Computational Experience with a GRG Method for Minimizing Fuel Consumption on Cyclic Natural Gas Networks

Abstract: We consider the problem of determining how to operate the compressor stations in a natural gas pipeline system so as to minimize the total amount of fuel consumed in the system while keeping the technological constraints. This problem is very important in the real world, since the amount of fuel used in a typical system is very high. Given the nature of the process, this problem is modeled as a non linear programming (NLP) problem, where the decision variables are the mass flow rate in each arc of the network and the pressure value at each node. In this work we present a mathematical model for a steady-state version of the problem, highlight its technical and computational issues, which makes a hard problem to solve, and report our computational experience with a Generalized Reduced Gradient Algorithms over a large database of instances under a variety of network topologies and a number of different compressor stations with data taken from industry. Results were encouraging, since we were able to obtain many local optimum solutions for instances with cyclic topologies, which are considered among the hardest to solve.