Keunju Song

Hi there! I am an incoming senior researcher at Korea Electronics Technology Institute (KETI). I received my Ph.D. degree in Electronic Engineering from Sogang University in 2026 and received my B.S. degree in Electronic Engineering from the Tech University of Korea in 2021. In 2025, I worked as a visiting graduate student at Argonne National Laboratory, contributing to physics-informed graph learning and grid foundation-model research for AC optimal power flow.

My research focuses on energy AI, power systems, graph neural networks, physics-informed learning, and foundation models. I am particularly interested in scalable and transferable AI methods that respect the topology and physical constraints of real power systems.

News

Jun 2026SABLE, a GPU-based power-flow accelerator for sparsity-aware batched learning, was released as a preprint.
Jun 2026Our physics-informed graph learning work for large-scale AC-OPF with topology changes appeared at the ICML AI4Science Workshop.
May 2026LUMINA, a grid foundation-model benchmark for AC-OPF surrogate learning, was released.
Mar 2026Our topology-transferable ACOPF foundation-model work appeared at the ICLR FM4Science Workshop.
2025Visited Argonne National Laboratory and contributed to physics-informed graph learning and grid foundation-model research.
2025Received the Best Graduate Student Award / Sogang Ricci Engineering Academic Award.
2024Our graph-based probabilistic PV forecasting work was published in IEEE Transactions on Sustainable Energy.
2024NICE_Forecast placed 11th overall and 3rd among student teams in HEFTCom24.

Research

Physics-Informed Graph LearningGraph neural networks for AC-OPF, topology changes, physical feasibility, and large-scale power-system learning.
Grid Foundation ModelsPretraining, transfer learning, zero-shot adaptation, and federated learning across heterogeneous grid topologies.
Renewable Energy IntelligenceSpatiotemporal probabilistic forecasting, missing-data robustness, and uncertainty-aware renewable generation learning.

Work Experience

Publications

SABLE: GPU-Based Power Flow Accelerator for Sparsity-Aware Batched Learning2026
arXiv preprint Paper
Suho Park, Keunju Song, Hongseok Kim
Physics-Informed Graph Learning Acceleration for Large-Scale AC-OPF with Topology Changes2026
ICML 2026 Workshop on AI4Science Paper
Keunju Song, Kyungnam Park, Sua Choi, Seunguk Kim, Tae-un Kim, Youngmin Choi, Sang-Won Min, Hongseok Kim
LUMINA: A Grid Foundation Model for Benchmarking AC Optimal Power Flow Surrogate Learning2026
arXiv preprint Paper
Hongwei Jin, Keunju Song, Zeeshan Memon, Yijiang Li, Stefano Fenu, Hongseok Kim, Liang Zhao, Kibaek Kim
LUMINA: Foundation Models for Topology Transferable ACOPF2026
ICLR 2026 Workshop on Foundation Models for Science Paper
Yijiang Li, Zeeshan Memon, Hongwei Jin, Stefano Fenu, Keunju Song, Sunash B Sharma, Parfait Gasana, Hongseok Kim, Liang Zhao, Kibaek Kim

View full publication list →

Awards

Services

Journal Reviewer

Last updated: August 2026