Energy integration represents an essential part of modern process/ plant design. Heat exchanger networks (HENs) are typically the prime target area for process integration and extensive work has been carried out in terms of optimizing their steady state design. Several other research efforts have addressed the issue of controllability and control structure selection, e.g. based on structural considerations and relative gain array arguments for decentralized control. Most research however has mainly focused on regulatory control of HENs. The operation of modern plants, on the other hand requires that product grade and load transitions be carried out quickly and efficiently, thereby placing an additional requirement on HENs. Present day HENs need to be capable of a flexible, fast transient operation that effectively employs available internal energy streams and external utilities.

In the present paper, we consider generic networks of counter current heat exchangers, designed to minimize the external utility consumption at steady state. Detailed dynamic models of these networks comprise of systems of hyperbolic partial differential equations (PDEs). We establish that in the limit of tight energy integration (i.e. when large heat transfer coefficients, leading to small temperature differences, are present), the resulting model exhibits two – time scale behavior, with the fast dynamics associated with the temperature dynamics of the individual streams and the slow dynamics associated with the total enthalpy content of the streams in a unit.

We propose a model reduction framework for the resulting system of stiff hyperbolic PDEs that allows us to derive non-stiff models in each time scale. We also propose a controller design framework that exploits this time scale separation to achieve temperature regulation in the fast time scale and transition between operating points in the slow time scale. The analysis also leads to a classification of the manipulated inputs (internal streams and external utilities) that act in each time scale and can be used to achieve these control objectives. These theoretical results will be illustrated by simulations on prototype HEN configurations.