Improved models for operation modes of complex compressor stations

Abstract We study combinatorial structures in large-scale mixed-integer (nonlinear) programming problems arising gas network optimization. propose a preprocessing strategy exploiting the observation that large part of complexity arises certain subnetworks. Our approach analyzes these subnetworks and structure flows within order to provide alternative models with stronger can be exploited by off-the-shelve solvers. In particular, we consider modeling operation modes for complex compressor stations (i.e., ones several in- or outlets) networks. refined model allows precompute tighter bounds each mode number variants based on bounds. procedure obtain from input data original model. This is nontrivial reduction graph representing flow through station an mode. evaluate our reference benchmark data, showing they reduce average running time between 10% easy instances 46% hard instances. Moreover, three four considered networks, search tree nodes at least halved, effectivity guide solver’s search.