Xing Wei
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Xing Wei has a total of 46 co-authors. Here are the top co-authors by number of shared publications:
- Huai Wang
Aalborg University
16 shared publications - Bo Yao
North China Electric Power University
15 shared publications - Yingzhou Peng
School of Electrical Engineering, Hunan University, Hunan, China
9 shared publications - Zhaoxin Wang
Department of Energy Technology, Aalborg University, Aalborg, Denmark
7 shared publications - Yichi Zhang
Aalborg University
6 shared publications - Jiahong Liu
Department of Energy, Aalborg University, Aalborg, Denmark
5 shared publications
Latest Publications
- 1 December 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (12) , 21460-21473
A Thermal Modeling Method for Power Capacitors Under Long Timescale Dynamic Operation Profiles
- 1 December 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (12) , 21460-21473
A Thermal Modeling Method for Power Capacitors Under Long Timescale Dynamic Operation Profiles
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73 (10) , 15396-15408
PE-TSFM: Self-Supervised Time-Series Learning for Generalizable Power Converter Health Monitoring Under Unseen Conditions
Data-driven health monitoring of power converters remains limited by poor generalization to unseen operating conditions. This work addresses this out-of-distribution (OOD) challenge by building a domain-specific time-series foundation model (PE-TSFM) that learns representations directly from large-scale unlabeled converter data. Unlike generic TSFMs trained on broad time-series datasets, the proposed PE-TSFM is pretrained entirely on domain data, enabling it to learn the physical relationships unique to power electronics. To further tailor the model to this domain, we introduce a dual-attention mechanism that captures both temporal patterns and inter-channel dependencies. While generic TSFMs primarily model temporal dependencies, the added channel attention (CA) captures inter-sensor physical relationships essential for converter degradation analysis. A dataset containing 141 million unlabeled timestamps from an operating power converter is used for pretraining. Experiments show that PE-TSFMs achieve 76%–92% accuracy under unseen operating conditions. In contrast, generic TSFMs achieve around 60% and conventional time-series models achieve around 40% accuracy. This result confirms the strong OOD generalization of the proposed PE-TSFM. Ablation studies further verify that the introduced CA mechanism significantly improves model performance. In addition, we conduct detailed studies on model scalability, hyperparameter sensitivity, and interpretability to provide a comprehensive understanding of the proposed approach.
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73 (10) , 15396-15408
PE-TSFM: Self-Supervised Time-Series Learning for Generalizable Power Converter Health Monitoring Under Unseen Conditions
Data-driven health monitoring of power converters remains limited by poor generalization to unseen operating conditions. This work addresses this out-of-distribution (OOD) challenge by building a domain-specific time-series foundation model (PE-TSFM) that learns representations directly from large-scale unlabeled converter data. Unlike generic TSFMs trained on broad time-series datasets, the proposed PE-TSFM is pretrained entirely on domain data, enabling it to learn the physical relationships unique to power electronics. To further tailor the model to this domain, we introduce a dual-attention mechanism that captures both temporal patterns and inter-channel dependencies. While generic TSFMs primarily model temporal dependencies, the added channel attention (CA) captures inter-sensor physical relationships essential for converter degradation analysis. A dataset containing 141 million unlabeled timestamps from an operating power converter is used for pretraining. Experiments show that PE-TSFMs achieve 76%–92% accuracy under unseen operating conditions. In contrast, generic TSFMs achieve around 60% and conventional time-series models achieve around 40% accuracy. This result confirms the strong OOD generalization of the proposed PE-TSFM. Ablation studies further verify that the introduced CA mechanism significantly improves model performance. In addition, we conduct detailed studies on model scalability, hyperparameter sensitivity, and interpretability to provide a comprehensive understanding of the proposed approach.
