Fault localization on transformer winding by frequency response analysis and evidential reasoning

Abstract

In this paper, a novel fault localization method combining frequency response analysis (FRA) and evidential reasoning process is proposed. FRA test is conducted at first so that we can obtain corresponding responses of several monitoring nodes on the winding. Traditional FRA interpretation methods simply compares those response curves, while in our work, evidential reasoning method is adopted to present beliefs about the fault locations. Extracted features from monitoring sensors are mapped into deviation matrix by which possibility assignments of each fault location is computed. So we finally obtain fault distribution probabilities. According to test calculation results, proposed method can give fault probabilities of each section of transformer winding, which correctly indicate the fault position.

Keywords

• fault diagnosis
• transformer windings
• matrix algebra
• fault location
• frequency response
• probability
• case-based reasoning
• power engineering computing
• feature extraction
• fault probabilities
• transformer winding
• fault position
• frequency response analysis
• fra test
• corresponding responses
• monitoring nodes
• traditional fra interpretation methods
• response curves
• evidential reasoning method
• fault location
• monitoring sensors
• fault distribution probabilities
• test calculation results
• fault localization method