K-Nearest Neighbor Based High Impedance Fault Detection for Radial Feeder

In the process of providing electrical supply to rural areas using radial feeders, typical operating voltage levels of 11KV, are commonly utilized. These systems connect the supply from sub-station to the consumer-end through distribution transformers. However, in such configurations, conductors may come into contact with tree branches or other conducting mediums, leading to a type of fault known as High Impedance Fault (HIF). HIFs are characterized by phase-to-ground faults with ground resistance, resulting in low fault currents that resemble the load current carrying capacity. These faults pose significant danger to living organisms and cannot detected due to their non-linear nature. This paper proposes a model employing a multilayer perceptron artificial neural network (ANN) and K-Nearest Neighbor (KNN) for the detection of HIFs in radial feeders. The proposed KNN method has achieved better results by assigning k = 2 in predicting the fault bus. The proposed model efficacy is demonstrated using the IEEE 13 bus radial distribution test system.