Dynamic Modelling of Large Inverter-Based Loads: A Data Centre Perspective

These systems offer high availability but introduce fast, nonlinear behaviours that are not well represented by conventional ZIP or composite load models (CMLDs). Reported operating experience in several jurisdictions indicates that simplified representations can miss important vulnerabilities, motivating higher-fidelity user-written models in both electromagnetic transient

(EMT) and phasor-domain transient (PDT) environments, with appropriate treatment of control and protection logic.

This work summarises key considerations for representing data centres in dynamic studies when vendor-specific models are unavailable. Data centres, commonly equipped with doubleconversion uninterruptible power supply (UPS) systems, tend to maintain near-constant power during disturbances, increasing current demand under low voltage subject to current limits and protection. Their converter controls and backup arrangements can produce responses that differ from conventional loads and inverter-based resources (IBRs), including sensitivity to voltage and frequency disturbances, phase-angle jumps, and interactions with weak grids and nearby

IBRs.

A modular, component-based modelling framework is proposed, informed by representative performance expectations such as fault ride-through (FRT), response under low system strength, and behaviour under unbalanced conditions. The framework focuses on observable outcomes at the connection point and avoids prescriptive definitions of internal design choices.

Key behaviours to capture include FRT, response to unbalanced and phase-jump conditions,

DC voltage and current stability, reactive power dynamics, and post-disturbance recovery trajectories, including conditional transfer to internal supply and staged reconnection. This modular approach provides a practical basis for representing data centres in dynamic studies and for selecting an appropriate level of model detail without relying on overly simplified assumptions.