A. Alonso Ayuso, L. F. Escudero, J. F. Monge Ivars
The aim of this work is twofold. First, presenting several alternatives for defining ambiguity sets in mathematical optimization under uncertainty to be given by a set of probability distributions (PD) from where to generate sets of scenarios to represent the realization of the uncertain strategic and operational parameters in a multistage tree along a time horizon. And second, designing a mixed integer linear optimization deterministic equivalent model to the meta stochastic one, where strategic and operational ambiguity sets are considered in a distributionally robust optimization for multistage multiscale stochastic optimization. Two of the alternatives to build the ambiguity sets that are proposed are based on phi-divergence and Wasserstein distance schemes. In case there is not available a Nominal Distribution to contrast from, the third alternative assumes that some statistic information (means, covariance, etc.) is available for building the set.
Keywords: Stochastic optimization, distributionally robust two-stage optimization, multistage multicale scenario treess, supply network design
Scheduled
Integer and Combinatorial Optimization
June 10, 2022 4:00 PM
Grade Hall
