AN APPROACH TO POWER DISTRIBUTION FAULT-DIAGNOSIS USING A NEURAL-NET BASED SUPERVISED CLUSTERING METHODOLOGY

This paper discusses the use of a new neural network supervised cluste ring method to perform fault diagnosis for power distribution networks . The neural network performs fault type classification, faulted feede r and faulted phase identification, and fault impedance estimation for grounded and ungrounded distribution networks. These fault diagnosis decisions are performed on a feeder by feeder basis. The theoretical d evelopment of the neural net is presented. Also the effectiveness of t he method is demonstrated using data generated by a fault simulation p rogram specially designed to facilitate sensitivity studies of distrib ution systems. This work represents the first time a supervised cluste ring neural net has been used for distribution fault diagnosis.