Power electronics is the backbone of future energy systems including data centers, electric vehicles, and grid-scale energy storage. These are high-impact applications, demanding increased efficiency, density, and control bandwidth from power electronics converters. To leverage the advances in semiconductor devices and the scaling laws of passive components, a promising trend is to adopt granular power architecture with magnetics integration for minimized power conversion stress and maximized component utilization.
In pursuit of this vision, this thesis first introduces a systematic approach to merging multiple magnetic components into one with matrix coupling. The benefits of matrix coupling in size reduction, ripple compression, and transient acceleration are quantified. The effectiveness of matrix coupling was verified by an “all-in-one-magnetics” dc-dc converter with 5.6x reduced inductor size and 8.5x faster transient speed. Two distinct architectures are then developed to answer two important questions: (1) how to deliver a massive amount of current to a tiny area (e.g., high-performance microprocessors) with extreme power density (above 724 W/in3 with a height below 6 mm); and (2) how to deliver power to a massive number of modular loads (e.g., hard drives, batteries, solar cells) with extreme energy efficiency (above 99%). The fundamental performance gains and limitations of these solutions are thoroughly evaluated. The matrix coupling theory and the two developed architectures push the boundaries of granular power electronics and pave the way toward high-performance power conversion systems for a wider range of applications.
Adviser: Minjie Chen