An active learning approach for identifying the smallest subset of informative scenarios for robust planning under deep uncertainty. (May 2020)

• Record Type: Journal Article
• Authors: Giudici, Federico; Castelletti, Andrea; Giuliani, Matteo; Maier, Holger R.
• Is Part Of: Environmental modelling & software. Volume 127(2020)
• Publication Date: 2020-05
• Subjects: Robust optimization -- Deep uncertainty -- Active learning -- Robust planning -- Hybrid energy systems; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2020.104681
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Deep uncertainty in future climate, socio-economic and technological conditions poses a great challenge to medium-long term decision making. Recently, several approaches have been proposed to identify solutions that are robust with respect to a large ensemble of deeply uncertain future scenarios. In this paper, we introduce ROSS (Robust Optimal Scenario Selection), a novel algorithm that uses an active learning approach for adaptively selecting the smallest scenario subset to be included into a robust optimization process. ROSS contributes a twofold novelty in the field of robust optimization under deep uncertainty. First, it allows the computational requirements for the generation of robust solutions to be considerably reduced with respect to traditional optimization methods. Second, it allows the identification of the most informative regions of the scenario set containing the scenarios to be included in the optimization process for generating a robust solution. We test ROSS on the real case study of robust planning of an off-grid hybrid energy system, combining diesel generation with renewable energy sources and storage technologies. Results show that ROSS enables computational requirements to be reduced between 23% to 84% compared with traditional robust optimization methods, depending on the complexity of the robustness metrics considered. It is also able to identify very small regions of the scenario set containing the most informative scenarios for generating a robust solution. Highlights: A novel algorithm for robust planning under deep uncertainty is proposed. The algorithm uses active learning for selecting the smallest subset of informative scenarios. Computational time is significantly reduced with respect to traditional methods. Most informative scenarios are highlighted and exploited within the optimization process.