The rapid expansion of artificial intelligence infrastructure is creating new challenges for power grid stability and generator reliability. A new technical analysis reveals that large-scale data center loads can generate persistent sub-synchronous power oscillations that directly threaten the mechanical integrity of nearby hydroelectric generators.
The research develops a comprehensive risk assessment framework using electromagnetic transient simulation to model how oscillatory loads propagate through the transmission network to hydro-generator terminals. The study employs a two-mass turbine-generator shaft model representing real-world generation units, allowing engineers to predict where resonance conditions occur and how severely shaft stress increases.
Key findings show that generator design significantly influences vulnerability. Kaplan-type hydro units—commonly used for variable-flow applications—appear more susceptible to fatigue than comparable Francis or Pelton designs due to lower inertia ratios that shift torsional modes to frequencies where data center oscillations operate. The analysis also demonstrates that reduced damping in aging generators substantially amplifies resonant response, compounding fatigue exposure.
The framework quantifies risk through a two-stage transfer function approach: first measuring how data center oscillations travel through the network to the generator, then calculating resulting torque amplification at the turbine shaft. A frequency-scan methodology identifies dangerous resonance regions, while classical Goodman diagram fatigue analysis translates simulated torque into practical safety factors.
As data centers become larger and more prevalent in regional grids, interconnection planners and utility operators face new technical constraints. The study provides actionable guidance for implementing oscillation limits during interconnection studies, establishing plant-level vibration monitoring systems, and protecting valuable hydro assets from accelerated aging. These findings bridge a critical gap between rapid data center growth and grid infrastructure resilience, ensuring that power system planning accounts for emerging load characteristics that differ fundamentally from traditional industrial and residential demand patterns.
