Differential Fault Detection Scheme for Islanded AC Microgrids Using Digital Signal Processing and Machine Learning Techniques

The conventional overcurrent-based fault detection schemes are not applicable to islanded microgrids due to their dependence on significant fault current transients. This paper introduces a differential protection-based fault detection, localization, and classification method by using Digital Signal Processing and machine learning techniques. The proposed method is capable of detecting any type of line faults in the system and provides information on the fault location and type of fault including faulty phase information. In this method, differential current and voltage signals from both the ends of a line are sampled processed using Fast Fourier Transform to extract the features. Two different classification models for fault detection and classification and one regression model for fault localization are developed using Support Vector Machine technique. The proposed method is extensively validated in an IEC-based and IEEE 34-bus based islanded microgrid models. Further, the performance of the proposed method is verified with clean as well as noisy data, where the noise is extracted from real system measurements. The extensive test results indicate that the proposed method is highly reliable in providing an effective protection measure for safe and secure microgrid operation.