4713. Accelerating energy system optimization on GPUs
Invited abstract in session ME-52: Decomposition approaches and structure-aware energy system optimization, stream Energy Markets & Systems.
Monday, 14:15-15:45Room: KOL – PC Raum 1
Authors (first author is the speaker)
| 1. | Benoît Legat
|
| INMA, UCLouvain |
Abstract
As energy system models grow in scale and complexity, exploiting domain-specific structure is critical for computational efficiency. This presentation highlights recent open-source advancements within the JuMP ecosystem that enable native GPU acceleration for large-scale energy system optimization. We will introduce GenOpt, a framework that allows practitioners to explicitly represent algebraic constraints as block structures that GPUs can directly exploit. We explore two approaches that leverages these blocks to accelerate automatic differentiation (AD): ExaModels, which evaluates constraints
Keywords
- Decomposition Techniques
Status: accepted
Back to the list of papers