Detection and Classification of Single Line to Ground Faults in Unbalanced Islanded Microgrids.

The distribution grid of the future will have a variety of generation resources. With the help of such distributed energy resources (DERs), the grid should be able to fragment and coalesce to form microgrids as needed for enhanced resiliency. In such microgrids, the distinction between different events such as faults and topology change/load step can be challenging. Especially for single line to grounds faults in an unbalanced microgrid, where zero sequence components are already present. The proposed method analyses current waveform through time-frequency multi-resolution analysis (MRA) to extract information without processing into sequence components. It then extracts faultfeatures using statistical methods applied to Shannon entropy and signal energy. These features quantify the changes in the current signal while incorporating the changing physics of the system. To make the method agnostic to the DER mix and topology changes, rather than using set thresholds, it implements machine learning based algorithm using Support Vector Machines (SVM) to detect and classify fault types. The method is tested against a variety of events simulated in MATLAB Simulink.