Year
2023
Season
Spring
Paper Type
Master's Thesis
College
College of Computing, Engineering & Construction
Degree Name
Master of Science in Electrical Engineering (MSEE)
Department
Engineering
NACO controlled Corporate Body
University of North Florida. School of Engineering
First Advisor
Dr. Touria El Mezyani
Second Advisor
Dr. Mona Nasseri
Abstract
This thesis covers the increasing usage of machine learning techniques in the field of power electronics. These uses are broad and efficient, but not without flaws. The major flaw being the requirement of time to collect data if there is not pre-existing datasets available. To combat this, a hybrid combination of existing physics-based modeling and data-driven modeling is proposed to limit the high computational strain of physics-models while simultaneously preventing the need for large pre-existing datasets. This method is then simulated and tested on a buck-converter with several disturbances applied to it to stimulate non-linear dynamics known as bifurcations that cannot be easily predicted by physics-based methods. The results show that hybrid methodology is more efficient and robust than data-driven methods alone and can efficiently predict the output of a system even when under the effect of bifurcations.
Suggested Citation
Harrison, Caleb, "A hybrid prediction approach using machine learning and physics based models applied to power electronic circuits" (2023). UNF Graduate Theses and Dissertations. 1177.
https://digitalcommons.unf.edu/etd/1177
Included in
Controls and Control Theory Commons, Power and Energy Commons, Signal Processing Commons