Talk Title: Data Driven Modelling in Embedded Power Electronics under Real Time and Hardware Constraints

Niloofar Jafarnia Embedded Development Engineer, embeX GmbH, and Researcher at TU Dortmund University

Abstract:

The talk addresses the specific challenges that arise when extending conventional microcontroller-based power electronic platforms with data driven components. It traces the workflow from precise problem formulation and targeted data acquisition through the construction of models, focusing on classical machine learning techniques and compact artificial neural networks motivated by universal approximation theory and constrained by physical system characteristics, to their deployment on embedded targets under stringent computational and real time requirements, with particular emphasis on maintaining feasibility on existing industrial hardware.


Brief Bio:

Niloofar Jafarnia is an Embedded Development Engineer in the Energy & Drive Department at embeX GmbH in Germany and a doctoral researcher at the Chair of Energy Conversion at TU Dortmund University. She received her M.Sc. degree in Embedded Systems Engineering from the University of Duisburg-Essen and her B.Sc. degree in Electrical Engineering. Her research interests include data driven modelling and machine learning for embedded power electronic systems, with a focus on resource constrained implementation and the interaction between physical system behaviour and algorithmic design.