Researchers Min-Seung Ko, Jae Woong Shim, and Hao Zhu from the University of Michigan have developed a new control framework to help manage the significant power demands of hyperscale AI datacenters. Their work, published in the IEEE Transactions on Power Systems, focuses on integrating different types of energy storage to stabilize the grid and improve efficiency.
Datacenters, especially those supporting AI operations, can cause sudden and significant fluctuations in power demand, which can stress the electrical grid and lead to frequency and voltage instability. To address this issue, the researchers propose a hybrid energy storage system (HESS) that combines a battery energy storage system (BESS) and a supercapacitor (SC). The HESS is designed to smooth out power demand by handling different components of the load separately.
The system uses a high-pass filter to divide the datacenter’s power demand into slow and fast components. The slow components are managed by the BESS through a leaky-integral controller, while the fast components are handled by the SC using a phase-lead proportional-derivative controller. This controller is enhanced with feedforward and ramp-tracking compensation to improve performance. Additionally, the system employs adaptive weighting and repetitive control mechanisms to further enhance its ability to respond to transient and periodic changes in demand.
The researchers tested their framework through case studies, which demonstrated that the HESS effectively suppresses both ramping and fluctuations in power demand. The system also helps stabilize the grid frequency and maintains sustainable state-of-charge (SoC) trajectories for both the BESS and the SC over prolonged, stochastic training cycles. This approach could be particularly useful for datacenters looking to reduce their impact on the grid and improve their energy efficiency.
The practical applications of this research are significant for the energy sector, particularly for datacenters and other facilities with variable power demands. By integrating different types of energy storage and using advanced control strategies, facilities can better manage their power consumption and contribute to grid stability. This can lead to more reliable and efficient energy systems, benefiting both the facilities and the broader electrical grid.
Source: IEEE Transactions on Power Systems
This article is based on research available at arXiv.