Volume 52

Global Sensitivity Analysis of Key Parameters for Data Center Power and Energy Systems Considering Reliability Yang Yu, Kui Shan, Shengwei Wang

Abstract

To provide continuous and trustworthy services to users, contemporary data centers rely significantly on a stable power supply. This power is generally obtained from main energy sources and is frequently supported by on-site uninterruptible power supply systems, which include battery storage and renewable energy solutions. However, a precise understanding of how various parameters within power and energy systems affect the reliability of the data center’s power supply remains largely unexplored. This research, therefore, evaluates the influence of several critical parameters in data center power and energy systems on reliability, utilizing two prevalent global sensitivity analysis techniques: the Morris method and the Sobol method. The findings reveal that the likelihood of power grid failure, the failure probability of diesel generator and the wind turbine failure rate are the primary determinants of reliability. Moreover, the Morris method proves to be more efficient than Sobol, reducing computation time by approximately 48.9% to 49.7%, while still delivering comparable outcomes in certain cases.

Keywords data center, reliability, power system, energy system, global sensitivity analysis