Comparative Analysis of Machine Learning Algorithms for Eccentricity Fault Classification in Salient Pole Synchronous Machine

Abstract

The paper performs a comparative study of Artificial Neural Networks (ANNs) and Convolutional Neural Networks (CNNs) for the classification of Static Eccentricity (SE) and Dynamic Eccentricity (DE) faults in a Salient Pole Synchronous Machine (SPSM). The SPSM was subjected to varying SE and DE severities, unbalanced source voltages, and load conditions. Stator and field current data were measured, and various time-domain and frequency-domain features were extracted from the above-mentioned data. Both networks were fed these features and compared based on classification accuracy, robustness, and computational complexity.