Abstract: Fault location is one of the important tasks of automated distribution systems. In this research, fast learning Radial Basis Function Neural networks (RBFNs) were employed to automatically locate the faults in distribution networks. The radial basis networks are simpler in structure, faster and more efficient than the conventional multilayer feed forward networks. These are functionally considered equivalent to more successful fuzzy connectionist hybrid models. An IEEE test distribution system was used for analyzing the potential and accuracy ofthese networks in the estimation of fault location information. The distribution network wassimulated and tested in MATLAB/SIMULINK. Three fundamental tasks of this RBFN Models, fault type classification, faulted line section detection and pin pointing of fault location on the faulted line were executed by multiple RBFN Models which were designed in MATLAB environment. All the required fault data for training and testing of the models was generated by triggering various fault scenarios on the simulated distribution network. The test results obtained demonstrate good degree of accuracy. This vital fault location information supplied by RBFNs can greatly support the search efforts of distribution substation repair crew in quickly pin pointing the faulty spot and restoring the power to the affected customers. This reduces the customer service interruption duration and thus contributes in enhancing the power system reliability and quality.
Surender Kumar Yellagoud, Purnachandra Rao Talluri and Gondlala N. Sreenivas, 2017. An Application of Fast Learning Radial Basis Function Networks for an Accurate Estimation of Fault Location in Electrical Distribution Networks. International Journal of Soft Computing, 12: 72-78.