The Flexible Energy Management of Artificial Intelligence Data Center Clusters

With the accelerating deployment of Artificial Intelligence Data Centers (AIDCs), new challenges have been introduced to energy systems due to the high-density computational workloads and substantial power demands. To address these issues, a flexible energy management framework for AIDC clusters has been proposed based on the concept of computing-power coordination. Within this framework, several models are integrated, including power consumption models in which workload characteristics are considered, a spatiotemporal scheduling model for elastic task migration, and operational models for electric energy routers (EERs), renewable energy systems, and storage devices. Coordinated power dispatch across different layers is enabled, allowing dynamic adjustment between power supply and IT demand. Through this work, contributions are made toward the advancement of sustainable AIDC design, where intelligent computing operations are aligned with evolving energy infrastructures, and both theoretical foundations and practical pathways for green, flexible, and future-oriented data centers are provided.