(237e) Synthesis of Heat Exchanger Networks with Optimal Placement of Utilities

Heat exchanger network synthesis has been widely studied because the economical and environmental benefits that its implementation produces. The approaches to solve the heat exchanger network synthesis problem have been classified as simultaneous and sequential techniques. Sequential techniques decomposed the problem to determine first utilities, number of units and area targets and then synthesize the network to satisfy theses targets; therefore, sequential techniques do no consider the trade-offs between the capital and operational costs, and as a consequence they may not yield an optimal network. On the other hand, simultaneous techniques solve the problem considering the trade-offs between those factors, and they are usually based on MINLP techniques. One of the first simultaneous techniques is the SYNHEAT model reported by Yee and Grossmann (1991). One interesting feature of the SYNHEAT model is that all the constraints are linear. It uses a superstructure in which the problem is decomposed into stages, and in each stage the process streams are divided and directed to all possible matches with the other streams, after which the streams are mixed at the same temperature. In the SYNHEAT model, the utilities are located at the extremes of the superstructure. However, because the temperature level for the utilities and process streams, there are cases when such situation may produce nonoptimal, or even infeasible solutions. For example, consider the case when the target temperature for a cold process streams is higher than the inlet temperature of the hot utility, but there is a hot process stream with an inlet temperature high enough to heat such cold process stream; in this case, the Yee and Grossmann superstructure must be modified to allow the intermediate use of utilities. In addition, because of temperature levels and film heat transfer coefficients, there may be cases when it would be preferable to locate the hot and cold utilities parallel to the process - process matches to yield a cheapest configuration. In this work, a new representation (see Figure 1) that allows the intermediate localization of utilities in the network configuration is proposed. All the constraints to model this configuration are linear, and the nonlinearities appear only in the objective function. The model is able to manipulate isothermal process streams that exchange their latent heat in addition to process streams that exchange sensible heat. The application of the proposed model is shown through several example problems, which show that the new formulation is able to yield better results than previously reported methodologies.