Optimal model switching for gas flow in pipe networks.

Rüffler, Fabian; Mehrmann, Volker; Hante, Falk M.

We consider model adaptivity for gas flow in pipeline networks. For each instant in time and for each pipe in the network a model for the gas flow is to be selected from a hierarchy of models in order to maximize a performance index that balances model accuracy and computational cost for a simulation of the entire network. This combinatorial problem involving partial differential equations is posed as an optimal switching control problem for abstract semilinear evolutions. We provide a theoretical and numerical framework for solving this problem using a two stage gradient descent approach based on switching time and mode insertion gradients. A numerical study demonstrates the practicability of the approach.

GAS flow; COMBINATORICS; PARTIAL differential equations; NATURAL gas; MATHEMATICAL optimization

Networks & Heterogeneous Media, 2018, Vol 13, Issue 4, p641

1556-1801

Academic Journal

10.3934/nhm.2018029