Fault Detection, Categorization, and Localization of Series Faults in Radial Distribution Networks With DGs Using DLANN-Based Method

This research describes a deep learning-based artificial neural network (ANN) approach for the detection, classification, and localization of series faults (open conductor faults) in power distribution networks. The ANN is trained using the basic elements of the voltage and current signals that are recovered at the relaying point as input attributes. After an ANN has been trained, it must be tested for fault scenarios that were not encountered during training. The proposed DLANN-based fault detector is evaluated in MATLAB/Simulink for all types of series faults in distribution lines with variations in fault location and fault inception time of 0.2 seconds, including one conductor open fault, two open conductor faults, and three open conductor faults. A modified IEEE 34-bus test system is used to test the methodology, and two scenarios, one with and one without distributed generation units, are modelled in the MATLAB environment. The detection and classification accuracy for the series faults were both achieved at 100%, and the fault location accuracy for series faults with and without DGs was obtained as 99.7% and 99.5%, respectively.