Resilience Enhancement of Electricity-Computing Interdependency System against Contingency

Abstract

Electrical power and computing power are highly interdependent especially due to the AI revolution. This dependency is largely underestimated during normal and recovery operation. To address this issue, an Electricity-Computing Interdependency System (ECIS) is developed in this paper, which first quantifies the interdependencies between ECIS and other critical infrastructures. The data center (DC) is modeled as a facility that transforms electrical power into computing power, and a recovery optimization strategy for ECIS is then proposed considering electrical power constraints, computing power supply, and infrastructure interdependencies. The case study demonstrates that the proposed strategy dynamically adjusts the DC electrical power quota according to system states, improving system resilience by more than 20% during power shortage.