(132f) Global Optimization of Sizing Problem in Pipe Networks

Raghunathan, A. - Presenter, United Technologies Research Center

Sizing optimization of pipes in a specified network has been a problem of interest for the last 50 years. The problem has usually been formulated as a non-convex and non-smooth mixed integer nonlinear optimization problem. Consequently, most approaches proposed to solve this problem do not guarantee convergence to a globally optimal solution. In this talk, we present a novel convex mixed integer nonlinear programming formulation for the sizing optimization and describe an algorithm based on outer-approximations for the efficient solution. The proof of global optimality only requires that the pressure drop function governing the fluid flow in the pipes be monotonically increasing in the flows. We also present new insights into the hydraulic calculation problem in piping networks. We show the problem of hydraulic calculations can be cast as a strictly convex nonlinear program. This result is central to showing the global optimality of the approach. We present results from optimizing pipe sizes in water distribution networks available from the literature.