High performance computing for energy system optimization models: Enhancing the energy policy tool kit. (May 2019)

• Record Type: Journal Article
• Authors: Sharma, Tarun; Glynn, James; Panos, Evangelos; Deane, Paul; Gargiulo, Maurizio; Rogan, Fionn; Gallachóir, Brian Ó
• Is Part Of: Energy policy. Volume 128(2019)
• Page Start: 66
• Page End: 74
• Publication Date: 2019-05
• Subjects: ESOMs Energy System Optimization Models -- HPC High Performance Computing; Energy system optimization models -- High performance computing -- Solution strategies; Energy policy -- Periodicals; Politique énergétique -- Périodiques; Electronic journals; 333.79
• DOI: 10.1016/j.enpol.2018.12.055
• Languages: English
• ISSNs: 0301-4215

Abstract

Abstract: Energy system optimization models (ESOMs) form a critical component of a suite of modelling tools used by policy makers to understand (i) evolving complexity in energy systems arising from intersectoral coupling and other considerations at different spatial and temporal resolutions and (ii) uncertainty and sensitivity to assumptions and model parameters which entails analysis of a multitude of scenarios. Such enquiries are partly restricted by increasing computational times which can range from hours to days. To appease this restriction, we report our attempts at formalizing the performance testing of running ESOMs on a High Performance Computing (HPC) facility. The goal is to provide an assessment of the potential of a HPC environment to minimize solution time. Reporting on the outcomes, we present the scaling performance of the Irish TIMES, ETSAP-TIAM and JRC EU TIMES models by demonstrating solution time improvement on scaling across components of a HPC facility. Such facilities permit parallel runs of model instances. We identify and characterize the benefits and trade-offs of forking as a strategy in solution time reduction. Such capability permits policy makers and modellers to pose and derive insights to increasingly relevant questions on inter-sectoral coupling and risks that energy systems face due to uncertainty. Highlights: Ported JRC EU TIMES, ETSAP TIAM and Irish TIMES to High performance computing. Significant reduction in solution time. Permits policy makers to pose and modellers to solve complex policy problems. Characterize forking as a solution strategy for batches of scenarios.