Working with our project partner Open Grid Europe, who operates one of the largest and most complex gas network in Germany, we developed a Navigational System Tool (NAVI) for supporting dispatchers. Based on a prognosis of supply and demand at entries and exits, the NAVI makes suggestions for active elements, e.g., whether or not to close a valve and/or to ramp up a compressor machine. A non-convex non-linear MINLP can be formulated to ‘exactly’ model this problem, however in practice it does not nearly scale to the necessary network size or number of time steps. Our current approach is split into two-stages, both of which utilise MILP as approximations of their non-linear counterparts: The first takes a broader approach, capturing the transient aspects of gas-transport but only using a simplified view of compressor stations. The second stage then realises decisions based on the flow / pressure values provided by the first-stage for complex intersection areas of the original network. We conclude with visualisations of our computational experiments.
Zuse Institute Berlin
I’m currently a PhD student at Technical University Berlin, writing my thesis within the Energy Group at the MODAL Research Campus in Zuse Institute Berlin. I’m interested in combinatorial as well as network optimisation problems. The major goal of my current research is the on-line creation of a complete set of decisions for effective gas network operation while fully encapsulating the transient constraints. My previous work was centred around Steiner Trees, with both my B.Sci / Masters being completed at The University of Melbourne. In addition, a brief stint between semesters was spent at Monash University working on a crew scheduling problem.