Automatic synthesis of optimum heat exchanger network configurations

TLDR

This paper develops an automatic method to generate heat‑exchanger network configurations that minimize investment cost while meeting minimum utility cost and limiting the number of units. The method uses linear programming and mixed‑integer linear programming transshipment models to build a superstructure of alternative matches, then applies nonlinear programming to prune zero‑flow connections and produce realistic designs. The approach is demonstrated on three example problems, showing its applicability.

Abstract

Abstract This paper addresses the problem of automatically generating heat exchanger network configurations that feature minimum investment cost subject to minimum utility cost and fewest number of units. Based on the linear programming and the mixed‐integer linear programming (MILP) transshipment models for heat integration, a superstructure that has embedded many alternative configurations is proposed. This superstructure, which has as units the matches predicted by the MILP transshipment model, includes options for series and parallel matching, as well as stream splitting, mixing, and bypassing. Many of the implied stream connections in the superstructure are reduced to zero flow with a nonlinear programming formulation that leads to realistic and practical designs. Theoretical properties as well as the implementation aspects of the proposed procedure for the automatic generation of networks are presented. The method is illustrated with three example problems.