The design of natural gas transmission involves a high number of alternatives that are normally reduced applying criteria based on previous experiences and/or usual engineering practices (Pipeline Design and Construction – A Practical Approach. Mohitpour, Golshan and Murray, 2003). In many instances, this design procedure is presented as a trial and error process among several candidate designs proposed beforehand. The usual decision variables are the route(s), the pipeline diameter (usually the same for the whole pipeline) and the compression/re-compression stations as well as the design pipeline pressure. The usual evaluation methods include the simulation of the chosen pipeline fro a variety of flowrates evaluating the work expenditure. Decisions regarding future expansions (addition of loops, which are new pipelines in parallel for some segments, and additional compressors) based on increasing forecasted demand are rarely proposed as part of the initial design procedure.

With the exception of a very nice paper by Edgar, Himmelblau and Bickel (Optimal design of Gas transmission Networks, SPE Journal, 1978), very little attention was given to the use of mathematical optimization methods to design gas transmission networks.

In this paper, we developed a model that builds on the concepts advanced by Edgar et al., and add some modern advances in process systems engineering. We propose a general superstructure of nodes, some being consumer nodes, some supplier nodes, some, simply transmission nodes. We also incorporate the time dimension and possible expansions through loops as well as added compressors. Finally, we optimize the NPV given a certain demand profile. Contracts between gas suppliers and gas consumers including penalties are also considered.

To solve this model we have prepared two versions. One is a straightforward MINLP version, which performs well when looping is not allowed. To ameliorate the computational difficulties an MILP model was constructed based on the discretization key variables, which we claim can capture a solution close to the global optimum.

The model is capable of determining the optimum design variables (routes, pipeline diameters, and compressors locations and capacities), strategic decisions (annual purchased amount that should be agreed and amount of gas that should be sold to the consumers) and the annual operational conditions (station pressures and flowrates through the pipes).