Transformer Winding Fault Classification and Condition Assessment Based on Random Forest Using FRA

At present, the condition assessment of transformer winding based on frequency response analysis (FRA) measurements demands skilled personnel. Despite many research efforts in last decade, there is still no definitive methodology for interpretation and FRA results, this a major challenge industrial application method. To overcome challenge, paper proposes (TCA) algorithm, which numerical indices, supervised machine learning technique to develop method automatic results. For purpose, random forest (RF) classifiers were developed first time identify classify different faults windings. Mainly, six common states classified research, i.e., healthy transformer, with saturated core, mechanically damaged winding, short-circuited open-circuited repeatability issues. In data from 139 performed more than 80 power transformers used. The database belongs having ratings, sizes, designs, manufacturers. results reveal that proposed TCA algorithm can effectively assess up 93% accuracy without much human intervention.