betaFeedback

WorksEngineering

Model-Free UPS Control Tuning for Improved Low-Voltage Ride-Through in AI Data Centers

Abstract

Voltage sags can trigger protection trips in uninterruptible power supply (UPS) systems, abruptly reducing the power that AI data centers draw from the grid. This paper improves the low-voltage ride-through (LVRT) capability of AI data centers equipped with double-conversion UPS systems by tuning their existing converter controls. We formulate a bound-constrained optimization problem to maximize the ride-through area, defined as the integral of ride-through time over a range of retained grid voltages. To solve this problem, we develop a batched projected zeroth-order optimization algorithm with adaptive moment estimation (BPZO-Adam). The algorithm evaluates candidate control settings using only simulated ride-through times, making it suitable for vendor-supplied black-box models. Simulations using the TAMU-ERCOT electromagnetic transient (EMT) AI data center model show that tuning converter control parameters increases the ride-through area by 10.6%. In a separate study, tuning the controls yields a larger ride-through area than increasing the DC-link capacitance tenfold.

Paper details

vlnc:W_twydudge
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC-BY-NC-ND-4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Engineering

Not recorded for this paper

View statement
Looking for related papers…