Data centers are experiencing unprecedented growth as the fourth industrial revolution’s supporting pillars and the engine for the future digitized world. However, data centers are carbon-intensive enterprises due to their massive energy consumption, and it is estimated that the data center industry will account for 8% of global carbon emissions by 2030. In Singapore, data centers consume large amounts of electricity and water and have grown significantly in capacity during the second decade of this century. My research focus on developing a physics-informed digital twin system that can precisely quantify the carbon footprint of each device in the data center, ranging from ICT system and facilities.
The digital twin system is the digital counterpart of a physical data center, which continuously synchronizes with the physical one via real-time sensory data from the data center. To address the data scarcity issue in real-world data centers, I propose adopting a physics-informed manner to build the digital twin system by empowering it with knowledge about the physical process inside a data center. It will make the system robust and extrapolate better in the small-data region. It also provides valuable prediction of the complicated data center dynamics, enabling predictive management. Furthermore, it can synthesize large amount of data to train an industrial AI-driven management system, which derives the carbon-aware data center operation policies. In summary, my research objective is to understand, quantify, and minimize the carbon footprint of a data center through the lens of a physics-informed digital twin and industrial AI.
