CARNEGIE MELLON UNIVERSITY
Chengyi Cai
PhD Student · Heinz College
Operations Research.
Machine Learning.
Energy Systems.
I use operations research and machine learning to study reliable, low-carbon electricity systems and the growing energy demand of AI data centers.
BACKGROUND
About me
I am a PhD student at Carnegie Mellon University's Heinz College, working with Peter Zhang and Woody Zhu. My research brings together operations research, machine learning, and energy systems, with a focus on infrastructure planning under uncertainty.
I am particularly interested in how electricity systems can respond to uncertain growth in AI data center demand, and when adapting investment decisions over time improves reliability and cost.
I received my M.S. in Public Policy and Management–Data Analytics from Carnegie Mellon University in 2026, and bachelor's degrees in International Affairs and International Relations and Japanese from Jinan University in 2024.
- Low-carbon electricity systems
- Robust & multistage optimization
- AI data centers & the grid
- Data-driven energy planning
WHAT’S NEXT
Upcoming activities
- NOVEMBER 2026
I will attend the annual meeting in San Francisco and present my research on multistage power grid capacity expansion under uncertain AI data center demand.
- OCTOBER 31, 2026
I plan to attend the INFORMS DMDA Workshop in San Francisco.
- PLANNED
ML×OR Workshop
I plan to attend and explore connections between machine learning and operations research.
SELECTED WORK
Research & projects
CURRENT RESEARCH · POWER SYSTEMS & OPTIMIZATION
Power grid planning under uncertain AI data center demand
I develop multistage robust optimization models for electricity generation capacity expansion. This work compares investment plans fixed in advance with policies that adapt as demand becomes known, examining when that flexibility creates value.
Current experiments use an ERCOT-based network to study how the location and timing of demand growth, construction lead times, and network constraints affect investment, operations, and reliability.
ENERGY & CLIMATE POLICY · RESEARCH WITH VALERIE KARPLUS
Decarbonizing industrial heat in China
At CMU's Scott Institute for Energy Innovation, I have studied renewable natural gas and industrial heat pumps using levelized cost of heat models and lifecycle emissions analysis. This work also examines feedstock potential, certification systems, and policy pathways for industrial decarbonization.
SELECTED PROJECT · FORECASTING & CAPACITY EXPANSION
Data-driven electricity planning for California
I developed a GenX-inspired predict-then-optimize pipeline for a CAISO generation planning project, combining renewable energy forecasts and representative-day clustering with capacity and dispatch optimization, and testing sensitivity to carbon prices.
SELECTED PROJECT · SPATIAL ANALYSIS
Where should data centers be built?
I used GIS-based multi-criteria analysis to study U.S. data center siting, combining county-level indicators of renewable energy, disaster risk, grid reliability, and electricity prices into a composite index and spatial clusters.
BEYOND THE MODELS
Data visualization & policy storytelling
Selected visualizations and coursework, including a story on the human impacts of the energy transition in West Virginia.
Explore the archive →