On NLP and MINLP Formulations and Preprocessing for Fuel Cost Minimization of Natural Gas Transmission Networks

Abstract: The problem of minimizing fuel consumption on natural gas pipeline networks is addressed. Both a nonlinear programming model and a mixed-integer nonlinear programming model are presented. A database containing many problem instances under different types of topologies is proposed and described. For a more efficient application of optimization algorithms, preprocessing techniques for this problem are presented, discussed, and computationally evaluated. It is found the use of three techniques provides a significant algorithm performance improvement reducing considerably many of the numerical difficulties inherent to this very complex problem. In addition, a preliminary computational evaluation of an outer approximation with equality relaxation and augmented penalty algorithm for MINLPs is presented. In initial findings, the algorithm reports promising results by finding optimal solutions to many problem instances.