Solving large-scale retrofit heat exchanger network synthesis problems with mathematical optimization methods

Heat exchanger network optimization is a standard problem in process design. Various mathematical models and heuristics have been developed to help the designer in constructing the network. Different target procedures, like the pinch analysis, are widely used both in academia and industry. Another approach to find cost optimal network structures is to use mathematical programming methods. The advantage with mathematical programming methods is that a rigorous optimization of the structure, sizes of heat exchangers and utility usage can be carried out, whereas the designer makes these decisions if purely pinch-based tools are used. Even if much effort has been put on research within this area, many of the mathematical models consider only grassroot design, whereas most practical cases today seem to be retrofit situations. In addition, these models are likely to be either rigorous but not solvable for bigger (large-scale, real life examples) or deficient and solvable for large-scale problems. This paper takes an attempt to address these problems simultaneously and to develop a rigorous optimization framework based on both a genetic algorithm and a deterministic MINLP-approach and to present an extended model for large-scale retrofit heat exchanger network design problems.