Hongfei of Littelfuse IXYS Westcode

Power semiconductor devices are the key building blocks in modern power electronics-dominated
power systems. However, under certain conditions, semiconductor devices can fail catastrophically, resulting
in failure and blackout. Temperature Sensi<ve Electrical Parameters (TSEPs) offer a non-invasive route to
monitor device health through measurable electrical signals. This talk will introduce two TSEP fusion strategies,
a mul<-parameter PCA-regression model and a gate current waveform feature-based method, developed for
automo<ve-grade IGBT and SiC MOSFET modules. The talk will also explain how electrical characterization is
conducted in industrial PressPack device engineering, covering output characterization measurements,
evaluation methodologies, and test platform considerations. Taken together, these two perspectives reflect
how the same electrical signals can be interpreted through fundamentally different aspects, one driven by
how a device degrades, the other by how a device is expected to perform.

Biography:

Dr Hongfei Chen (Member, IEEE) received the BSc degree in Electrical Engineering and Automation from Three Gorges University, Yichang, China, in 2016, the MEng degree in Electrical and Electronic Engineering (Distinguished) from the University of Leicester, Leicester, U.K., in 2018, and the PhD degree from the University of Leicester in 2025. She is currently a Product Engineer at Littelfuse IXYS Westcode, Chippenham, U.K. Dr Chen is specialised in power semiconductor device process and design optimisation, electrical-thermal characterisation, and performance validation. She has been leading the evaluation process of high-power PressPack in HVDC transmission, traction, data centre UPS, and industrial inverter applications. Her research interests include advanced semiconductor process qualification, temperature-sensitive electrical parameter evaluation, and condition monitoring.

Address:United Kingdom

Mostefa of University of Leicester

Data-driven control is emerging as a paradigm shift in power electronics, as it offers an effective alternative to classical model-based techniques that rely on accurate knowledge of system structure and parameters. The presentation explores the shift from reliance on detailed physical models to purely data-driven, model-free control in power conversion systems’ control design. Using grid-connected inverters as a case study, the presentation introduces a practical framework in which system behavior is captured in real time using a generic autoregressive with exogenous input (ARX) model identified via recursive least squares (RLS), without requiring any prior knowledge of the system topology or parameters.

The presentation begins by revisiting the limitations of classical predictive deadbeat controllers under parameter uncertainty. We then illustrate how online identification enables predictive control laws to be derived directly from data, thus transforming deadbeat control into an effective measurement-driven algorithm. An adaptive predictive controller with online parameter estimation is introduced to illustrate how real-time identification can compensate for these mismatches and restore prediction accuracy and closed-loop stability.

Building on this adaptive concept, the framework is extended to a fully model-free deadbeat predictive current control (MF-DBPC) strategy, which requires no prior knowledge of system parameters or physical structure. We then present our proposed unified model-free predictive control framework for grid-connected power conversion systems. The central idea is to replace physics-based modeling with a generic ARX model, identified online directly from measured voltage and current data using an RLS estimator.

The key concepts, including control theory, system identification, and real-time implementation, are presented in a tutorial manner. Experimental verification of the proposed control approach is also presented to emphasize its practical feasibility and to highlight its potential as a scalable and implementation-friendly solution for next-generation power conversion systems.

Biography:

Dr Mostefa Kermadi is a Lecturer in the School of Engineering at the University of Leicester. His research focuses on the control of power electronic converters, data-driven and model-free control, and the application of artificial intelligence for power conversion and energy management in renewable energy systems. His recent work emphasizes data-driven control strategies for power conversion, particularly for robust grid integration of renewables under uncertain operating conditions. Mostefa is an IEEE Senior Member and was listed by Stanford University among the world’s top 2% scientists in Electrical Engineering and Energy fields in 2023, 2024, and 2025.

Address:United Kingdom