(316d) Hierarchical Control of Networks Featuring Large Solvent Recycle

Motivated by the cost benefits resulting from the reduced consumption of make-up solvent and the environmental regulations demanding minimal discharge of toxic chemicals, solvent recovery and recycle is one of the essential features of integrated chemical processes. Examples of such processes include those in which a non-aqueous solvent is used as the reaction media, or as an extractant to purify a product or remove a contaminant. A characteristic common to most of these cases is the  presence of a large solvent recycle when compared to the fresh solvent feed.

In our previous work (Baldea and Daoutidis, 2007; Kumar and Daoutidis, 2002) it was demonstrated that  networks featuring flows of different orders of magnitude give rise to multi-time scale dynamics,  and thus present significant control challenges.  In this work, the defining characteristics of processes with significant solvent recycle are identified and a systematic framework for their analysis and control is developed. As a first step, a generic prototype network that captures the essential features of such processes is proposed. It is shown that the  material balance model of this network can be  recast in a singularly perturbed form,  resulting in two-time scale dynamics.  A model reduction framework is then developed to obtain reduced order models that capture the dynamics of this network in the fast and slow time scales.   Finally,  a hierarchical control strategy resulting from this time scale multiplicity is proposed. Specifically, the control objectives related to the individual units are addressed in the fast time scale, whereas those associated with the solvent recycle loop are addressed in the slow time scale.

A process for the production of 5-hydroxymethylfurfural (HMF), a key intermediate in the production of biomass derived fuels and chemicals, is considered as a case study. HMF synthesis is based on the acid catalyzed dehydration of sugars, and is the intermediate product of consecutive reactions. In order to improve its yield, HMF needs to be separated from the reaction medium with solvent extraction being a commonly used approach. In previous work (Torres et al., 2010) a continuous process for the production of HMF using solvent extraction was designed and optimized. The optimal steady state was characterized by a large solvent recycle flow when compared to the fresh solvent feed, a feature representative of the prototype network identified earlier. The model reduction and hierarchical control strategy developed for the prototype network is applied to this system, and model based controllers are derived to address control objectives (product purity and holdups) in each time scale. The effectiveness of the control scheme is demonstrated through closed loop simulations of the system under various operating scenarios.