Quantitative impacts of energy system flexibility provision and design/operation on data center reliability and the key influential parameters

AI applications drive the dramatic increase in data centers and their power demands. Data centers are usually equipped with uninterruptible power supply systems (e.g., batteries) and backup generators to ensure continuous and reliable operation. Battery storage systems also provide significant energy flexibility capacity for high renewable power grids, attracting increasing attention. However, the engagement in energy flexibility services could potentially compromise the reliability of different tiers of data centers to some extent and raise critical concerns. This study, therefore, comprehensively assesses the impacts of energy flexibility provision on the operational reliability of data centers across different tiers and their key influential design and operation parameters from the lifecycle aspect for the first time. The data center reliability is assessed by adopting probability-related (i.e., probability of power insufficiency) and energy-related (i.e., quantity of power insufficiency) criteria, concerning both design and operation parameters. Two typical global sensitivity analysis methods, FAST and PAWN, are adopted to identify the key influential parameters. The results show that engaging up to 40% of battery capacity for energy flexibility provision has an insignificant impact on data center reliability. The energy flexibility provision and the specifications of the backup generators are more influential for high-tier data centers in low-reliability power grids. The power grid reliability is more influential for lower tiers of data centers in high-reliability power grids. The battery backup time is more influential under a low ratio of battery engagement for energy flexibility provision.