Generalized disjunctive programming for heat exchanger networks synthesis

By introducing the conditional modeling technique into the mixed integer nonlinear programming (MINLP) model of heat exchanger networks synthesis (HENS), a generalized disjunctive programming (GDP) model for HENS was proposed. The GDP model adopts a set of logical variables to accomplish the conditional selection of constraints. Compared with the original MINLP model for HENS, the GDP model is more straightforward in the modeling stage. Moreover, the dynamic selection of constraints greatly reduces the complexity in the NLP sub-problems. An automatic elimination of the cost-computation terms for non-existing units also facilitates the solving process. Together with the GDP modeling of HENS, a novel hybrid method, with Tabu Search (TS) in the outer-layer and gradient-based optimizer in the inner-lay, was proposed to solve this GDP model. A comparison with traditional MINLP models for HENS was also conducted via case studies, and the advantages in model's simplicity and computational efficiency of using GDP can be obviously observed. The results show that the proposed GDP with the hybrid solver is effective in dealing with HENS.