AI Data Centers: A Review of Power Quality Challenges and Mitigation Strategies

The accelerated growth of AI data centers has introduced a new class of large and highly volatile loads to the power grid. Unlike traditional data centers, which are dominated by relatively steady and predictable IT loads, AI data centers exhibit rapid power transients, leading to increased harmonic injection and a higher likelihood of power system instability. This work examines the characteristics of AI data centers’ workload and evaluates methods to mitigate them. Specifically, artificial load modulation, GPU ramp-rate tuning, power capping, and the integration of multi-level energy storage to provide supplementary power during dynamic load fluctuations are reviewed. Several energy storage methods are discussed to reduce transients due to power oscillations at AI data center interconnections and to improve the electric grid’s power quality. New power delivery architecture and system-level coordination are also identified as promising approaches to alleviating the power quality challenges.