Multi-Junction Photonic Power Converters: AI Enhanced Design Optimization

Photovoltaic devices containing InGaAs absorbers, lattice matched to InP, have shown excellent performance in many applications. We explore this material' potential in photonic power converter (PPC) applications. We have developed a machine learning empowered computational framework to explore...

Full description

Saved in:
Bibliographic Details
Main Authors: Hinzer, Karin, Hunter, Robert F.H., Wilson, D. Paige, Forcade, Gavin P., Beattie, Meghan N., Valdivia, Christopher E., Hahn, Oliver, St-Arnaud, Louis-Philippe, Pellegrino, Carmine, Lackner, David, Grinberg, Yuri, Krich, Jacob J., Walker, Alexandre W., Helmers, Henning
Format: Conference Proceeding
Language:English
Subjects:
Dimensionality reduction
Indium phosphide
Junctions
Lattices
Machine learning algorithms
Optoelectronic devices
Performance evaluation
Photonics
Photovoltaic systems
Space exploration
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Photovoltaic devices containing InGaAs absorbers, lattice matched to InP, have shown excellent performance in many applications. We explore this material' potential in photonic power converter (PPC) applications. We have developed a machine learning empowered computational framework to explore design space for optoelectronic devices. We apply dimensionality reduction and clustering machine learning algorithms to identify optimal ten-junction C-band PPC designs. We introduce a photo current figure of merit to find optimal designs with low computational cost. We compare modeled performance to experimental devices for 1, 2 and 10 junctions. We discuss the role of luminescent coupling in devices under test, and show increased effect when devices have a back reflector. We compare results obtained using dimensionality reduction with traditional Beer-Lambert calculations.
ISSN:2995-1755
DOI:10.1109/PVSC57443.2024.10749088