A Sugeno-Type ANFIS approach for fast and accurate fault diagnosis in a distribution system

Power consumers are always expecting quality and continuous electric supply from utilities in order to prevent financial losses. Therefore, utilities have to provide high quality and reliableelectric power supply. For this purpose, an accurate and fast fault diagnosis in distribution systems should be provided to reduce the customer average interruption duration index. This paper presents an intelligent technique for fast and accurate fault diagnosis of power distribution systems. The fault diagnosis functions include correct fault type classification, precise fault location, accurate identification of faulty protection devices and current operation status of protection devices in a distribution system. To determine fault types and fault points, the proposed intelligent technique uses post-fault three-phase root-mean-square currents to train the Sugeno-type parallel-series adaptive neuro-fuzzy inference system (ANFIS). Meanwhile, the current operating status of protection devices are presented as binary ones or zeros by ANFIS prediction using fault point geometrical coordinates. The simulation results show that the technique has an average maximum percentage error of 0.02% for predicting the type of faults and 2.8% in determining the fault points. Average maximum errors of 0.053 and 0.182 were obtained for operating status of main and back up protection devices, respectively. The proposed technique is validated through simulations using the PSS-ADEPT commercial software package. The results showed thatthe Sugeno-type ANFIS approach provides fast and precise fault diagnosis that can be implemented in a distribution network.