Robust optimal control method based on constrained neural network for liquid-cooled data centers

Shengli Qu, Yuxiang Guo, Yiwei Feng, Chuang Wang, Ziwen Xing and Kaiwen Duan

Energy, 2025, vol. 333, issue C

Abstract: Data centers are infrastructures that provide arithmetic for the Information Technology industry which has developed rapidly in recent years. The refrigeration systems in data centers operate continuously throughout the year and consume a lot of energy, so it is crucial to reduce the energy consumption of the systems. The goal of refrigeration systems in data centers is to maintain the chip temperatures within the safe ranges to avoid irreversible damage. However, the lower the target chip temperature of the cooling system, the more energy it consumes. Therefore, the control strategy that keeps the chip temperatures exactly at the upper limits is the most energy-efficient strategy. However, the uncertainty makes the optimal control strategy difficult to be applied in practice. Therefore, this paper innovatively proposes a robust optimal control method to overcome the negative effect of uncertainty on the optimal control strategy. The results indicate that when the scaling factor of the loss function is equal to 40, the actual chip temperatures under the robust optimal control strategy are always within the safe ranges. In addition, the robust optimal control method can save at least 24 % energy compared with conservative control strategies with safety margin above 1.5 °C.

Keywords: Data center cooling; Uncertainty; Robustness; Optimal control strategy; Constrained neural network; Loss function; Energy-saving (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0360544225031317
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:energy:v:333:y:2025:i:c:s0360544225031317

DOI: 10.1016/j.energy.2025.137489

Access Statistics for this article

Energy is currently edited by Henrik Lund and Mark J. Kaiser

More articles in Energy from Elsevier
Bibliographic data for series maintained by Catherine Liu ().