Principal Investigator Paul Barton
Natural gas is often produced and distributed by complex pipeline networks involving 10s of fields and 100s of wells that feed into natural gas liquefaction plants or the gas distribution network of a nation/continent. This project is investigating rigorous optimization (mathematical programming) formulations for short term (approx. 90 days) production planning in gas production systems. A mathematical programming formulation of this problem has the potential to generate automatically feasible production plans that are guaranteed to be globally optimal with respect to the chosen cost metric.
Production planning problems involve both binary (0–1) and continuous decision variables to be selected at optimal values. Binary variables are used to represent the temporal sequencing of events such as bringing elements of the network on or off line. Logical relationships that must be enforced between elements of the network and events can be represented by constraints in terms of these discrete variables. Continuous variables are used to represent quantities such as flowrates, compositions, inventories, etc. Nonlinearities in the model will arise from phenomena such as the mixing streams of differing composition and multiphase flow pressure relationships. These nonlinearities will make the feasible region of the problem nonconvex. Similarly, the complexities inherent in an economic model reflecting considerations such as tariffs, taxes and opportunities such as spot cargos, make the objective function for the problem nonlinear and nonconvex. The combination of these features yields a mixed-integer nonlinear optimization problem in which some of the participating functions are nonconvex (nonconvex MINLP). We are applying and extending our existing results on global optimization algorithms for MINLP to the gas production planning problem.