The GOOGLE Little Box Challenge – Ultra-Compact GaN-Based Single-Phase DC/AC Power Conversion

Abstract

With the ambition of achieving a cost reduction in solar power and microgrid technology, an increased efficiency of Uninterruptible Power Supplies (UPS) or the ability to use an electric vehicle’s battery as backup power during a power outage, Google and IEEE initiated the Google Little Box Challenge (GLBC) back in July 2014 to build the worldwide smallest 2 kW / 450 V DC / 240 V AC single-phase PV inverter with η > 95 % CEC weighted efficiency and an air-cooled case temperature of less than 60 °C by using latest power semiconductor technology and innovative converter concepts, advertising $1 million prize money. The challenging specifications and the attractive prize money created a remarkable interest in the power electronics community, which led to the participation of 2000+ teams – companies, research institutes and universities – in the GLBC. Finally, 100+ teams submitted technical descriptions of realized systems. Out of these applications, 18 finalists including ETH Zurich were invited to submit their hardware prototypes for final testing. This dissertation reports all major findings and key lessons learned from the participation of the Power Electronic Systems Laboratory (PES) of ETH Zurich in the GLBC and during the research conducted afterwards to investigate encountered problems which could not be analyzed in detail because of the tight schedule of the competition. The power-density benchmark established by the realized inverter prototypes and considering the achievements of other GLBC finalists, indicates that a 20 times higher power-density compared to the current state-of-the-art in industry is principally possible. All necessary aspects to realize an extreme power-density converter in accordance with the GLBC specifications are discussed. First, a review of suitable converter topologies and advanced control concepts and component technologies to achieve a high power-density, adopted by the GLBC finalists and/or described in the scientific literature, is provided. Different bridge-leg control techniques (e.g. constant frequency PWM vs. Triangular Current Mode (TCM)), the selection of WBG power transistors, the implementation of compact High Frequency (HF) inductors and the selection of suitable capacitors, are discussed among other relevant topics. Guided by the insights from a preceding multi-objective design optimization (virtual prototyping), two hardware implementations of the Little Box inverter concepts developed at ETH Zurich are presented and accompanied with experimental results to support the claimed performances with respect to efficiency, η, and power-density, ρ. The initial prototype implementation, a GaN based full-bridge inverter with TCM control and variable switching frequency up to the MHz range, was ranked among the top ten contributions of the GLBC finalists. The continued research following the conclusion of the GLBC in October 2015, resulted in an improved understanding of key technologies and allowed to further improve component models for more accurate Pareto optimization results. Novel experimental methods to accurately determine the soft-switching losses in GaN and SiC semiconductors, to characterize the behavior of ceramic capacitors subject to a large-signal excitation at low-frequencies and to analyze unexpectedly high core losses in multi-airgap MnZn ferrite inductors, developed for the first version of the Little Box inverter, are reported. By means of the gained insights and the consideration of an alternative inverter concept, i.e. a DC/|AC|-buck converter operated with constant 140 kHz PWM and a subsequent low-frequency unfolding inverter, an improved version of the Little Box inverter, almost twice as compact as the initial version, is realized. Because of the fluctuating power with twice the mains-frequency intrinsic to single-phase AC systems, active power buffer concepts with additional auxiliary converters are employed in the developed Little Box inverter systems in order to substitute bulky electrolytic capacitors and shrink the volume of the necessary energy storage element to compensate the fluctuating power. In particular, two promising concepts, a full-power processing buck-type and a partial-power processing Series Voltage Compensator (SVC)-type power buffer, are comparatively evaluated. The investigated partial-power auxiliary converter approach is subsequently also applied to a 380 V DC / 48 V DC series-resonant LLC converter, where a novel control concept achieves tight output voltage regulation for changing input voltage without varying the switching frequency or duty-cycle of the main DC/DC converter. Since the auxiliary converter only processes a small share of the rated power, only a marginal ηρ-impairment of the overall converter must be accepted compared to a system with constant voltage transfer ratio. In conclusion, the main research findings and lessons learned over the course of the dissertation are summarized and an outlook on expected future power-density improvements is provided including a discussion of necessary advances of the involved component technologies. Finally, further insights on the origin of the observed excess cores losses in multi-airgap inductors and a comprehensive comparison of two promising ceramic capacitor technologies to implement ultra-compact power pulsation buffers are provided in the Appendices.