
ABSTRACT As core components of energy conversion, power semiconductor devices are confronted with reliability challenges. The junction temperature is an important parameter that characterizes the operating status and reliability of power devices, the extraction of which is of great significance for the long‐term safety of the converter. Since the insulated gate bipolar transistors (IGBT) and power metal‐oxide‐semiconductor field‐effect transistors (MOSFET) are widely used in high‐power applications as fully controllable devices, this article focuses on conducting a review of junction temperature extraction methods for these devices, including classification and evaluation. First, four types of junction temperature extraction methods are compared, including optical methods, physical contact methods, analytical calculation methods, and temperature‐sensitive electrical parameter (TSEP) methods. Then, considering the advantages of TSEP methods, the TSEP methods proposed in recent years have been analyzed, which have been divided into steady‐state methods and transient‐state methods based on the working state of the power device. Finally, a comprehensive comparison of TSEP methods has been evaluated in terms of linearity, sensitivity, and other features. Furthermore, gallium nitride (GaN) devices, as another widely used wide‐bandgap power device, operate differently from IGBTs and power MOSFETs, and reliability research on GaN devices is still in the exploratory stage. This article also discusses the junction temperature methods for GaN devices to promote their reliability application in high power‐density converters.
The full-bridge submodule-based modular multilevel converter (MMC) can achieve three-phase 50 Hz to single-phase 16.7 Hz direct AC/AC conversion, making it suitable as railway power supplies. However, the presence of 16.7 Hz complicates the coupling between the single-phase and three-phase sides, causing odd-frequency circulating current components, which influence power quality and increase power losses. To address this issue, this paper proposes hybrid circulating current suppression strategies combining repetitive control and feedforward compensation. First, the improved odd-harmonic repetitive control (ORC) and improved 4k +/- 1-order harmonic repetitive control (4k +/- 1-RC) are proposed to suppress the circulating currents. Each consists of a proportional (P) controller connected in parallel with the repetitive controller, followed by a proportional-integral (PI) controller to accelerate the dynamic response. Second, to further improve the dynamic response of strategies, a common-mode (CM) insertion indices compensation method is introduced as a feedforward component for the strategies. Finally, comparative simulation and HIL experiments conducted under identical parameter settings demonstrate that both proposed CCSCs achieve satisfactory circulating current suppression performance. The 4k +/- 1-RC-based CCSC exhibits a smoother transient response, whereas the ORC-based CCSC provides better harmonic suppression.
Input-series multi-inverter architectures divide the DC bus voltage among modules, which suits high-voltage wireless power transfer but makes module input-voltage balance sensitive to component tolerances and coupling nonuniformity. In multi-transmitter LCC-S topologies this sensitivity has a structural origin: the shared-receiver admittance matrix decomposes into a dissipative diagonal term and a dominant rank-one coupling term whose condition-number lower bound quantifies the fragility of uniform open-loop sharing. A design-orientated equivalent model complements this structural finding by tracing the mismatch to branch impedance dispersion and reflected cross-coupling. The paper develops a hybrid equalisation method in which divider capacitors suppress switching-cycle voltage oscillation and phase-shift modulation compensates the residual RMS imbalance. Switching-level simulation shows that a supervisory PI phase-shift loop cuts the module-voltage span by 95.7% compared with uncompensated open-loop operation. On a 320 V four-inverter prototype, the RMS voltage deviation rate falls from 13.6% with passive balancing alone to 6.1% under hybrid equalisation at 80.7% measured DC-to-load efficiency.
To meet the stringent requirements for high bandwidth and low error in current sensors used for condition monitoring of power converters, embedded printed circuit board (PCB) Rogowski coil sensors have become a promising solution. However, existing PCB Rogowski coil sensors exhibit inherent droop and offset errors when monitoring the switching currents of power converters. To address this issue, this paper proposes a droop and offset error suppression method for a PCB self-integrating Rogowski coil sensor. By optimising the parameters of the self-integrating integrator circuit, the proposed method mitigates waveform distortion in silicon carbide (SiC) MOSFET pulse-current monitoring with conventional sensors. Experimental results show that the method reduces the droop rate of a 50 & micro;s pulse to 3.9%, satisfying the requirement of less than 5%, and limits the offset voltage to within 5 V. The improved measurement accuracy enables abnormal states such as short circuit and overheating to be reflected more accurately, providing accurate current information for reliability-oriented condition monitoring of power converters.
Bond-wire fatigue and gate oxide degradation are the main reliability issues for silicon carbide (SiC) MOSFETs. Accurately separating these two degradation modes and identifying the dominant failure mechanism in operation is crucial for improving device reliability. In this paper, a dynamic diagnosis method for separating bond-wire-fatigue and gate oxide degradation in SiC MOSFETs is proposed. A switching transient analytical model considering bond-wires related parasitic parameters is established first. The analytical analysis reveals that the drain current slope during the turn-on process can comprehensively characterise both bond wires and gate oxide degradation. Furthermore, numerical analysis is conducted to quantitatively investigate the dependence of the drain current slope on bond-wire fatigue and gate oxide degradation. It is found that the relative change in drain current slope induced by bond-wire fatigue decreases with increasing gate resistance, whereas that caused by gate oxide degradation decreases with increasing gate voltage. Based on this observation, a dynamic diagnostic method is proposed to separate the two degradation modes by obtaining the drain current slope under different driving parameters. Finally, equivalent degradation experiments are carried out to validate the proposed dynamic diagnosis method. This work provides a foundation for reliability monitoring and control of SiC MOSFETs in actual switching operation.