Protection Scheme Based on k-Nearest Neighbour Algorithm for Fault Detection Classification and Section Identification in DC Microgrid

摘要

The rapid increase in the power requirement along with the stress on reducing the dependence on fossil fuels has propelled the integration of green sources of energy such as PV and wind energy. The increase in the use of DC power-based devices along with the ease of integrating renewable sources has led to significant interest in DC microgrids. However, the absence of zero crossing, bi-directional current flow and the dependence of the fault current magnitude on the operating modes (islanded and grid connected) posses significant challenges in designing a reliable protection scheme for DC microgrid. In this regard, a k-nearest neighbour (kNN)-based scheme has been proposed in the present work to perform the task of mode detection, fault detection/classification and section identification in DC microgrid. The algorithm does not involve a module for extracting features from the post-fault time waveforms, thereby leading to faster execution of the protection tasks. The scheme has been extensively validated for varying fault scenarios in terms of accuracy and computational cost.