Projects

Current Projects

1. Characterizing Emerging Large Loads with LLM-Driven Models

Power Systems Engineering Research Center (PSERC), Award Number T-77

PI, with Co-PI Chee-Wooi Ten (Michigan Tech) and Co-PI Lang Tong (Cornell)

Project Period: August, 2026 - July, 2028

1. • Managing large electrical loads, such as data centers, industrial campuses, and high-performance computing facilities, is becoming a critical challenge for modern power systems, with direct implications for grid reliability, market operations, and renewable integration. These tens-to-hundreds-of-megawatt facilities exhibit highly dynamic, non-stationary demand patterns driven by computing workloads, cooling cycles, and operational policies. However, confidentiality, security concerns, and the absence of standardized data-sharing limit access to detailed measurements. To address this challenge, we will create a statistically and operationally realistic synthetic dataset to complement the limited real data available. Building on this foundation, the project will develop Large Language Model (LLM)-driven large load classification, disaggregation, and anomaly detection tools that work seamlessly with ISO, utility, and vendor analytics platforms. These capabilities will enable more accurate facility characterization, behind-the-meter analysis, and reliability assessment, preparing the T&D system for the next generation of high-impact loads.

2. A Proof-of-Concept Framework for Power System Synthetic Data Generation via AI Agentic Programming

GLRC/ICC Rapid Seedling Research Funding, Michigan Tech

PI, with Co-PI Jie Wu (Michigan Tech)

Project Period: August, 2026 - December, 2026

• Artificial Intelligence (AI) and the electric power system are becoming increasingly interconnected. AI technologies require reliable and flexible electricity, while the grid increasingly relies on AI-based tools to manage renewable energy, storage, electric vehicles, emerging large loads, and reliability challenges. A major barrier is that realistic power system data are difficult to access and share due to customer, operational, and infrastructure sensitivities. This project addresses that challenge by developing reusable, open-source software tools for generating trustworthy synthetic power system data under user-defined conditions. The tools will integrate power engineering constraints with AI-assisted programming methods to produce, test, and document physically meaningful and reproducible datasets. The expected outcome is an open-source prototype that includes generator code, selected synthetic datasets, evaluation methods, and documentation. This work will support privacy-preserving, FAIR, and reproducible data sharing in power engineering, while providing student training and preliminary results for future externally funded research.