PRESENTATION
AI-based Controllers for Power Electronics Converters
This presentation explores the development and implementation of AI-based control strategies for power electronics converters, focusing on reinforcement learning (RL) controllers for the 5-level Packed U-Cell (PUC5) grid-connected inverter. Traditional model-based controllers, such as Model Predictive Control (MPC) and Sliding Mode Control (SMC), achieve precise current regulation but are heavily dependent on system parameters. In contrast, reinforcement learning provides a model-free approach capable of handling nonlinearities and uncertainties. The proposed RL-based controller leverages a Proportional-Integral (PI) reward function, streamlining the training process by decoupling voltage balancing from the RL control objective. Instead, voltage balancing is managed via redundant switching states, significantly reducing the action space and training complexity. This study highlights the potential of RL as a robust and adaptive control solution for multilevel inverters, offering a promising alternative to conventional control methods.
Hani Vahedi
Delft University of Technology
Hani Vahedi (Senior Member, IEEE) received his PhD with honors from École de Technologie Superieure (ÉTS), University of Quebec, Montreal, Canada, in 2016. He received the Best PhD Thesis Award from ETS for the academic year of 2016-2017. He has published more than 90 technical papers in IEEE conferences and Transactions. He also published a book on Springer Nature and a book chapter in Elsevier. He was a co-chair of the IEEE Industrial Electronics Society (IES) Student & Young Professionals (S&YP) committee and is currently serving as the IES Chapters Coordinator. He has been co-organizing special sessions and SYP forums at IEEE international conferences. He is also the associate editor for IEEE Transactions on Industrial Electronics, IEEE Open Journal of Industrial Electronics Society, and IEEE Open Journal of Power Electronics. He is the inventor of the PUC5 converter, holds multiple US/world patents, and transferred that technology to the industry, where he developed the first bidirectional electric vehicle DC charger based on his invention. After 7 years of experience in industry as a power electronics designer and chief scientific officer, he joined the Delft University of Technology, where he is currently an assistant professor at the DCE&S group, working towards the electrification of industrial processes for clean energy transition. He is also leading the 24/7 Energy Hub project at The Green Village of TU Delft, implementing a local microgrid with renewable energy resources, green Hydrogen production, and energy storage systems as the future of the clean energy transition. His research interests include multilevel converter topologies, control and modulation techniques, and their applications in the electrification of industrial processes and clean energy transition, such as smart grids, renewable energy conversion, electric vehicle chargers, green hydrogen production (electrolyzers), and fuel-cell systems.