Hyperscale AI datacenters generate complex, workload-driven power fluctuations that appear as time-varying disturbances at grid interconnection points. Unlike conventional loads that vary smoothly, AI workloads—particularly model training and fine-tuning operations—create structured active-power swings that can excite generator frequency dynamics and threaten grid stability. Traditional peak- or average-load representations fail to capture these characteristics, requiring new mitigation strategies.
Researchers have proposed a hybrid energy storage system with differentiable predictive control (HESS-DPC) to address this challenge at the datacenter source. The framework combines battery energy storage systems (BESS) for longer-duration deviations with supercapacitors (SC) for rapid, high-frequency variations. A frequency-based rule-based controller initially allocates power deviations between these components based on their operational characteristics.
To improve performance beyond fixed-frequency decomposition approaches, the team trained a residual predictive control policy offline that computes optimal command corrections around the baseline rule-based controller while maintaining real-time safety constraints. This hybrid approach leverages the strengths of both deterministic control and machine learning optimization.
Simulation testing on the NPCC 140-bus system demonstrated substantial improvements. The HESS-DPC framework reduced residual grid-side power deviations during critical workload transitions, maintained supercapacitor state-of-charge within sustainable operating ranges over extended periods, and achieved more than 80 percent reduction in generator peak-to-peak frequency deviations. The worst-affected generator's frequency response improved from 15.1 mHz to 1.3 mHz.
These results validate that local, source-side power smoothing can effectively mitigate frequency disturbances before they propagate into bulk power systems. As AI datacenter deployments accelerate globally, such mitigation technologies become increasingly important for maintaining grid reliability and reducing the need for expensive transmission upgrades or additional conventional generation capacity.
