Session: Modeling Oriented Software and Methods (1)
Chair: Jean-Paul Watson
Cluster: Computational Software
Talk 1: Advances in Modeling and Solving Stochastic Programs
Speaker: Jean-Paul Watson
Abstract: TBd
Talk 2: (Canceled)What's new in Pyomo?
Speaker: Bethany Nicholson
Abstract: TBD
Talk 3: Implicit and explicit model reformulations for nonlinear optimization
Speaker: Robert Parker
Abstract: Variable aggregation is a common presolve technique in (mixed-integer) linear optimization for reducing the numbers of variables and constraints seen by the core algorithm. We study this technique in the context of nonlinear local optimization with the goal of improving solver robustness by taking a “more feasible” path to the solution. In contrast to mixed-integer linear optimization, nonlinear local optimization allows the possibility of an implicit reduction. We first provide a case study in which a well-chosen implicit reduction significantly improves convergence reliability of a moving bed reactor optimal control problem. Second, we propose an algorithm for automatically detecting largeexplicit reductions and present its behavior on example problems from chemical engineering.
