A Novel Machine Learning-Based Load-Adaptive Power Supply System for Improved Energy Efficiency in Datacenters
Michael Chrysostomou, Nicholas G. Christofides, Demetris Chrysostomou · IEEE Access · 2021
Power Supplies are a key part of the modern Internet and Communications Technologies (ICT) industry. Modern Uninterruptible Power Supply (UPS) systems are modular and as such, consists of several Power Supply Units (PSUs). Various PSU designs are used to optimize efficiency of operation at specific loading conditions, but this engenders inefficient operation at other loading conditions. To optimize the energy efficiency in various loading conditions, this paper proposes a novel power supply multiplexing system engaging different combinations of PSUs, controlled through machine learning techniques to maximize efficiency depending on the loading conditions. Each PSU combination is given a state number. Because of the vast number of combinations (States) that can occur in such systems and due to the redundancy requirements, that need to be met during multiplexing, machine learning techniques are adopted. It is shown that through the novel proposed system an efficiency improvement of over 78% can be achieved in low loading conditions, and an average 5.23% efficiency improvement in all loading conditions.