PhD defence by Yingzhou Peng on Advanced Converter-level Condition Monitoring for Power Electronics Components
09.12.2020 kl. 09.00 - 12.00
Yingzhou Peng, Department of Energy Technology, will defend the thesis "Advanced Converter-level Condition Monitoring for Power Electronics Components"
Advanced Converter-level Condition Monitoring for Power Electronics Components
Professor Huai Wang
Professor Francesco Iannuzzo
Associate Professor Amjad Anvari-Moghaddam, Dept. of Energy Technology, Aalborg University (Chairman)
Associate Professor Sanjib Kumar Panda, National University of Singapore, Singapore
Senior Lecturer Suresh Perinpanayagam, Cranfield University, UK
Applications, such as traction and offshore wind, demand cost-effective and robust condition monitoring solutions for operation optimization and predictive maintenance. There is still a gap between academic research and industrial applications in condition and health monitoring in power electronic converters from the signal extraction, calibration, and data processing perspectives. This Ph.D. study proposes two converter-level condition monitoring methods that have promising features to bridge the gap.
The first method is based on a converter-level on-state voltage measurement
approach. The on-state voltage, such as saturation voltage of IGBTs and forward voltage of diodes, is a widely used electrical parameter for junction and health condition estimation of power semiconductor switches. Conventional methods are mainly at the component-level by adding a measurement circuit for each switch in its gate driver circuit. It suffers from two challenges: 1) gate driver is not always accessible, especially for power converters with design freeze, and 2) it has a relatively high cost and complexity.
This study proposes a single measurement circuit connected in the middlepoints
of the phase legs of a single-phase inverter or three-phase inverter. It can measure on-state voltages of all the power switches, including diodes. Three variants of the circuits have been proposed as well: 1) with external power supply as existing component-level methods do; 2) with a self-power scheme to exclude the power supply which is the most expensive component in the measurement circuit, and 3) simple passive solution without the need of both power supply and self-power circuit. The passive solution reduces the implementation cost to 10% of the widely used component-level on-state voltage measurement circuit for a three-phase inverter application. It provides a plug-and-play solution with better accessibility to both existing and new designs of power electronic converters. Based on the obtained on-state voltage information, the junction temperature and health status estimation are demonstrated on the case study of single-phase inverter.
The second method is based on the digital twin concept without any additional hardware circuitry. This study chooses a Buck converter as the case study to prove the concept by using its digital twin for the degradation status monitoring of MOSFETs and capacitors. The analytical models of the power stage and controllers are built up for the Buck converter. The in-situ component parameters, such as the on-state voltage of MOSFETs, capacitance, and equivalent series resistance, are estimated based on the existing feedback signals from the physical prototype and the corresponding calculated values from the digital twin. A data cluster method is proposed, which eliminates the calibration requirement for excluding the impact of other factors on the estimated component parameters, such as load level and temperature. The proposed method enables the degradation monitoring of MOSFETs and capacitors in the Buck converter without calibration and additional sensor. The application of the concept has been extended to a single-phase inverter.
The above methods have been experimentally verified, besides the theoretical
analyses. An industry-oriented prototype has been designed for the converter-level on-state voltage measurement. The results of the Ph.D. research have been presented in 1 patent, 5 journal papers, and a few other conference publications.
THE DEFENCE IN ENGLISH - all are welcome.
Streaming info TBD.
Department of Energy Technology