home

Development of a Converter-Based Emulator for Data Center Power Distribution System

Application

Data centers have been constructed around the world, leading to a considerable power consumption. It is reported that in 2014, 70 TWh energy is consumed by data centers in the U.S., accounting for 1.8% of the national electricity supply [1]. Influence of the fast-growing data centers on the power system is nonnegligible. The huge power consumption with fluctuations could lead to undesired consequences in power grid stability. Also, as an electric load with a high-density of power electronics, data center has multiple operation modes which result in dynamic response on power grid. In this project, a converter-based emulator of the typical data center AC power distribution system is developed to investigate the dynamic performances of the power network. Transient conditions are identified, and dynamic interactions between power grid and data centers are characterized. An adaptive linearized average model is proposed and implemented as an emulator in hardware testbed (HTB) to predict the grid dynamic performances in different transient conditions.

Research

Illustration of the structure, modeling and emulation of the data center system.

The project aims to develop a converter-based emulator for data center power distribution system. To achieve this, approaches in several aspects are conducted:

  1. Data center survey: commonly applied AC power distribution data centers are surveyed comprehensively to characterize the system architecture, operational principle, power electronics topologies, and protection schemes.
  2. System simulation: simulation model of the data center power distribution system is built on MATLAB Simulink, mainly including the electrical system and the cooling system. The electrical system is composed of a centralized uninterrupted power supply (UPS), power distribution unit (PDU), rack-level power supply unit (PSU), server motherboard, and IT loads [2]. And the cooling system consists of cooling tower, chiller, cooling water pumps, computer room air handler (CRAH), server fans [3]. Power converters and the associated controls are used in the model.
  3. Adaptive linearized average model: average model of each component is derived and linearized by discrete-time equations with 10 kHz sample frequency. Also, an adaptive model is derived to predict dynamic performances in different conditions.
  4. Implementation on HTB: The linearized model of the data center is implemented in one of the voltage source inverters (VSI) in the HTB, which is a multi-converter based hardware testbed platform developed to perform real power testing and emulate the transmission-level power grid [4]. By programming the VSI, data center is emulated as one of the loads within the power system, to predict the dynamic performance of the grid network.
  5. Emulation of transient conditions: dynamic transitions of the data center load mainly happens during the voltage sags when UPS switches the operation mode, backup energy switch online/offline, motor dynamic response, etc. Also, the data center backup energy can be used to provide some grid ancillary services like voltage support and frequency regulation, which induces more interactions between power grid and data centers.

How WBG Can Help

With the converter-based real-time emulator, various transient responses or interactions between the grid and data centers can be demonstrated accurately with real power, which helps fulfill the evaluation of the dynamic performances in power systems. Also, the emulator can be used to investigate grid support functions of data centers like voltage support and frequency regulation.

Personnel Involved

Students
  • Jingjing Sun
Faculty

References

[1] P. T. Krein, "Data center challenges and their power electronics," CPSS Transactions on Power Electronics and Applications, vol. 2, no. 1, pp. 39-46, 2017.
[2] ANSI/BICSI-002, Data center design and implementation best practices, 2014.
[3] Joshi, Yogendra, and Pramod Kumar, eds. Energy efficient thermal management of data centers. Springer Science & Business Media, 2012.
[4] J. Wang, Y. Song, W. Li, J. Guo, and A. Monti, "Development of a universal platform for hardware in-the-loop testing of microgrids," IEEE Transactions on Industrial Informatics, vol. 10, no. 4, pp. 2154-2165, 2014.