Voltage and Current Sensor Fault Diagnosis Method for Traction Converter with Two Stator Current Sensors

Hongwei Tao, Tao Peng, Chao Yang, Jinqiu Gao, Chunhua Yang, Weihua Gui

TL;DR

This paper presents a novel fault diagnosis method for sensors in high-speed train traction converters, specifically addressing incipient faults in grid current, DC-link voltage, and stator current sensors. By leveraging the DC-link voltage residual and its differential, the method effectively detects, locates, and identifies the type of sensor faults, even with only two stator current sensors, enhancing train safety and reliability.

What Problem Does This Paper Address?

High-speed trains rely heavily on traction converters, which are critical components for their operation. However, faults within these converters, particularly sensor faults, can lead to inaccurate feedback signals, degrading system performance and potentially causing cascading failures or safety hazards. Existing fault diagnosis methods often focus on power device failures or require multiple sensors for current measurement, which can be costly. Furthermore, many methods struggle to distinguish between different types of sensor faults (e.g., offset vs. scaling) or are limited to specific components. This paper tackles the challenge of developing a unified, cost-effective, and accurate fault diagnosis system for incipient sensor faults in both the rectifier and inverter sections of a traction converter, using a minimal sensor configuration (two stator current sensors).

Methods

The proposed method centers around analyzing the residual of the DC-link voltage.

  1. DC-Link Model: A DC-link model is established based on Kirchhoff's current law, relating the DC-link voltage to the currents flowing through the converter legs.
  2. Fault Detection: The core idea is to compare the measured DC-link voltage with an estimated value derived from the DC-link model and measured currents. The difference, or residual, is used for fault detection. To mitigate estimation and accumulation errors, a reconstructed residual is employed. If this reconstructed residual exceeds a predefined threshold ($h_1$), a fault is detected.
  3. Fault Location: The differential of the reconstructed DC-link voltage residual is then analyzed. The behavior of this differential, particularly its change in magnitude and its relationship with the converter's command signals (PWM signals), is used to pinpoint the faulty sensor.
  4. Fault Type Identification: For current sensors, the method further differentiates between offset and scaling faults. This is achieved by observing whether the differential of the residual changes periodically or remains relatively constant, and by analyzing its behavior in relation to specific combinations of command signals.

The method is designed to handle incipient faults, meaning it can detect subtle deviations before they escalate into major issues. It also accounts for the interdependence between rectifier and inverter components to reduce false alarms.

Datasets

The paper does not explicitly mention training on specific datasets. Instead, it relies on a hardware-in-the-loop (HIL) test platform to evaluate the proposed method. This platform emulates the traction converter and motor, allowing for controlled injection of various sensor faults under realistic operating conditions.

Key Insights and Ideas

  • Unified Diagnosis: The method provides a single framework for diagnosing faults in grid current, DC-link voltage, and stator current sensors, simplifying the diagnostic process.
  • Minimal Sensor Requirement: The ability to diagnose stator current sensor faults with only two sensors is a significant cost-saving advantage for practical applications.
  • Incipient Fault Detection: Focusing on incipient faults is crucial for preventing minor issues from escalating into severe system failures, thereby enhancing overall safety.
  • DC-Link Residual as a Universal Indicator: The DC-link voltage residual proves to be a versatile signal that can be influenced by faults in multiple sensors, making it an effective starting point for diagnosis.
  • Command Signals as Diagnostic Tools: The clever use of PWM command signals in conjunction with the residual differential allows for precise fault localization and type identification.

Key Results

The paper demonstrates the effectiveness of the proposed method through extensive HIL tests.

  • Robustness to Operating Conditions: The method showed robustness to changes in train speed (200 km/h to 300 km/h) and load torque, with no misdiagnoses observed.
  • Accurate Detection, Location, and Type Identification: The HIL results confirm that the method can accurately detect, locate, and distinguish between offset and scaling faults for all considered sensors (grid current, DC-link voltage, and stator currents).
  • Timely Diagnosis: Faults were diagnosed within a short time frame, enabling prompt intervention. For instance, in the case of a DC-link voltage sensor fault, detection occurred at $t_2$ and location at $t_3$ after the fault occurrence at $t_1$.
  • Validation Across Fault Types: The paper presents detailed HIL results for offset and scaling faults in the DC-link voltage sensor, grid current sensor, and stator current sensors, validating the method's comprehensive coverage.

Limitations and Future Work

  • Threshold Determination: The paper mentions that thresholds ($h_1$ and $h_2$) are obtained through extensive testing and represent a trade-off between detection speed and accuracy. Further research could explore adaptive or self-tuning threshold mechanisms.
  • Computational Load: While the method is designed for real-time application, the complexity of calculations, especially for the differential and its analysis, might be a consideration for extremely resource-constrained embedded systems.
  • Interdependence of Command Signals: The discussion on command signals highlights that their differences are crucial. While the paper states these differences are generally present, extreme scenarios where they are consistently similar could potentially impact diagnosis.
  • Future Work: The authors suggest that the proposed method is suitable for traction converters. Future work could involve extending this methodology to more complex converter topologies or investigating its application in other high-power systems. Further exploration into the impact of different PWM strategies on the diagnostic performance could also be beneficial.

Summary updated

Abstract

The traction converter is one of the key components of high-speed trains. Current and voltage sensor faults in the converter may lead to feedback values deviation and system degradation, which will bring security risks to the train. This paper proposes a real-time fault diagnosis method for grid current, DC-link voltage and stator current sensor faults in the traction converter with two stator current sensors, which can not only detect and locate faults but also identify the types of faults. Moreover, the faults considered in this paper are incipient. First, the DC-link model is established, and the fault is detected by the residual of the DC-link voltage. Next, the differential of DC-link voltage residual is calculated, which is applied to fault location. Then, according to the change of the differential values, different fault types are determined. Finally, the hardware-in-the-loop (HIL) platform is built and the effectiveness and accuracy of the proposed method are verified by the HIL tests.
BibTeX
@inproceedings{tao2022voltage,
  author = {Hongwei Tao and Tao Peng and Chao Yang and Jinqiu Gao and Chunhua Yang and Weihua Gui},
  title = {{Voltage and Current Sensor Fault Diagnosis Method for Traction Converter with Two Stator Current Sensors}},
  booktitle = {Sensors},
  year = {2022},
  url = {https://openreview.net/forum?id=HEc216kGIR},
}
Figure