Resilience of data centre power system: modelling of sustained operation under outage, definition of metrics, and application

摘要

A novel criterion for quantifying the resilience of power systems supplying data centres is formulated to measure the system's ability to sustain functionality even during an outage. By comparative analysis of two alternative data centre power systems covering apparatus of electrical power supply and environmental control, it is shown that reliability and availability alone are insufficient as metrics to gauge different designs. The gap is bridged by the proposed resilience analysis to further evaluate situations of single and double outages. As a complement to the indicators of single point of failure and double point of failure, respectively, $N - 1$N-1 and $N - 2$N-2 security criteria, the novel metrics of a single point of reduced availability and double point of reduced availability are proposed. These criteria identify those single subsystems or subsystem pairs causing system availability to drop below requested levels in periods when they are out of service. The metrics so offer information on the overall system's availability during times of maintenance and failures. Thanks to this understanding, it is shown that a guided reduction of the number of subsystems considering their relative importance can lead to designs offering desirable trade-offs in terms of complexity, reliability, availability, and resilience.