While data-driven stochastic optimization has been established as a powerful tool for many applications, two important practical issues are often overlooked: (i) Data is assumed to be available at no cost, (ii) Information on data quality, which may differ based on measurement and underlying privacy-protection mechanisms, is ignored. This talk presents a novel modification of the established data-driven distributionally robust optimization (DRO) approach using the Wasserstein metric that addresses these two shortcomings. First, I discuss the ability of the Wasserstein metric to measure data quality and show how the Wasserstein DRO approach can be modified to accommodate data from multiple sources with individual transport budgets. Second, using an exemplary application from power system operations, I show how the resulting optimization problem implicitly computes the value of data given its quality and the context of the decision-making problem at hand. This creates an interpretable economic signal for efficient payments towards data owners and paves the way for a comprehensive economic analysis of data as a resource
Bio: Robert Mieth is a Leopoldina Postdoctoral Fellow at the Department of Electrical and Computer Engineering at Princeton University. He received the Doctorate in Engineering (Dr.-Ing.) from the Technical University of Berlin, Germany, in 2021. From 2018 to 2020 he was a Visiting Scholar and, from 2021 to 2022, a Postdoctoral Researcher at the Department of Electrical and Computer Engineering of New York University's Tandon School of Engineering. His research interests include risk analysis, stochastic optimization, data methods, and machine learning for modern power system operations and electricity markets.