Characteristics of Partial Discharge of Typical Defects in Converter Transformer Under Operation

Converter transformers, operating in complex electric fields for extended periods, are prone to insulation degradation. To investigate partial discharge (PD) characteristics caused by internal insulation deterioration, this study developed an AC/DC PD simulation platform replicating real-world operating conditions. Four typical insulation defects—turn-to-turn, spacer block, electrostatic ring, and tap changer—were modeled, with experiments conducted in a 35 kV transformer tank using pure AC and composite AC/DC voltages (1:1, 1:3, 1:5, 1:7 ratios). The 48 Phase-Resolved Partial Discharge (PRPD) spectra were processed through image cropping, grayscale conversion, and dimension standardization, with 84% training and 16% testing datasets. Convolutional neural networks (CNNs) achieved an overall classification accuracy of 93.7% on the test set. Field validation using ±400 kV real-world converters confirmed the network’s precision in PD pattern recognition, providing technical support for insulation condition diagnosis and fault prediction in converter transformers.