The paper proposes a novel fault detection and classification scheme for EHV power transmission lines using genetic algorithm-based neural networks. The application concerned is fault classification for EHV lines with a unified power factor corrector (UPFC), since fault classification is a key part of protective relaying schemes. After the genetic algorithm-based neural network is briefly discussed in general, EMTP based digital simulation results of a UPFC transmission system are presented. The generation of training/test data and preprocessing of these data for neural networks are then described. The paper places special emphasis on the performance comparison between a genetic algorithm-based neural network and a backpropagation network-based scheme.