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73 (10) , 15409-15421
Physics-Informed Neural Network Based Noninvasive LC-Filter Parameter Estimation in Inverters
This article proposes a novel physics-informed neural network (PINN) parameter estimation method that integrates circuit physical models with deep learning for three-phase inverter in uninterruptible power supply (UPS) system. Due to manufacturing tolerances and continuous operation, the AC-side LC filter of inverter in UPS is subject to degradation and parameter drift, which can affect service performance and equipment lifetime. The proposed noninvasive approach avoids additional hardware and demonstrates strong performance trained by small-scale dataset. Pretrained neural networks and transfer learning strategy are employed to enhance training efficiency, model generalization, and noise immunity, enabling reliable and faster convergence. Moreover, a novel trigger-based sampling and data selection strategy is proposed to improve efficiency in parameter identification. An experimental prototype was built for verification, and AC film capacitors are used in the prototype. The experiment results indicate that the estimation error for inductor remains within 2.5% and the error for capacitor is approximately 5% under normal operating conditions.
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73 (10) , 15409-15421
Physics-Informed Neural Network Based Noninvasive LC-Filter Parameter Estimation in Inverters
This article proposes a novel physics-informed neural network (PINN) parameter estimation method that integrates circuit physical models with deep learning for three-phase inverter in uninterruptible power supply (UPS) system. Due to manufacturing tolerances and continuous operation, the AC-side LC filter of inverter in UPS is subject to degradation and parameter drift, which can affect service performance and equipment lifetime. The proposed noninvasive approach avoids additional hardware and demonstrates strong performance trained by small-scale dataset. Pretrained neural networks and transfer learning strategy are employed to enhance training efficiency, model generalization, and noise immunity, enabling reliable and faster convergence. Moreover, a novel trigger-based sampling and data selection strategy is proposed to improve efficiency in parameter identification. An experimental prototype was built for verification, and AC film capacitors are used in the prototype. The experiment results indicate that the estimation error for inductor remains within 2.5% and the error for capacitor is approximately 5% under normal operating conditions.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (7) , 10496-10502
A Flexible Hold-Up Time Application for AC/DC Converters Based on Active Capacitor
This letter proposes a DC-side hold-up-time extension method for AC/DC converters using a low-voltage active capacitor. A low-voltage H-bridge with an auxiliary energy-storage capacitor is inserted in series with the DC-link capacitor and operated in three modes (charging, bypass during normal operation, and hold-up during AC line-cycle dropout) without modifying the original converter topology or control loops. Closed-form hold-up-time analysis and benchmark evaluation are provided to guide the tradeoff between the volume and power loss of the active capacitor. Experiments on a commercial 3.3 kW/400 V server power supply demonstrate a 20.7 ms hold-up time at full load, while reducing capacitor volume by 34.1% compared with a passive solution and reducing measured power loss by 32% compared with a baby-boost solution (13.4 W). Other electrical and thermal performances are also verified by experiments in terms of operating temperature, cascaded stability, transient response to load variation, and response to output voltage variation. Cost and electrolytic-capacitor lifetime are also evaluated, showing comparable prototype cost and $B_{5}$ lifetime to the passive solution.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (7) , 10496-10502
A Flexible Hold-Up Time Application for AC/DC Converters Based on Active Capacitor
This letter proposes a DC-side hold-up-time extension method for AC/DC converters using a low-voltage active capacitor. A low-voltage H-bridge with an auxiliary energy-storage capacitor is inserted in series with the DC-link capacitor and operated in three modes (charging, bypass during normal operation, and hold-up during AC line-cycle dropout) without modifying the original converter topology or control loops. Closed-form hold-up-time analysis and benchmark evaluation are provided to guide the tradeoff between the volume and power loss of the active capacitor. Experiments on a commercial 3.3 kW/400 V server power supply demonstrate a 20.7 ms hold-up time at full load, while reducing capacitor volume by 34.1% compared with a passive solution and reducing measured power loss by 32% compared with a baby-boost solution (13.4 W). Other electrical and thermal performances are also verified by experiments in terms of operating temperature, cascaded stability, transient response to load variation, and response to output voltage variation. Cost and electrolytic-capacitor lifetime are also evaluated, showing comparable prototype cost and $B_{5}$ lifetime to the passive solution.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (7) , 11322-11336
Application-Oriented Power Semiconductor Switches Condition Monitoring Based on Data-Driven Method
