the most adopted dynamic models make at least partial steady- state assumptions. In particular the gas-phase dynamics and the tray mixing dynamics are neglected. The computation of steady-state distillation models is notoriously difficult, the main reason being that it consists of a large number of heavily coupled, algebraic equations, the size of which proliferates with the number of components and number of trays. Solving such a steady-state model implies solving the large-scale algebraic equation system for a single set of roots that represents the equilibrium state of the distillation. The numerical methods being used for this purpose have usually a very small convergence radius, which, in the past, gave rise to long lasting research: C. D. Holland and others (1988, 1993), Taylor and Lucia (1996), Rabeau et al. (1997) and more recently Grossmann et al. (2005). Similar convergence problems are observed for partially simplified dynamic models, mainly for the algebraic parts. The aim of this research is to create a fully dynamic distillation model that is robust with respect to the initial conditions and under normal operating conditions can be computed an order-of-magnitude faster than real-time. Here the authors mainly compare