- 일시: 2026년 6월 2일(화) 오후4시
- 장소: R관 715호
- 연사: 홍준호 교수, Department of Electrical and Computer Engineering, University of Michigan-Dearborn
- 제목: Generative AI for Power Grid Operations
- 초록: Recent advances in generative artificial intelligence (GenAI) have created new opportunities for improving situational awareness and operational intelligence in modern power systems. This presentation introduces a GenAI-based framework that enables large language models (LLMs) to understand supervisory control and data acquisition (SCADA) systems through multi-modal analysis of numerical measurements, communication information, operational rules, and human-machine interface (HMI) displays. The proposed approach integrates Set-of-Mark (SoM)-based visual guidance with power system operational knowledge to help GenAI models interpret SCADA environments, understand system topology, and identify abnormal operating conditions. Unlike traditional machine learning approaches that rely primarily on fixed training datasets, the framework combines contextual reasoning, physical consistency validation, and visual interpretation to analyze complex system behaviors. Several case studies are presented using IEEE benchmark systems and hardware-in-the-loop (HIL) environments, demonstrating the ability of GenAI to detect system anomalies, operational inconsistencies, abnormal state transitions, communication irregularities, and visualization mismatches in SCADA displays. The presentation also discusses the strengths and limitations of GenAI for engineering applications, including explainability, scalability, contextual understanding, and integration with trusted power system solvers. The results highlight the potential of GenAI as an intelligent orchestration and diagnostic assistant for future energy management systems and digital substations.
- 주관: 김홍석 교수 연구실