Date of Graduation

12-2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Engineering (PhD)

Degree Level

Graduate

Department

Electrical Engineering

Advisor/Mentor

Alan Mantooth

Committee Member

Juan Balda

Second Committee Member

Simon Ang

Third Committee Member

Roger Dougal

Keywords

400 V DC, Cascaded Totem-pole Bridgeless PFC, Data Centers, Large Signal Model, Model Predictive Control, Ultracapacitor

Abstract

The demands of high performance cloud computation and internet services have increased in recent decades. These demands have driven the expansion of existing data centers and the construction of new data centers. The high costs of data center downtime are pushing designers to provide high reliability power supplies. Thus, there are significant research questions and challenges to design efficient and environmentally friendly data centers with address increasing energy prices and distributed energy developments.

This dissertation work aims to study and investigate the suitable technologies of power interface and system level configuration for high efficiency and reliable data centers.

A 400 V DC-powered data center integrated with solar power and hybrid energy storage is proposed to reduce the power loss and cable cost in data centers. A cascaded totem-pole bridgeless PFC converter to convert grid ac voltage to the 400 V dc voltage is proposed in this work. Three main control strategies are developed for the power converters. First, a model predictive control is developed for the cascaded totem-pole bridgeless PFC converter. This control provides stable transient performance and high power efficiency. Second, a power loss model based dual-phase-shift control is applied for the efficiency improvement of dual-active bridge converter. Third, an optimized maximum power point tracking (MPPT) control for solar power and a hybrid energy storage unit (HESU) control are given in this research work. The HESU consists of battery and ultracapacitor packs. The ultracapacitor can improve the battery lifetime and reduce any transients affecting grid side operation.

The large signal model of a typical solar power integrated datacenter is built to analyze the system stability with various conditions. The MATLAB/Simulink™-based simulations are used to identify the stable region of the data center power supply. This can help to analyze the sensitivity of the circuit parameters, which include the cable inductance, resistance, and dc bus capacitance. This work analyzes the system dynamic response under different operating conditions to determine the stability of the dc bus voltage. The system stability under different percentages of solar power and hybrid energy storage integrated in the data center are also investigated.

Share

COinS