Massively Parallel GPU AC Powerflow Solver under Topology Modifications
ACM SIGEnergy Energy Informatics Review · 01.09.2026 · 6 (3) · 231-239
DOI: 10.1145/3861130.3861148 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM SIGEnergy Energy Informatics Review, online erschienen am 11.11.2026. Verfasst von Mikhail Farber und 2 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben.
Abstract (Originalsprache)
Grid operation in modern power grids requires solving a large number of power-flow problems with high precision. Applications such as topology optimization or static security analysis compute many similar, but slightly perturbed, grid topologies arising, e.g., from the failure of a single element. The traditional approach of using CPU-based Newton-Raphson solvers struggles under these workloads. Past research has shown strong potential to accelerate computations that are very similar using graphical processing units (GPUs). We present an accelerated solver based on the Fast Decoupled method within a preconditioned conjugate gradient scheme. Using custom CUDA kernels that leverage in-place updates, we batched hundreds of thousands of topologies into VRAM, enabling us to solve even larger contingency analysis tasks in a single batch. Despite high hardware utilization, we observe only a moderate speedup relative to a highly optimized CPU baseline. We hypothesize that the memory-bound nature of the AC powerflow problem, characterized by low arithmetic intensity, presents fundamental challenges for current GPU acceleration paradigms when compared to highly optimized multi-core CPUs.
Zitieren
APA
Farber, M., Westerbeck, N., & Koch, T. (2026). Massively Parallel GPU AC Powerflow Solver under Topology Modifications. ACM SIGEnergy Energy Informatics Review, 6(3), 231-239. https://doi.org/10.1145/3861130.3861148
Vancouver
Farber M, Westerbeck N, Koch T. Massively Parallel GPU AC Powerflow Solver under Topology Modifications. ACM SIGEnergy Energy Informatics Review. 2026;6(3):231-239. doi:10.1145/3861130.3861148
BibTeX
@article{Farber2026,
author = {Farber, Mikhail and Westerbeck, Nico and Koch, Thorsten},
title = {Massively Parallel GPU AC Powerflow Solver under Topology Modifications},
journal = {ACM SIGEnergy Energy Informatics Review},
year = {2026},
volume = {6},
number = {3},
pages = {231--239},
doi = {10.1145/3861130.3861148}
}
EUVIMED