ANFIS approach for locating precise fault points with coordinated geometries in a test distribution system

To maintain high reliability and quality in electrical power delivery, it is essential for utilities to accurately locate fault points and identify the type of fault so as to achieve a low customer average interruption duration index. Unless fault locations are known accurately, power cannot be promptly restored. This paper presents an intelligent technique for locating fault points in term of their geometrical coordinates in a power distribution system. The proposed technique implements parallel-series adaptive neuro-fuzzy inference systems for the purpose of locating fault points and identifying the fault type. The inputs of the proposed system design are the post-fault three-phase root-mean-square currents, and the output provides information about the fault points and types of faults. The technique produces satisfactory results with an average maximum percentage error of 1.3% and 2.8% respectively for fault X, Y coordinates prediction and 0.02% for fault type identification. The proposed technique is validated through simulations using the commercial software package PSS-ADEPT. The results show that parallel-series ANFIS approach is a fast and effective technique to determine precise fault location in a distribution system.