Abstract:
Data centers are experiencing a surge in power consumption due to the increasing complexity of artificial intelligence (AI) models. The fluctuating power demand associated with AI training workloads can introduce sub-synchronous harmonics (SSHs), potentially destabilizing the utility grid. Power hardware-in-the-loop (PHIL) simulation offers a promising approach to study these harmonics and assess grid compliance. However, PHIL simulation itself might cause instability and inaccuracies, mainly due to PHIL interface where software and hardware are connected. This paper evaluates various PHIL interfaces in the data center application to study SSHs and fault scenarios . A real-time simulation of a data center is conducted using the RTDS platform, and the grid power spectrum is analyzed via Fast Fourier Transform (FFT). Based on the results, recommendations are provided for selecting the most appropriate PHIL interface for data center applications.
J. Dadkhah, Y. Qi and Y. Zhang, "PHIL Interface Evaluation in Data Center Application," 2026 IEEE Power & Energy Society General Meeting (PESGM), Montreal, QC, Canada, 2026, pp. 1-5.
KEYWORDS: Data center, PHIL, Artificial Intelligence, real-time simulation