IMPROVED SEQUENTIAL STRATEGY FOR THE SYNTHESIS OF NEAR-OPTIMAL HEAT-EXCHANGER NETWORKS

A sequential framework (LP, MILP, NLP) for Heat Exchanger Network Synt hesis (HENS) is described that reduces significantly the numerical pro blems caused by non-linearities, discontinuities and combinatorial exp losion in simultaneous MINLP models. The multiple trade-offs in HENS r elated to level of heat recovery, network topology and area distributi on as well as the trade-off between total annual cost, network complex ity and operability is dealt with by iteration, user interaction and a new MILP transportation model. This model, which is based on the same idea as our vertical MILP transshipment model, is crucial in the sequ ential approach, since it provides sets of matches that give designs w ith close to minimum area and cost when supplied to an NLP model for n etwork generation and optimization. In addition to the framework itsel f, this paper will focus on specific issues related to the sequential approach such as: (1) improving the selection of matches in the MILP m odel by accounting for temperature driving forces, (2) reducing the co mbinatorial problem of the MILP model by a tighter formulation, and (3 ) improving the robustness and efficiency of the NLP model by increase d use of information from the MILP model and the addition of convex es timators. Results will be presented for some of these problem areas.