A data-centre cooling control technology based on physics-informed offline reinforcement learning — jointly developed by AIR, the Tsinghua Wuxi Institute of Applied Technology and Caituo Cloud Computing (Shanghai) — has passed the scientific and technological achievement evaluation organised by the Chinese Association of Refrigeration. The panel concluded that the overall technology has reached an internationally advanced level, and that the Time-Reversal Symmetry Thermodynamic Model (TTDM) at its core has achieved international leadership. The work points to a transferable paradigm for energy saving and carbon reduction in multi-variable, strongly coupled industrial systems.
https://air.tsinghua.edu.cn/info/1007/2537.htm