k-NN Algorithm Based Approach for the Detection of Faulty Sections in Underground Distribution Network

The occurrence of faults in electrical distribution networks is a significant concern due to the potential damage to electrical equipment, system instability, and the disruption of reliable energy. Detecting faults in a timely and accurate manner is crucial, especially in complex underground distribution networks with branches, non-homogeneous cables, and various loads. The inherent complexity of such networks poses challenges in pinpointing faulty sections, necessitating specialized detection methods. This research focuses on utilizing the k-Nearest Neighbors (k-NN) algorithm for detecting faulty sections in an underground distribution network. Practical data, obtained from Tenaga Nasional Berhad Malaysia (TNB), including measurements of current swell and voltage sags for 17 network sections, was used to calculate the Euclidean Distance. Given that 70% of faults in the distribution network are attributed to Single Line to Ground Fault (SLGF), this study specifically targets this fault type. Additionally, various fault resistances were tested to observe the k-NN algorithm's performance. The results indicate that the k-NN algorithm successfully detected faulty sections, demonstrating effectiveness across different fault resistances and ranks. This research contributes valuable insights into improving fault detection mechanisms in underground distribution networks.