A Comparison between NLP and MILP Formulations of Organic Rankine Cycle Systems for Optimization

Abstract

Abstract In chemical process optimization, often one type of formulation can be more advantageous due to financial constraints, compatibility issues with modelling software, the availability of optimization solvers and/or the importance of establishing global optimality. Historically, the availability and computational efficiency of linear program (LP) solvers, or even mixed integer linear program (MILP) solvers, were significantly better than some nonlinear program (NLP) solvers, which was a key incentive to reformulate nonlinear optimization problems to LPs or MILPs. However, with the current advanced NLP solvers and the recent improvements in optimization modelling platforms and techniques, it might now be redundant for practitioners to reformulate certain real-world NLP problems into integer/linear programming problem. This paper shows that with the right integration between the NLP formulation and the state-of-the-art solvers, it is possible to achieve significantly better optimization performance for real-world organic Rankine cycle systems.