This paper proposes an application-oriented condition monitoring method for power semiconductor switches, which constructs a robust health indicator with high robustness, low sampling rate, and low data storage requirements that decouples the effect of operation conditions on the health indicator. First, an effective application-oriented data acquisition and collection strategy is presented based on high-level synchronous-triggered sampling of the switching cycle and data selected from the peak point of the fundamental cycle current, which retains features strongly correlated with the degradation of power semiconductor switches while reducing sampling rate, data transmission and storage requirements. Then, a concept of mission profile alignment method is proposed to decouple the effects of operation conditions on the health indicator. It avoids the extensive offline calibration requirement that is difficult to be applied in the field. Finally, the reconstruction error of the conditional autoencoder network is used as a robust health indicator, which further enhances the immunity of condition monitoring against variations of operation conditions. A case study with the three-phase grid-connected inverter for the PV system is demonstrated to prove the effectiveness of the proposed method.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (7) , 11322-11336
Application-Oriented Power Semiconductor Switches Condition Monitoring Based on Data-Driven Method
This paper proposes an application-oriented condition monitoring method for power semiconductor switches, which constructs a robust health indicator with high robustness, low sampling rate, and low data storage requirements that decouples the effect of operation conditions on the health indicator. First, an effective application-oriented data acquisition and collection strategy is presented based on high-level synchronous-triggered sampling of the switching cycle and data selected from the peak point of the fundamental cycle current, which retains features strongly correlated with the degradation of power semiconductor switches while reducing sampling rate, data transmission and storage requirements. Then, a concept of mission profile alignment method is proposed to decouple the effects of operation conditions on the health indicator. It avoids the extensive offline calibration requirement that is difficult to be applied in the field. Finally, the reconstruction error of the conditional autoencoder network is used as a robust health indicator, which further enhances the immunity of condition monitoring against variations of operation conditions. A case study with the three-phase grid-connected inverter for the PV system is demonstrated to prove the effectiveness of the proposed method.
- 9 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (9) , 16333-16352
Discharge Profile-Based Converter-Level Health Monitoring for Power Semiconductor Module in Traction Inverter Application
This paper proposes a converter-level health monitoring scheme for power semiconductor modules in traction inverter applications, implemented via the standard inverter DC-link discharge operation mode after system shutdown. The on-state voltage information of all power devices in the inverter is acquired in combined forms from the inverter three-phase output terminals, minimizing invasiveness to the traction inverter while reducing implementation costs due to fewer sensor requirements. By leveraging the low-dynamic features of discharge operation, data acquisition remains robust under relaxed sampling conditions and immune to parasitic parameter interference, further reducing monitoring complexity and uncertainty. To enable effective decision-making in field operation, the operating and environmental variables affecting the acquired on-state voltage combinations are identified, and a degradation model-based data calibration scheme is provided to process historical raw data. Based on this monitoring scheme, proactive maintenance for the power module in traction inverters can be scheduled prior to failure, thereby enhancing system availability and reducing maintenance costs. Proof-of-concept experiments are conducted to verify its effectiveness and feasibility for practical application.
- 9 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (9) , 16333-16352
Discharge Profile-Based Converter-Level Health Monitoring for Power Semiconductor Module in Traction Inverter Application
This paper proposes a converter-level health monitoring scheme for power semiconductor modules in traction inverter applications, implemented via the standard inverter DC-link discharge operation mode after system shutdown. The on-state voltage information of all power devices in the inverter is acquired in combined forms from the inverter three-phase output terminals, minimizing invasiveness to the traction inverter while reducing implementation costs due to fewer sensor requirements. By leveraging the low-dynamic features of discharge operation, data acquisition remains robust under relaxed sampling conditions and immune to parasitic parameter interference, further reducing monitoring complexity and uncertainty. To enable effective decision-making in field operation, the operating and environmental variables affecting the acquired on-state voltage combinations are identified, and a degradation model-based data calibration scheme is provided to process historical raw data. Based on this monitoring scheme, proactive maintenance for the power module in traction inverters can be scheduled prior to failure, thereby enhancing system availability and reducing maintenance costs. Proof-of-concept experiments are conducted to verify its effectiveness and feasibility for practical application.
