Open-Switch Fault Diagnosis of Three-Phase PWM Converter Systems for Magnet Power Supply on EAST
Xi Deng, Chenguang Wan, Li Jiang, Ge Gao, Ya Huang
Abstract
The three-phase two-level pulsewidth modulation (PWM) converter operates in rectification state or active inverter state according to the corresponding experimental requirements of Experimental Advanced Superconducting Tokamak. However, the current fault diagnosis methods proposed for PWM converters usually deal with only the rectification state or inverter state. To solve this problem, this article proposes a diagnosis method to detect single and double open-circuit faults under both the rectification state and active inverter state. First, an inception-based neural network diagnosis model is proposed, in which 1 × 1, 3 × 3, 5 × 5, and 7 × 7 convolution and 3 × 3 max-pooling processes are conducted in parallel at the input and the output and stacked together to generate the final output. Second, the model is trained using three-phase alternating current current signals to determine the location of open-switch faults with a large number of simulation data samples and fewer experimental data samples. Finally, the experimental result shows that the proposed model can accurately detect approximately 99.14% of the open-switch faults within 12.83 ms (<3/4 cycle) without an additional sensor. Furthermore, the result shows that the proposed model is robust to load voltage transient conditions and nonunity power factor conditions.