Abstract: Over the past decades, electric power systems have experienced major changes due to two main drivers: the liberalization of energy markets and the rise of renewable energy. While market liberalization has placed increased focus on optimal utilization of generation and transmission assets, the variability of renewable energy generation has lead to higher levels of uncertainty in system operation. Put together, the electric grid today presents a fascinating, complicated challenge for stochastic optimization.

In this talk, I will first discuss the optimal power flow (OPF) problem and our work in developing chance-constrained OPF formulations that are tractable for large-scale instances. To achieve tractability, our chance-constrained OPF formulations restrict the recourse actions to affine control policies. Motivated by the wish to discover more optimal recourse policies, we investigated the use of machine learning to obtain optimal solutions directly from the input parameters. In the second part of the talk, I will present our approach to learning for optimization, which is based on discovery of active sets.

Bio: Line Roald received her M.Sc. in mechanical engineering (2012) and Ph.D. in electrical engineering (2016) from ETH Zurich, Switzerland. She is a postdoctoral researcher in the Advanced Network Science Initiative at Los Alamos National Laboratory, and will join University of Wisconsin, Madison as a faculty member in August 2018. Her research focuses on modelling and optimization of energy systems under uncertainty.