New dynamic phasor framework tackles grid oscillations from AI data centers

Subsynchronous oscillations (SSOs) pose a growing risk to grid stability as inverter-based resources (IBRs) and massive loads like AI data centers proliferate. Traditional electromagnetic transient (EMT) simulations can capture SSOs but become computationally prohibitive for large systems. In a new arXiv paper, Fiaz Hossain and colleagues from Penn State and Mitsubishi Electric introduce a generalized dynamic phasor (DP) framework that models grid-following (GFL) and grid-forming (GFM) IBRs in the dq-frame, while representing synchronous generators, transmission networks, and loads in the pnz-frame. The framework is linear and time-invariant, enabling efficient eigen-analysis to identify and mitigate poorly damped SSO modes.

The authors validate their approach using PSCAD/EMTDC on the IEEE first benchmark for subsynchronous resonance and a modified 4-machine system, then demonstrate scalability on the IEEE 68-bus system with two GFL IBRs. They propose two damping solutions: a decentralized controller optimized via particle swarm optimization (PSO), and replacing one GFL IBR with GFM control. Critically, the study also examines the impact of AI data center loads on primary frequency response and multi-mass turbine dynamics, highlighting a new interaction between computing infrastructure and power system stability.