- 22 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (6) , 9513-9527
A Converter-Level Health Monitoring Method for Traction Inverter Application
This article proposes a converter-level health monitoring method for detecting critical degradation in either dc-link capacitors or power semiconductor modules in traction inverters. The outcome can be used for condition-based maintenance decisions to determine whether the inverter or its power stage as a whole needs to be replaced. The proposed method is based on the dc-link voltage profile during the capacitor discharge operation mode after a traction system is already shut down, which is a standard operation mode for safety reasons. Therefore, the method relies only on existing resources from traction inverters without the need for additional hardware or software. The theoretical foundation of the method and practical considerations for implementation are discussed. Proof-of-concept experiments are performed to validate its effectiveness and feasibility for practical application.
- 22 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (6) , 9513-9527
A Converter-Level Health Monitoring Method for Traction Inverter Application
This article proposes a converter-level health monitoring method for detecting critical degradation in either dc-link capacitors or power semiconductor modules in traction inverters. The outcome can be used for condition-based maintenance decisions to determine whether the inverter or its power stage as a whole needs to be replaced. The proposed method is based on the dc-link voltage profile during the capacitor discharge operation mode after a traction system is already shut down, which is a standard operation mode for safety reasons. Therefore, the method relies only on existing resources from traction inverters without the need for additional hardware or software. The theoretical foundation of the method and practical considerations for implementation are discussed. Proof-of-concept experiments are performed to validate its effectiveness and feasibility for practical application.
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (5) , 7830-7846
Integrated Short-Circuit Protection Method in On-State Voltage Measurement Circuit for SiC MOSFETs
This article presents an integrated approach for condition monitoring (CM) and short-circuit (SC) protection in silicon carbide (SiC) mosfets. This design integrates CM and SC protection functionalities within a single circuit by employing an improved on-state voltage measurement circuit, thereby enhancing system compactness. The on-state voltage measurement circuit is specifically configured to interface with the desaturation pin of standard gate driver ICs, which enables effective SC fault protection without significantly increasing circuit complexity or component count. By leveraging shared components, this integrated method simplifies the circuit architecture, reduces space requirements, and mitigates the noise interference or measurement error issues common in traditional integration methods. Simulation and experimental results demonstrate that the proposed method achieves accurate on-state voltage measurements and reliable SC protection under various operating conditions. Therefore, this integrated method has the potential to improve the safety and durability of power electronic systems, particularly in high-frequency SiC mosfet applications.
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (5) , 7830-7846
Integrated Short-Circuit Protection Method in On-State Voltage Measurement Circuit for SiC MOSFETs
This article presents an integrated approach for condition monitoring (CM) and short-circuit (SC) protection in silicon carbide (SiC) mosfets. This design integrates CM and SC protection functionalities within a single circuit by employing an improved on-state voltage measurement circuit, thereby enhancing system compactness. The on-state voltage measurement circuit is specifically configured to interface with the desaturation pin of standard gate driver ICs, which enables effective SC fault protection without significantly increasing circuit complexity or component count. By leveraging shared components, this integrated method simplifies the circuit architecture, reduces space requirements, and mitigates the noise interference or measurement error issues common in traditional integration methods. Simulation and experimental results demonstrate that the proposed method achieves accurate on-state voltage measurements and reliable SC protection under various operating conditions. Therefore, this integrated method has the potential to improve the safety and durability of power electronic systems, particularly in high-frequency SiC mosfet applications.
