A Hybrid Method of Self Organizing Maps with Statistical Feature Extraction for Accurate and Efficient Partial Discharge Recognition and Clustering

摘要

Partial discharge is the phenomena that affecting the health of power transformer. The problem with delay in identifying will deteriorate the transformer insulation condition and ultimately reduced the network security and reliability. In this paper, author proposed a hybrid method combining pinnacle statistical features with self organizing method for partial discharge recognition and clustering to replace the conventional way. Overall, the proposed method achieved decent clustering result with fast computation time (less than 10 seconds)