Comparative Analysis of Partial Discharge Pattern Recognition Using Deep Learning and Machine Learning

Partial Discharge (PD) defect type analysis is important for evaluating insulation performance. A machine learning feature extraction algorithm is presented for AC PD pattern data collected in the laboratory, along with deep learning algorithms for PD pattern images and PD time series data. In addition, data is collected under conditions different from those used for artificial intelligence (AI) training, and algorithm performance is evaluated and compared.