Fault diagnosis for three-phase PWM rectifier based on deep feedforward network with transient synthetic features

Area of Science:

• Electrical Engineering
• Power Electronics
• Fault Diagnosis Systems

Background:

• Three-phase PWM rectifiers are crucial in industry but susceptible to Insulated Gate Bipolar Transistor (IGBT) open-circuit faults.
• Such faults degrade system performance through voltage fluctuations and current harmonics without immediate system failure.

Purpose of the Study:

• To develop a fault diagnosis method for IGBT open-circuit faults in three-phase PWM rectifiers.
• To reduce reliance on complex fault mathematical models by utilizing transient phase current data.

Main Methods:

• Feature analysis of fault phase current under open-circuit conditions.
• Training a deep feedforward network classifier using synthesized historical fault data.
• Online experimental validation of the proposed diagnosis method.

Main Results:

• The proposed method achieved an average fault diagnosis accuracy of 97.85% using transient synthetic fault data.
• Classifiers trained with synthetic features showed over 1% performance improvement compared to those trained with original transient features.
• Online experiments confirmed accurate localization of faulty IGBTs.

Conclusions:

• The developed deep feedforward network method effectively diagnoses IGBT open-circuit faults in three-phase PWM rectifiers.
• Employing synthetic transient features enhances diagnosis accuracy and reliability.
• The multi-group result determination strategy further boosts the robustness of the fault diagnosis system.