Simultaneous optimization models for heat integration systems.

摘要

Several models are presented in this thesis for the systematic optimal design of grassroots and retrofit heat integration systems. Significant improvements over previous design methods are obtained due to the simultaneous nature of the proposed models. These models are not heuristic-based and can account for trade-offs between the different types of costs explicitly. Furthermore, the models provide a framework for automating the design of heat integration systems.; For grassroots problems, a simple yet general heat integration representation is introduced which is directly applicable to various problems. The representation is first formulated as an NLP for the simultaneous area and energy targeting of heat exchanger networks and the modeling of multi-stream heat exchangers. The model is then extended to an MINLP for the synthesis of heat exchanger networks. The entire network, including matches, area requirement, number of units as well as the configuration are defined by its solution. Finally, the representation is embedded within a process superstructure or flowsheet for the simultaneous optimization or synthesis of the process and its heat exchanger network. The significance of all these models is that through the use of the proposed representation, heuristics for heat integration, such as the consideration of the pinch point, specification of the heat recovery approach temperature (HRAT), or specification of minimum number of units are not necessary. Instead, the models can minimize cost by explicitly considering the trade-offs between the different design parameters. Furthermore, design constraints for heat integration can be easily incorporated into the formulations. Also, it is shown that certain simplifying assumptions make the solution of the models very efficient.; For the retrofit case, two models are presented. The first is an MILP assignment transshipment model which only accounts for the structural modifications that may be needed to change the level of energy recovery for the existing network. The second is a detailed MINLP model based on a novel superstructure which can accurately account for the many different arrangements of the existing equipment. Also, a prescreening stage is proposed to determine whether or not a retrofit project is economically feasible. This MINLP model, like grassroots models, does not require heuristic assumptions but optimizes energy and modification capital costs simultaneously. Many different types of retrofit possibilities, some of which are not so obvious, are considered by the model.