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (5) , 8392-8408
A Robust Health Monitoring Method for DC-Link Capacitor Banks in Aerospace Motor Drives
This article presents a health monitoring scheme for dc-link capacitor banks in aerospace motor drives, implemented through the standard postlanding discharge operation of the motor drive. The capacitance percentage of the dc-link capacitor bank relative to its healthy baseline is directly evaluated using solely the dc-link voltage profile without additional hardware upgrades or complex computations, minimizing cost and complexity concerns. Furthermore, to ensure monitoring reliability and effectiveness, practical implementation solutions are provided to address challenges arising from varying operating conditions and unbalanced capacitor degradation within the bank. Compared with existing methods, the proposed scheme offers superior practical deployability and operational robustness. Experimental validation through component-level accelerated aging tests and converter-level discharge tests demonstrates capacitance percentage evaluation error within 0.6%, and reliable detection of critical degradation in individual capacitors within the dc-link capacitor bank under varying operating conditions, enabling proactive maintenance before failure occurs.
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (5) , 8392-8408
A Robust Health Monitoring Method for DC-Link Capacitor Banks in Aerospace Motor Drives
This article presents a health monitoring scheme for dc-link capacitor banks in aerospace motor drives, implemented through the standard postlanding discharge operation of the motor drive. The capacitance percentage of the dc-link capacitor bank relative to its healthy baseline is directly evaluated using solely the dc-link voltage profile without additional hardware upgrades or complex computations, minimizing cost and complexity concerns. Furthermore, to ensure monitoring reliability and effectiveness, practical implementation solutions are provided to address challenges arising from varying operating conditions and unbalanced capacitor degradation within the bank. Compared with existing methods, the proposed scheme offers superior practical deployability and operational robustness. Experimental validation through component-level accelerated aging tests and converter-level discharge tests demonstrates capacitance percentage evaluation error within 0.6%, and reliable detection of critical degradation in individual capacitors within the dc-link capacitor bank under varying operating conditions, enabling proactive maintenance before failure occurs.
- 13 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (4) , 4472-4485
Adaptive Current Threshold for Rapid Short-Circuit Protection of SiC MOSFETs
Due to the limited short-circuit (SC) withstand time of silicon carbide metal-oxide-semiconductor field-effect transistors (SiC mosfets), their reliable operation requires fast SC protection. This article proposes an ultrafast SC protection method for SiC mosfets. This method leverages the characteristic of the load current to remain continuous throughout the converter operation, enabling the adaptive adjustment of the SC detection threshold. Unlike traditional methods that utilize a fixed threshold to detect fault currents, the proposed adaptive threshold strategy allows for dynamic adjustments according to the load current level, thereby improving the detection speed across all types of SC faults. The circuit structure and design considerations for implementing this strategy are discussed. Experimental results under 450-V multipulse testing and half-bridge inverter operation demonstrate the effectiveness of this SC protection method. The detection speed is significantly improved compared to the traditional current detection method and the desaturation detection method.
- 13 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (4) , 4472-4485
Adaptive Current Threshold for Rapid Short-Circuit Protection of SiC MOSFETs
Due to the limited short-circuit (SC) withstand time of silicon carbide metal-oxide-semiconductor field-effect transistors (SiC mosfets), their reliable operation requires fast SC protection. This article proposes an ultrafast SC protection method for SiC mosfets. This method leverages the characteristic of the load current to remain continuous throughout the converter operation, enabling the adaptive adjustment of the SC detection threshold. Unlike traditional methods that utilize a fixed threshold to detect fault currents, the proposed adaptive threshold strategy allows for dynamic adjustments according to the load current level, thereby improving the detection speed across all types of SC faults. The circuit structure and design considerations for implementing this strategy are discussed. Experimental results under 450-V multipulse testing and half-bridge inverter operation demonstrate the effectiveness of this SC protection method. The detection speed is significantly improved compared to the traditional current detection method and the desaturation detection method.