The exponential growth of Artificial Intelligence (AI) applications is driving an unprecedented increase in the number, size, and energy demand of data centers worldwide. This trend raises urgent challenges in terms of efficiency and sustainability, making the design of innovative power conversion systems a strategic priority. This thesis investigates the Integrated Active Filter Rectifier (IAFR), the first part of the quasi-single-stage Integrated Active Filter isolated Rectifier (IAF-iR) AC/DC converter. The topology under investigation combines rectification and power factor correction in an efficient and compact solution. The work covers the theoretical modeling of the IAFR, its validation through PLECS simulations and Hardware-In-the-Loop (HIL) testing, and the detailed design of the magnetic component required by the IAFR, namely the current-injection inductor. Special attention is posed to practical constraints such as volume minimization, thermal dissipation, and efficiency targets. A prototype inductor is real- ized using Litz wires in an optimized winding distribution, and its electrical parameters are experimentally validated. The results confirm the capability of the IAFR to achieve high efficiency (up to 99.7%), sinusoidal input currents, and compact magnetic design, making the IAFR a promising candidate for future high-performance power supply units in AI data centers. Furthermore, the study highlights the potential advantages of emerging semiconductor technologies, such as Gallium Nitride (GaN) based Monolithic Bidirectional Devices (MBDs), which are expected to further enhance efficiency and power density in future implementations.
La crescita esponenziale delle applicazioni di AI sta determinando un incremento senza precedenti nel numero, nelle dimensioni e nei consumi energetici dei data center a livello mondiale. Questa tendenza solleva sfide urgenti in termini di efficienza e sostenibilità, rendendo la progettazione di sistemi innovativi di conversione dell’energia una priorità strategica. Questa tesi analizza l’IAFR, la prima parte del convertitore AC/DC quasi–single–stage IAF-iR. La topologia in esame combina rettificazione e correzione del fattore di potenza in una soluzione compatta ed efficiente. Il lavoro affronta la modellazione teorica dell’IAFR, la sua validazione tramite simulazioni in PLECS e test HIL, nonché la progettazione dettagliata del componente magnetico richiesto dall’IAFR, ovvero l’induttore di iniezione di corrente. Particolare attenzione è rivolta ai vincoli pratici quali la minimizzazione del volume, la dissipazione termica e gli obiettivi di efficienza. È stato realizzato un prototipo di induttore utilizzando filo Litz con una distribuzione degli avvolgimenti ottimizzata, e i suoi parametri elettrici sono stati validati sperimentalmente. I risultati confermano la capacità dell’IAFR di raggiungere un’elevata efficienza (fino al 99.7%), correnti di ingresso sinusoidali e un design magnetico compatto, rendendolo un candidato promettente per le future unità di alimentazione ad alte prestazioni nei data center dedicati all’AI. Inoltre, lo studio evidenzia i potenziali vantaggi delle tecnologie a semiconduttore emergenti, come i MBDs basati su GaN, che si prevede possano incrementare ulteriormente l’efficienza e la densità di potenza nelle implementazioni future.
Analysis and Design of a High-Efficiency Integrated Active Filter Rectifier
GARBUGGIO BIGAGLIA, GIACOMO
2024/2025
Abstract
The exponential growth of Artificial Intelligence (AI) applications is driving an unprecedented increase in the number, size, and energy demand of data centers worldwide. This trend raises urgent challenges in terms of efficiency and sustainability, making the design of innovative power conversion systems a strategic priority. This thesis investigates the Integrated Active Filter Rectifier (IAFR), the first part of the quasi-single-stage Integrated Active Filter isolated Rectifier (IAF-iR) AC/DC converter. The topology under investigation combines rectification and power factor correction in an efficient and compact solution. The work covers the theoretical modeling of the IAFR, its validation through PLECS simulations and Hardware-In-the-Loop (HIL) testing, and the detailed design of the magnetic component required by the IAFR, namely the current-injection inductor. Special attention is posed to practical constraints such as volume minimization, thermal dissipation, and efficiency targets. A prototype inductor is real- ized using Litz wires in an optimized winding distribution, and its electrical parameters are experimentally validated. The results confirm the capability of the IAFR to achieve high efficiency (up to 99.7%), sinusoidal input currents, and compact magnetic design, making the IAFR a promising candidate for future high-performance power supply units in AI data centers. Furthermore, the study highlights the potential advantages of emerging semiconductor technologies, such as Gallium Nitride (GaN) based Monolithic Bidirectional Devices (MBDs), which are expected to further enhance efficiency and power density in future implementations.| File | Dimensione | Formato | |
|---|---|---|---|
|
GarbuggioBigaglia_Giacomo.pdf
Accesso riservato
Dimensione
13.59 MB
Formato
Adobe PDF
|
13.59 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/93732