Case Study: Carnegie Mellon University achieves 20% energy savings with Microsoft Azure

Carnegie Mellon Sees a Way to Cut Energy Use by 20 Percent with Cloud Machine Learning Solution

Carnegie Mellon University’s Center for Building Performance and Diagnostics faced the challenge of managing large, disparate sensor data across campus to reduce equipment failures and wasted energy. Traditional monitoring provided real-time dashboards but lacked predictive analytics that were fast, easy to implement, and accessible to nontechnical building managers and technicians for proactive fault detection and operational optimization.

CMU extended its PI System and Azure infrastructure with Azure Machine Learning and Power BI to build models for fault detection (using valve water temperature as a proxy), predict solar radiation, and forecast morning indoor temperatures to optimize heating start times. The cloud solution cut model setup time from weeks to days, enabled collaborative, self-service analytics, and—based on experimental results—could reduce energy use by about 20%, potentially saving several hundred thousand dollars annually if deployed campus-wide.

Carnegie Mellon University

Bertrand Lasternas

Researcher